2018-11-17 Sandra Loosemore <sandra@codesourcery.com>
[official-gcc.git] / gcc / ChangeLog
blob7362f269c1cca6c0fbc98ce38cc630276925134d
1 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
3         PR c++/4025
4         * doc/invoke.texi (C++ Dialect Options): Clarify usage of
5         -fno-implicit-templates.
7 2018-11-17  Jakub Jelinek  <jakub@redhat.com>
9         PR tree-optimization/87546
10         * tree-vect-patterns.c (vect_look_through_possible_promotion): Add
11         min_precision variable, initially set it to orig_precision, only does
12         something if op_type's precision is <= min_precision and update
13         min_precision whenever calling set_op.
15 2018-11-16  Jan Hubicka  <hubicka@ucw.cz>
17         PR ipa/87957
18         * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL.
20 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
22         * doc/invoke.texi (Option Summary): Fix whitespace and line
23         breaks in @gccoptlist environments.
24         (Warning Options): Likewise.
25         (Optimize Options): Likewise.
26         (PowerPC SPE Options): Likewise.
27         (RS/6000 and PowerPC Options): Likewise.
29 2018-11-16  Jeff Law  <law@redhat.com>
31         * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus
32         gcc_assert.
34 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
36         PR middle-end/23197
37         * doc/invoke.texi (Optimize Options): Update options enabled by
38         fprofile-generate, -fprofile-use, and -fauto-profile.
40 2018-11-16  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
42         PR target/87927
43         * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
44         Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
45         * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
46         the byte_op hook.
47         * target.h: Add psi, pdi, pti to struct asm_int_op definition.
48         * targhooks.c (default_print_patchable_function_entry): Assert
49         asm_int_op does not return a NULL string.
50         * varasm.c (integer_asm_op): Return the op for a partial int type
51         when the requested size does not correspond to an integer type.
52         * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
53         * doc/tm.texi: Regenerate.
55 2018-11-12  Jason Merrill  <jason@redhat.com>
57         * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
59 2018-11-16  Michael Meissner  <meissner@linux.ibm.com>
61         * config/rs6000/constraints.md (wF constraint): Remove power9
62         fusion documentation.  Just document wF's use for power8 fusion.
63         * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
64         (fusion_gpr_addis): Delete power9 fusion support.  Change power8
65         fusion support to require the upper 12 bits to be all 0's or all
66         1's.
67         (fusion_gpr_mem_load): Add comment.
68         (fusion_addis_mem_combo_load): Remove power9 fusion support.
69         (fusion_addis_mem_combo_store): Delete.
70         (fusion_offsettable_mem_operand): Delete.
71         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
72         power8 fusion here.
73         (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
74         (POWERPC_MASKS): Delete power9 fusion.
75         * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
76         (fusion_p9_p): Delete.
77         (expand_fusion_p9_load): Delete.
78         (expand_fusion_p9_store): Delete.
79         (emit_fusion_p9_load): Delete.
80         (emit_fusion_p9_store): Delete.
81         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
82         fusion support.
83         (rs6000_option_override_internal): Set power8 fusion based on
84         whether we are tuning for power8.  Delete power9 fusion support.
85         (rs6000_opt_masks): Delete -mpower9-fusion switch.
86         (emit_fusion_load): Rename emit_fusion_load_store to
87         emit_fusion_load, and drop fusion store support.  Update callers.
88         (emit_fusion_load_store): Likewise.
89         (emit_fusion_gpr_load): Likewise.
90         (fusion_p9_p): Delete.
91         (expand_fusion_p9_load): Delete.
92         (expand_fusion_p9_store): Delete.
93         (emit_fusion_p9_load): Delete.
94         (emit_fusion_p9_store): Delete.
95         * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete.
96         (GPR_FUSION): Delete.
97         (FPR_FUSION): Delete.
98         (power9 fusion peephole2s): Delete.
99         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
100         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
101         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
102         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
103         (fusion_p9_<mode>_constant): Delete.
104         * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
105         power9 fusion switch.
106         * doc/md.texi (PowerPC constraints): Update wF constraint
107         documentation for power8 fusion only.
109 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
111         PR rtl-optimization/87475
112         * cfgrtl.c (patch_jump_insn): Allow redirection failure for
113         CROSSING_JUMP_P insns.
114         (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
116 2018-11-16  Uros Bizjak  <ubizjak@gmail.com>
118         PR target/88051
119         * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
120         * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
121         (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
123 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
125         PR middle-end/88032
126         * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
127         is VOIDmode, in which case use int_mode.  Similarly for op1_mode.
129 2018-11-16  Eric Botcazou  <ebotcazou@adacore.com>
131         * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
132         location_t parameter.
133         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
134         (switch_decision_tree::do_jump_if_equal): Likewise.
135         * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
136         of switch statement to emit_case_nodes.
137         (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
138         set it on the newly built GIMPLE comparison statement.
139         (switch_decision_tree::do_jump_if_equal): Likewise.
140         (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
141         in calls to do_jump_if_equal as well as recursive calls.
143 2018-11-16  Andi Kleen  <ak@linux.intel.com>
145         * config/i386/i386.md: Allow memory operands to ptwrite.
147 2018-11-16  Michael Matz  <matz@suse.de>
149         * system.h (PRsa): New macro.
150         (SIZE_AMOUNT): Cast number to uint64_t.
151         * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
152         (pool_usage::dump_footer): Likewise and also use PRIu64.
153         * bitmap.h (bitmap_usage::dump): Likewise.
154         * ggc-common.c (ggc_usage::dump): Likewise.
155         * ggc-page.c (ggc_print_statistics): Likewise.
156         * input.c (dump_line_table_statistics): Likewise.
157         * mem-stats.h (mem_usage::dump): Likewise.
158         (mem_usage::dump_footer): Likewise.
159         * rtl.c (dump_rtx_statistics): Likewise.
160         * tree-cfg.c (dump_cfg_stats): Likewise.
161         * tree-dfa.c (dump_dfa_stats): Likewise.
162         * tree-phinodes.c (phinodes_print_statistics): Likewise.
163         * tree-ssanames (ssanames_print_statistics): Likewise.
164         * vec.c (vec_usage::dump): Likewise.
165         (vec_usage::dump_footer): Likewise.
167 2018-11-16  Richard Biener  <rguenther@suse.de>
169         PR tree-optimization/88011
170         * tree-vrp.c (extract_range_from_binary_expr): Fix error in
171         replacing set_value_range_to_undefined and
172         set_value_range_to_varying with method calls.
174 2018-11-16  Ilya Leoshkevich  <iii@linux.ibm.com>
176         * config/s390/s390.md
177         (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
178         assembler.
179         (*r<noxa>sbg_<mode>_sll): Likewise.
180         (*r<noxa>sbg_<mode>_srl): Likewise.
181         (*r<noxa>sbg_sidi_srl): New pattern.
183 2018-11-16  Jerome Lambourg  <lambourg@adacore.com>
185         * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
186         Default to TARGET_VXWORKS_RTP.
187         (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
188         of TARGET_VXWORKS_RTP.
189         * config/vxworksae.h: Also define TARGET_VXWORKS_HAVE_CTORS_DTORS.
190         * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
191         of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
193 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
195         PR c++/25759
196         * doc/extend.texi (Common Type Attributes): Make it explicit
197         that attribute "packed" can apply to C++ classes.
199 2018-11-15  Martin Sebor  <msebor@redhat.com>
201         PR c++/87541
202         PR c++/87542
203         * tree.c (type_argument_type): New function.
204         * tree.h (type_argument_type): Declare it.
205         * gcc/doc/extend.texi (alloc_align): Update and clarify.
206         (alloc_size, nonnull, sentinel): Same.
208 2018-11-15  Andrew Stubbs  <ams@codesourcery.com>
209             Kwok Cheung Yeung  <kcy@codesourcery.com>
211         * tree-vect-stmts.c (vectorizable_store): Don't ICE when
212         int_mode_for_size fails.
213         (vectorizable_load): Likewise.
215 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
217         * doc/ux.texi (Group logically-related diagnostics): Move
218         discussion of auto_diagnostic_group into this new subsection.
219         Give an example of where this grouping is used.
221 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
223         PR other/19165
224         * Makefile.in (OBJS): Move json.o to...
225         (OBJS-libcommon): ...here and add diagnostic-format-json.o.
226         * common.opt (fdiagnostics-format=): New option.
227         (diagnostics_output_format): New enum.
228         * diagnostic-format-json.cc: New file.
229         * diagnostic.c (default_diagnostic_final_cb): New function, taken
230         from start of diagnostic_finish.
231         (diagnostic_initialize): Initialize final_cb to
232         default_diagnostic_final_cb.
233         (diagnostic_finish): Move "being treated as errors" messages to
234         default_diagnostic_final_cb.  Call any final_cb.
235         (default_diagnostic_finalizer): Add diagnostic_t param.
236         (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
237         diagnostic_finalizer callback.
238         * diagnostic.h (enum diagnostics_output_format): New enum.
239         (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
240         (struct diagnostic_context): Add "final_cb".
241         (default_diagnostic_finalizer): Add diagnostic_t param.
242         (diagnostic_output_format_init): New decl.
243         * doc/invoke.texi (-fdiagnostics-format): New option.
244         * dwarf2out.c (gen_producer_string): Ignore
245         OPT_fdiagnostics_format_.
246         * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
247         * lto-wrapper.c (append_diag_options): Ignore it.
248         * opts.c (common_handle_option): Handle it.
250 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
252         PR tree-optimization/88015
253         * graphite-isl-ast-to-gimple.c
254         (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
255         for dump_enabled_p.
256         * graphite-sese-to-poly.c (build_poly_scop): Likewise.
258 2018-11-15  Richard Biener  <rguenther@suse.de>
260         PR middle-end/88029
261         * gimple.c (gimple_call_flags): Union flags from decl, type
262         and call fntype.
263         * trans-mem.c (is_tm_pure_call): Simplify.
265 2018-11-15  Richard Biener  <rguenther@suse.de>
267         PR tree-optimization/88031
268         * tree-vect-loop.c (vectorizable_reduction): Move check
269         for multiple types earlier so we get the expected dump.
270         Simplify calls to vectorizable_condition.
271         * tree-vect-stmts.h (vectorizable_condition): Update prototype.
272         * tree-vect-stmts.c (vectorizable_condition): Instead of
273         reduc_def and reduc_index take just a flag.  Simplify
274         code-generation now that we can rely on the defs being set up.
275         (vectorizable_comparison): Remove unused argument.
277 2018-11-15  Richard Biener  <rguenther@suse.de>
279         PR tree-optimization/88030
280         * tree-complex.c (need_eh_cleanup): New global.
281         (update_complex_assignment): Mark blocks that need EH update.
282         (expand_complex_comparison): Likewise.
283         (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
284         perform EH cleanup and schedule CFG cleanup if that did anything.
286 2018-11-15  Jakub Jelinek  <jakub@redhat.com>
288         PR rtl-optimization/88018
289         * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
290         on the presence of fallthru edge, rather than if it is a USE or not.
292 2018-11-15  Richard Biener  <rguenther@suse.de>
294         PR middle-end/87917
295         * tree-data-ref.c (analyze_miv_subscript): Guard calls to
296         analyze_subscript_affine_affine properly.
298 2018-11-15  Xianmiao Qu  <xianmiao_qu@c-sky.com>
300         * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
301         (GLIBC_DYNAMIC_LINKER): Define.
302         (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
304 2018-11-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
306         PR tree-optimization/84648
307         * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
308         (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
309         by calling adjust_cond_for_loop_until_wrap.
311 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
313         PR other/56334
314         * doc/extend.texi (Common Function Attributes): Clarify linker
315         restrictions on "aligned" attribute.
316         (Common Variable Attributes): Likewise.  Mention that linker
317         restrictions don't apply to stack-allocated variables.
319 2018-11-15  Joern Rennecke  <joern.rennecke@riscy-ip.com>
321         * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
322         as an lvalue.
324 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
326         PR lto/55102
327         PR lto/56700
328         * doc/invoke.texi (Optimize Options): Remove bad example about
329         interaction between -flto and -O.  Replace it with a note that
330         you need to compile with -O and not just link.  Copy-edit -flto
331         discussion to reduce verbiage and improve flow.
333 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
335         PR middle-end/59658
336         * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
337         also suppress many optimizations.  Alphabetize option lists for
338         -O1, -O2, and -Os.  Add list of options disabled with -Og, and
339         correct documentation for those options to say that.
340         * opts.c (default_options_table): Sort table by level and option
341         name, to make it easier to correlate to the manual.
343 2018-11-14  Uros Bizjak  <ubizjak@gmail.com>
345         * config/i386/i386.c (ix86_print_operand_address_as): Simplify
346         printing of the names of segment registers.
348 2018-11-14  Nathan Sidwell  <nathan@acm.org>
350         PR debug/88006
351         PR debug/87462
352         * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
353         type list.
355 2018-11-14  David Malcolm  <dmalcolm@redhat.com>
357         * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
359 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
361         P1236R1 - Signed integers are two's complement
362         * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
364         PR bootstrap/86739
365         * hash-map.h (hash_map::iterator::reference_pair): New class.
366         (hash_map::iterator::operator*): Return it rather than std::pair.
368 2018-11-14  Jeff Law  <law@redhat.com>
370         * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
371         iff the operand is a constant.
373 2018-11-14  Aldy Hernandez  <aldyh@redhat.com>
375         * gimple-ssa-evrp-analyze.c
376         (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
377         ignore_equivs_equal_p to equal_p.
378         * ipa-cp.c (meet_with_1): Use equal_p instead of
379         ignore_equivs_equal_p.
380         * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
381         * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
382         (value_range::operator==): Remove.
383         (value_range::operator!=): Remove.
384         (vrp_prop::visit_stmt): Use equal_p.
385         * tree-vrp.h (value_range): Remove operator==, operator!=,
386         ignore_equivs_equal_p.
387         * vr-values.c (update_value_range): Use equal_p.
389 2018-11-14  Michael Matz  <matz@suse.de>
391         PR middle-end/86575
392         * gimplify.c (collect_fallthrough_labels): Add new argument,
393         return location via that, don't modify statements.
394         (warn_implicit_fallthrough_r): Adjust call, don't use
395         statement location directly.
397 2018-11-14  Richard Biener  <rguenther@suse.de>
399         PR middle-end/87985
400         * tree-data-ref.c (split_constant_offset): Add wrapper
401         allocating a cache hash-map.
402         (split_constant_offset_1): Cache results of expanding
403         expressions from SSA def stmts.
405 2018-11-14  Richard Biener  <rguenther@suse.de>
407         PR middle-end/88021
408         * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
409         to lambda_int.
410         (lambda_vector_mult_const): Likewise.
411         (lambda_matrix_right_hermite): Use lambda_int temporaries.
413 2018-11-14  Wilco Dijkstra  <wdijkstr@arm.com>
414             Jackson Woodruff  <jackson.woodruff@arm.com>
416         PR 71026/tree-optimization
417         * match.pd: Simplify floating point comparisons.
419 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
421         PR rtl-optimization/87817
422         * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
423         last argument has low 8 bits clear, fold to 0.
425 2018-11-14  Iain Sandoe  <iain@sandoe.co.uk>
427         * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
428         MAX_OFILE_ALIGNMENT as needed.
430 2018-11-14  Richard Biener  <rguenther@suse.de>
432         PR tree-optimization/88019
433         * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
434         COND_EXPR arguments to non-trapping overflow.
436 2018-11-14  Claudiu Zissulescu  <claziss@synopsys.com>
438         * config/arc/arc.h (reg_class): Reorder registers classes, remove
439         unused register classes.
440         (REG_CLASS_NAMES): Likewise.
441         (REG_CLASS_CONTENTS): Likewise.
442         (FIXED_REGISTERS): Make lp_count fixed.
443         (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
444         (PROGRAM_COUNTER_REGNO): Remove.
445         * config/arc/arc.c (arc_conditional_register_usage): Remove unused
446         register classes, use constants for register numbers, remove
447         reg_class_contents references.
448         (arc_process_double_reg_moves): Add asserts.
449         (arc_secondary_reload): Remove LPCOUNT_REG reference, use
450         lra_in_progress predicate.
451         (arc_init_reg_tables): Remove unused register classes.
452         (arc_register_move_cost): Likewise.
453         (arc_preferred_reload_class): Likewise.
454         (hwloop_optimize): Update rtx patterns involving lp_count register.
455         (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
456         * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
457         CHEAP_CORE_REGS.  Former one will be used for LRA.
458         ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS.  Former
459         one will be used for LRA.
460         ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS.  Former
461         one will be used for LRA.
462         ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
463         Former one will be used for LRA.
464         ("f"): Delete constraint.
465         ("k"): Likewise.
466         ("e"): Likewise.
467         ("l"): Change it from register constraint to constraint.
468         * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
469         (movhi_insn): Likewise.
470         (movsi_insn): Update pattern.
471         (arc_lp): Likewise.
472         (dbnz): Likewise.
473         (stack_tie): Remove 'b' constraint letter.
474         (R4_REG): Define.
475         (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
476         (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
477         (R57_REG, R59_REG, PCL_REG): Likewise.
478         (ILINK1_REGNUM): Renamed to ILINK1_REG.
479         (ILINK2_REGNUM): Renamed to ILINK2_REG.
480         (Rgp): Remove.
481         (SP_REGS): Likewise.
482         (Rcw): Remove unused reg classes.
483         * config/arc/predicates.md (dest_reg_operand): Just default on
484         register_operand predicate.
485         (mpy_dest_reg_operand): Likewise.
486         (move_dest_operand): Use macros instead of constants.
487         * config/arc/arc.opt (mlra): Switch to lra as default.
489 2018-11-14  Richard Biener  <rguenther@suse.de>
491         PR tree-optimization/87974
492         * tree-vect-loop.c (vectorizable_reduction): When computing
493         the vectorized reduction PHI vector type ignore constant
494         and external defs.
496 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
498         PR tree-optimization/87977
499         * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
500         stmt, build a new one and replace the old one with it.  Formatting fix.
501         Call release_ssa_name (x) if !has_other_use and !delete_div.
502         (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
503         verify lhs of stmt is still def.
505 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
507         PR rtl-optimization/87507
508         * lower-subreg.c (operand_for_swap_move_operator): New function.
509         (simple_move): Strip simple operators.
510         (find_pseudo_copy): Likewise.
511         (resolve_operand_for_swap_move_operator): New function.
512         (resolve_simple_move): Strip simple operators and swap operands.
514 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
516         PR tree-optimization/87898
517         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
518         (ipa_simd_modify_function_body): Remove debug stmts where the first
519         argument was changed into a non-decl.
521 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
523         PR rtl-optimization/87899
524         * lra-lives.c (start_living): Update white space in comment.
525         (enum point_type): New.
526         (sparseset_contains_pseudos_p): New function.
527         (update_pseudo_point): Likewise.
528         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
529         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
530         handling.  Move early exit after adding conflicts.
531         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
532         if regno is already live.  Remove all handling of program points.
533         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
534         after adding conflicts.  Remove all handling of program points and
535         ignore_reg_for_conflicts.
536         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
537         and do not guard call to mark_pseudo_live.
538         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
539         and do not guard call to mark_pseudo_dead.
540         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
541         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
542         Use new function update_pseudo_point.  Handle register copies by
543         removing the source register from the live set.  Handle INOUT operands.
544         Update to the next program point using the unused_set, dead_set and
545         start_dying sets.
546         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
548 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
550         * builtins.c: Replace "source_location" with "location_t".
551         * diagnostic-show-locus.c: Likewise.
552         * diagnostic.c: Likewise.
553         * dumpfile.c: Likewise.
554         * gcc-rich-location.h: Likewise.
555         * genmatch.c: Likewise.
556         * gimple.h: Likewise.
557         * gimplify.c: Likewise.
558         * input.c: Likewise.
559         * input.h: Likewise.  Eliminate the typedef.
560         * omp-expand.c: Likewise.
561         * selftest.h: Likewise.
562         * substring-locations.h (get_source_location_for_substring):
563         Rename to...
564         (get_location_within_string): ...this.
565         * tree-cfg.c: Replace "source_location" with "location_t".
566         * tree-cfgcleanup.c: Likewise.
567         * tree-diagnostic.c: Likewise.
568         * tree-into-ssa.c: Likewise.
569         * tree-outof-ssa.c: Likewise.
570         * tree-parloops.c: Likewise.
571         * tree-phinodes.c: Likewise.
572         * tree-phinodes.h: Likewise.
573         * tree-ssa-loop-ivopts.c: Likewise.
574         * tree-ssa-loop-manip.c: Likewise.
575         * tree-ssa-phiopt.c: Likewise.
576         * tree-ssa-phiprop.c: Likewise.
577         * tree-ssa-threadupdate.c: Likewise.
578         * tree-ssa.c: Likewise.
579         * tree-ssa.h: Likewise.
580         * tree-vect-loop-manip.c: Likewise.
582 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
584         * doc/extend.texi: Fix typo in the weakref description.
586 2018-11-13  Richard Biener  <rguenther@suse.de>
588         PR tree-optimization/86991
589         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
590         group building until we have successfully detected the SLP
591         reduction.
592         (vect_is_simple_reduction): Remove fixup code here.
594 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
596         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
597         (dump_gimple_stmt): Use it.
598         (dump_gimple_stmt_loc): Likewise.
599         (dump_gimple_expr): Likewise.
600         (dump_gimple_expr_loc): Likewise.
601         (dump_generic_expr): Likewise.
602         (dump_generic_expr_loc): Likewise.
603         (dump_printf): Likewise.
604         (dump_printf_loc): Likewise.
605         (dump_dec): Likewise.
606         (dump_dec): Likewise.
607         (dump_hex): Likewise.
608         (dump_symtab_node): Likewise.
609         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
610         Guard dump call with dump_enabled_p.
611         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
612         * graphite-optimize-isl.c (optimize_isl): Likewise.
613         * graphite.c (graphite_transform_loops): Likewise.
614         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
615         * tree-parloops.c (parallelize_loops): Likewise.
616         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
617         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
618         (vect_prune_runtime_alias_test_list): Likewise.
619         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
620         (vect_estimate_min_profitable_iters): Likewise.
621         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
622         (vect_build_slp_tree_2): Likewise.
623         (vect_supported_load_permutation_p): Likewise.
624         (vect_slp_analyze_operations): Likewise.
625         (vect_slp_analyze_bb_1): Likewise.
626         (vect_slp_bb): Likewise.
627         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
628         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
629         (pass_slp_vectorize::execute): Likewise.
630         (increase_alignment): Likewise.
632 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
634         PR ipa/87955
635         * ipa-inline.c (report_inline_failed_reason): Guard calls to
636         cl_target_option_print_diff and cl_optimization_print_diff with
637         if (dump_file).
639 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
641         * doc/invoke.texi (-fsave-optimization-record): Note that the
642         output is compressed.
643         * optinfo-emit-json.cc: Include <zlib.h>.
644         (optrecord_json_writer::write): Compress the output.
646 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
648         * tree-vrp.c (value_range_base::dump): Dump type.
649         Do not use INF nomenclature for 1-bit types.
650         (dump_value_range): Group all variants to common dumping code.
651         (debug): New overloaded functions for value_ranges.
652         (value_range_base::dump): Remove no argument version.
653         (value_range::dump): Same.
655 2018-11-13  Richard Biener  <rguenther@suse.de>
657         PR tree-optimization/87931
658         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
659         nested cycles we support to latch computations vectorizable_reduction
660         handles.
662 2018-11-13  Martin Liska  <mliska@suse.cz>
664         PR tree-optimization/87885
665         * cfghooks.c (account_profile_record): Rename
666         to ...
667         (profile_record_check_consistency): ... this.
668         Calculate missing num_mismatched_freq_in.
669         (profile_record_account_profile): New function
670         that calculates time and size of a function.
671         * cfghooks.h (struct profile_record): Remove
672         all tuples.
673         (struct cfg_hooks): Remove after_pass flag.
674         (account_profile_record): Rename to ...
675         (profile_record_check_consistency): ... this.
676         (profile_record_account_profile): New.
677         * cfgrtl.c (rtl_account_profile_record): Remove
678         after_pass flag.
679         * passes.c (check_profile_consistency): Do only
680         checking.
681         (account_profile): Calculate size and time of
682         function only.
683         (pass_manager::dump_profile_report): Reformat
684         output.
685         (execute_one_ipa_transform_pass): Call
686         consistency check before clean upand call account_profile
687         after a clean up is done.
688         (execute_one_pass): Call check_profile_consistency and
689         account_profile instead of using after_pass flag..
690         * tree-cfg.c (gimple_account_profile_record): Likewise.
692 2018-11-13  Martin Liska  <mliska@suse.cz>
694         PR sanitizer/87930
695         * config/i386/i386.c (ix86_option_override_internal): Error
696         about usage -mabi=ms and -fsanitize={,kernel-}address.
698 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
700         * config/aarch64/aarch64.c
701         (aarch64_vectorize_preferred_vector_alignment): Change return type to
702         poly_uint64.
703         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
704         alignment being a poly int.
705         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
706         return type to poly_uint64.
707         * target.def (default_preferred_vector_alignment): Likewise.
708         * targhooks.c (default_preferred_vector_alignment): Likewise.
709         * targhooks.h (default_preferred_vector_alignment): Likewise.
710         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
711         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
712         poly int.
713         (vect_update_misalignment_for_peel): Likewise.
714         (vect_enhance_data_refs_alignment): Likewise.
715         (vect_find_same_alignment_drs): Likewise.
716         (vect_duplicate_ssa_name_ptr_info): Likewise.
717         (vect_setup_realignment): Likewise.
718         (vect_can_force_dr_alignment_p): Change alignment parameter type to
719         poly_uint64.
720         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
721         mask with a compile time variable vector alignment.
722         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
723         int.
724         (vect_do_peeling): Exit early if vector alignment is not constant.
725         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
726         a poly int.
727         (vectorizable_store): Likewise.
728         (vectorizable_load): Likweise.
729         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
730         poly_uint64.
731         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
732         poly int.
733         (vect_can_force_dr_alignment_p): Change alignment parameter type to
734         poly_uint64.
736 2018-11-13  Richard Biener  <rguenther@suse.de>
738         PR tree-optimization/87962
739         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
740         detect outer reduction for disqualifying in-loop uses.
742 2018-11-13  Richard Biener  <rguenther@suse.de>
744         PR tree-optimization/87967
745         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
746         for constants for the scalar loop.
748 2018-11-13  Alan Modra  <amodra@gmail.com>
750         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
751         integer constants when -mcmodel=medium.
753 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
755         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
756         * config/arc/elf.h (PROFILE_HOOK): Define.
757         * config/arc/linux.h (PROFILE_HOOK): Likewise.
759 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
761         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
762         jump table data in the text section.
764 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
766         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
767         to fit the eh_return pattern.
768         * config/arc/arc.md (eh_return): Define.
769         (VUNSPEC_ARC_EH_RETURN): Likewise.
770         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
771         new implementation.
772         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
775 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
777         * common/config/arc/arc-common.c (arc_option_optimization_table):
778         Millicode optimization is default on for size optimizations.
779         * config/arc/arc-protos.h (arc_check_multi): New function.
780         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
781         (ENTER_LEAVE_START_REG): Define.
782         (ENTER_LEAVE_END_REG): Likewise.
783         (arc_override_options): Disable millicode when long calls option
784         is on.
785         (arc_frame_info): Change it from int to bool.
786         (arc_compute_frame_size): Clean up.
787         (arc_save_restore): Remove.
788         (frame_save_reg): New function.
789         (frame_restore_reg): Likewise.
790         (arc_enter_leave_p): Likewise.
791         (arc_save_callee_saves): Likewise.
792         (arc_restore_callee_saves): Likewise.
793         (arc_save_callee_enter): Likewise.
794         (arc_restore_callee_leave): Likewise.
795         (arc_save_callee_milli): Likewise.
796         (arc_restore_callee_milli): Likewise.
797         (arc_expand_prologue): Reimplement to emit enter/leave
798         instructions.
799         (arc_expand_epilogue): Likewise.
800         (arc_check_multi): New function.
801         * config/arc/arc.md (push_multi_fp): New pattern.
802         (push_multi_fp_blink): Likewise.
803         (pop_multi_fp): Likewise.
804         (pop_multi_fp_blink): Likewise.
805         (pop_multi_fp_ret): Likewise.
806         (pop_multi_fp_blink_ret): Likewise.
807         * config/arc/arc.opt (mmillicode): Update option.
808         (mcode-density-frame): New option.
809         * config/arc/predicates.md (push_multi_operand): New predicate.
810         (pop_multi_operand): Likewise.
811         * doc/invoke.texi (ARC): Update ARC options information.
814 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
816         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
817         * config/arc/arc.c (operands_ok_ldd_std): New function.
818         (mem_ok_for_ldd_std): Likewise.
819         (gen_operands_ldd_std): Likewise.
820         * config/arc/arc.md: Add peephole2 rules for std/ldd.
822 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
824         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
825         test on stack_usage_file.
826         (lang_dependent_init): Do not open the .su file if generating LTO.
828 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
830         PR rtl-optimization/87918
831         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
832         simplify_gen_relational rather than simplify_gen_binary.
834 2018-11-13  Richard Biener  <rguenther@suse.de>
836         * tree-ssanames.h (set_range_info): Use value_range_base.
837         (get_range_info): Likewise.
838         * tree-ssanames.c (set_range_info): Likewise.
839         (get_range_info): Likewise.
840         * tree-vrp.c (value_range_base::union_helper): Split
841         out common parts of value_range[_base]::union_.
842         (value_range_base::union_): Update.
843         (value_range::union_): Likewise.
844         (determine_value_range_1): Use value_range_base.
845         (determine_value_range): Likewise.
846         * tree-vrp.h (value_range_base::union_helper): Move ...
847         (value_range::union_helper): ... from here.
849 2018-11-13  Alan Modra  <amodra@gmail.com>
851         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
852         offset for PRE_DEC.
853         (rs6000_secondary_reload_gpr): Don't call find_replacement.
855 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
857         PR middle-end/59634
858         * doc/invoke.texi (Optimize Options): Clarify that the
859         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
860         apply to data cache size.
862 2018-11-13  Alan Modra  <amodra@gmail.com>
864         * config/rs6000/predicates.md (logical_const_operand),
865         (logical_operand): Correct comment.
866         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
868 2018-11-13  Alan Modra  <amodra@gmail.com>
870         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
872 2018-11-13  Alan Modra  <amodra@gmail.com>
874         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
875         description.
876         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
877         covered by alternative.
878         (movcc_internal1): Ignore h for register preference.
879         (mov<mode>_hardfloat64): Likewise.
880         (mov<mode>_softfloat): Ignore c, l, h for register preference.
882 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
884         PR preprocessor/47823
885         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
886         documentation to...
887         (Pragmas): ...here.
888         * doc/extend.texi (Pragmas): Note additional pragmas documented
889         in the CPP manual.
891 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
893         PR middle-end/86677
894         PR middle-end/87528
895         * tree-scalar-evolution.c (expression_expensive_p): Make
896         BUILTIN POPCOUNT as expensive when backend does not define it.
898 2018-11-12  Fredrik Noring  <noring@nocrew.org>
900         * config.gcc: Update with-llsc defaults for MIPS r5900.
902 2018-11-12  Martin Liska  <mliska@suse.cz>
904         PR gcov-profile/87442
905         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
906         options.
907         * doc/invoke.texi: Document them.
908         * tree-profile.c (parse_profile_filter): New.
909         (parse_profile_file_filtering): Likewise.
910         (release_profile_file_filtering): Likewise.
911         (include_source_file_for_profile): Likewise.
912         (tree_profiling): Filter source files based on the
913         newly added options.
915 2018-11-12  Martin Liska  <mliska@suse.cz>
917         PR target/87903
918         * doc/extend.texi: Add missing values for __builtin_cpu_is and
919         __builtin_cpu_supports for x86 target.
921 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
923         PR middle-end/21110
924         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
925         have to be Pmode.
926         (Arithmetic): Likewise for "lo_sum".
928 2018-11-12  Renlin Li  <renlin.li@arm.com>
930         PR target/87815
931         * dse.c (get_stored_val): Add check for compile-time constantness
932         of gap.
934 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
936         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
937         (ARMv8_5a): New fgroup.
938         (armv8.5-a): New arch.
939         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
940         sb and predres.
941         * config/arm/arm-tables.opt: Regenerate.
942         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
943         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
944         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
945         Add matching rules for -march=armv8.5-a and extensions.
946         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
947         Add sb and predres to all armv8-a except armv8.5-a.
949 2018-11-12  Richard Biener  <rguenther@suse.de>
951         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
952         overload for single value.
953         (value_range[_base]::set_nonnull): New.
954         (value_range[_base]::set_null): Likewise.
955         (value_range): Document bitmap copying behavior, mark
956         copy constructor and assignment operator deleted.
957         (value_range::move): New.
958         (value_range::set_and_canonicalize): Default bitmap to zero.
959         (set_value_range_to_nonnull): Remove.
960         (set_value_range_to_null): Likewise.
961         (set_value_range): Likewise.
962         (set_value_range_to_value): Likewise.
963         (extract_range_from_unary_expr): Work on value_range_base.
964         (extract_range_from_binary_expr_1): Likewise.  Rename to...
965         (extract_range_from_binary_expr): ... this.
966         * tree-vrp.c (value_range::update): Clear equiv bitmap
967         if required.
968         (value_range::move): New, move equiv bitmap.
969         (value_range_base::set_undefined): Avoid assignment.
970         (value_range::set_undefined): Likewise.
971         (value_range_base::set_varying): Likewise.
972         (value_range::set_varying): Likewise.
973         (set_value_range): Remove.
974         (value_range_base::set): New overload for value.
975         (value_range::set): Likewise.
976         (set_value_range_to_nonnull): Remove.
977         (value_range_base::set_nonnull): New.
978         (value_range::set_nonnull): Likewise.
979         (set_value_range_to_null): Remove.
980         (value_range_base::set_null): New.
981         (value_range::set_null): Likewise.
982         (range_is_null): Work on value_range_base.
983         (range_is_nonnull): Likewise.
984         (ranges_from_anti_range): Likewise.
985         (extract_range_into_wide_ints): Likewise.
986         (extract_range_from_multiplicative_op): Likewise.
987         (extract_range_from_binary_expr): Likewise.  Update for API changes.
988         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
989         handling.
990         (value_range::intersect_helper): Avoid copy and assignment.
991         (value_range::union_helper): Likewise.
992         (determine_value_range_1): Adjust.
993         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
994         Avoid assignment by using move.
995         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
996         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
997         Likewise.
998         * tree-ssanames.c (get_range_info): Likewise.
999         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
1000         * vr-values.c (vr_values::get_value_range): Adjust.
1001         (vr_values::update_value_range): Likewise.
1002         (symbolic_range_based_on_p): Work on value_range_base.
1003         (vr_values::extract_range_from_binary_expr): Use value_range_base.
1004         (vr_values::extract_range_from_unary_expr): Likewise.
1005         (vr_values::extract_range_from_cond_expr): Avoid assignment.
1006         (vr_values::extract_range_from_comparison): Adjust.
1007         (vr_values::check_for_binary_op_overflow): Use value_range_base.
1008         (vr_values::extract_range_basic): Adjust.
1009         (vr_values::adjust_range_with_scev): Likewise.
1010         (vr_values::vrp_visit_assignment_or_call): Likewise.
1011         (vr_values::get_vr_for_comparison): Change API to avoid
1012         assignment and copy construction.
1013         (vr_values::compare_name_with_value): Adjust accordingly.
1014         (vr_values::compare_names): Likewise.
1015         (vr_values::extract_range_from_phi_node): Avoid assignment and
1016         bogus in-place modify of equiv bitmap.
1017         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
1018         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
1019         for extract_range_from_unary_expr API change.
1020         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
1022 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
1024         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
1026 2018-11-12  Richard Biener  <rguenther@suse.de>
1028         * tree-vrp.h (value_range_base::symbolic_p,
1029         value_range_base::constant_p, value_range_base::zero_p,
1030         value_range_base::singleton_p): Move from value_range.
1031         (value_range::dump): Add.
1032         * gimple-ssa-evrp-analyze.c
1033         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
1034         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
1035         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
1036         Use set_varying.
1037         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
1038         (value_range::constant_p): Likewise.
1039         (value_range::singleton_p): Likewise.
1040         (value_range_base::dump): Add.
1041         (set_value_range_to_undefined): Remove.
1042         (set_value_range_to_varying): Likewise.
1043         (range_int_cst_p): Take value_range_base argument.
1044         (range_int_cst_singleton_p): Likewise.
1045         (value_range_constant_singleton): Likewise.
1046         (vrp_set_zero_nonzero_bits): Likewise.
1047         (extract_range_from_multiplicative_op): Use set_varying.
1048         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
1049         (extract_range_from_unary_expr): Likewise.
1050         (dump_value_range_base): Change to overload of dump_value_range.
1051         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
1052         (vrp_prop::visit_stmt): Likewise.
1053         (value_range::intersect_helper): Likewise.
1054         (value_range::union_helper): Likewise.
1055         (determine_value_range_1): Likewise.
1057 2018-11-12  Richard Biener  <rguenther@suse.de>
1059         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
1060         (set_value_range_to_null): Likewise.
1061         * vr-values.c (vr_values::extract_range_from_comparison):
1062         Clear equiv for constant singleton ranges.
1064 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
1066         * config/i386/sse.md: Combine VFIXUPIMM* patterns
1067         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1068         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1069         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
1070         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1071         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1072         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
1074 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1076         PR c/69502
1077         * doc/extend.texi (Common Type Attributes): For the align type
1078         attribute, copy language about decreasing alignment from the
1079         corresponding variable attribute.
1081 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1083         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
1084         -frounding-math.
1086 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1088         PR c++/43105
1089         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
1090         -frtti and -fno-rtti code.
1092 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1094         PR c/26366
1095         * doc/extend.texi (Other Builtins): Document probability associated
1096         with __builtin_expect.
1098 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
1100         PR target/87928
1101         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
1102         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
1103         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
1104         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
1106 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1108         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
1110 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1112         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
1114 2018-11-11  Richard Biener  <rguenther@suse.de>
1116         * tree-vrp.h (class value_range_base): New base class for
1117         value_range containing all but the m_equiv member.
1118         (dump_value_range_base): Add.
1119         (range_includes_zero_p): Work on value_range_base.
1120         * tree-vrp.c (value_range_base::set): Split out base handling
1121         from...
1122         (value_range::set): this.
1123         (value_range::set_equiv): New.
1124         (value_range_base::value_range_base): New constructors.
1125         (value_range_base::check): Split out base handling from...
1126         (value_range::check): this.
1127         (value_range::equal_p): Refactor in terms of
1128         ignore_equivs_equal_p which is now member of the base.
1129         (value_range_base::set_undefined): New.
1130         (value_range_base::set_varying): Likewise.
1131         (value_range_base::dump):Split out base handling from...
1132         (value_range::dump): this.
1133         (value_range_base::set_and_canonicalize): Split out base handling
1134         from...
1135         (value_range::set_and_canonicalize): this.
1136         (value_range_base::union_): New.
1137         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
1138         for m_vr.
1139         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
1140         instead of value_range everywhere.
1141         (ipcp_vr_lattice::print): Use dump_value_range_base.
1142         (ipcp_vr_lattice::meet_with): Adjust.
1143         (ipcp_vr_lattice::meet_with_1): Likewise.
1144         (ipa_vr_operation_and_type_effects): Likewise.
1145         (propagate_vr_across_jump_function): Likewise.
1146         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
1147         (ipa_get_value_range): Likewise.
1148         (ipa_set_jfunc_vr): Likewise.
1149         (ipa_compute_jump_functions_for_edge): Likewise.
1151 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
1153         PR middle-end/65703
1154         * doc/invoke.texi (Optimize Options): Add @opindex entries
1155         for the positive forms of -fno-xxx and -mno-xxx options
1156         that were lacking them.
1158 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
1160         * combine.c (make_more_copies): Only make an intermediate copy if the
1161         dest of a move is a pseudo.
1163 2018-11-09  Maya Rashish  <coypu@sdf.org>
1165         PR target/87221
1166         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
1167         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
1169 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
1171         PR driver/41179
1172         PR middle-end/65703
1173         * doc/invoke.texi (Optimize Options): Clarify default behavior
1174         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
1176 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1178         PR tree-optimization/87940
1179         * expr.c (string_constant): Don't strip NOPS in subexpressions.
1180         Fold PLUS_EXPR correctly.
1182 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
1184         PR target/87762
1185         * config/s390/s390.c (s390_safe_relative_long_p): New function.
1186         (annotate_constant_pool_refs): Skip insns which support
1187         relative addressing.
1188         (annotate_constant_pool_refs_1): New helper function.
1189         (find_constant_pool_ref): Skip insns which support relative
1190         addression.
1191         (find_constant_pool_ref_1): New helper function.
1192         (replace_constant_pool_ref): Skip insns which support
1193         relative addressing.
1194         (replace_constant_pool_ref_1): New helper function.
1195         (s390_mainpool_start): Adapt to the new signature.
1196         (s390_mainpool_finish): Likewise.
1197         (s390_chunkify_start): Likewise.
1198         (s390_chunkify_finish): Likewise.
1199         (pass_s390_early_mach::execute): Likewise.
1200         (s390_prologue_plus_offset): Likewise.
1201         (s390_emit_prologue): Likewise.
1202         (s390_emit_epilogue): Likewise.
1204 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
1206         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
1207         but unsupported lastprivate with conditional modifier.
1209 2018-11-09  Jeff Law  <law@redhat.com>
1211         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
1212         unused argument better.  Add gcc_unreachable to silence warning.
1214 2018-11-09  Martin Sebor  <msebor@redhat.com>
1216         PR middle-end/81824
1217         * attribs.c (has_attribute): New helper function.
1218         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
1219         * attribs.h (decls_mismatched_attributes): Declare.
1220         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
1221         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
1222         * common.opt (-Wattribute-alias): Take an argument.
1223         (-Wno-attribute-alias): New option.
1224         * doc/extend.texi (Common Function Attributes): Document copy.
1225         (Common Variable Attributes): Same.
1226         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
1227         (-Wattribute-alias): Document new option argument.
1229 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
1231         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
1232         matching criteria.  Remove unused array initializer.
1234 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
1235             Jinsong Ji  <jji@us.ibm.com>
1237         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
1238         constraints by introducing a new temporary.
1239         (_mm_cvtss_si64): Likewise.
1241 2018-11-09  Martin Liska  <mliska@suse.cz>
1243         * common.opt: Add -fipa-stack-alignment flag.
1244         * doc/invoke.texi: Document it.
1245         * final.c (rest_of_clean_state): Guard stack
1246         shrinking with flag.
1248 2018-11-09  Martin Liska  <mliska@suse.cz>
1250         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
1251         to ...
1252         (ipa_discover_variable_flags): ... this.
1253         * common.opt: Come up with new flag -fipa-reference-addressable.
1254         * doc/invoke.texi: Document it.
1255         * ipa-reference.c (propagate): Call the renamed fn.
1256         * ipa-visibility.c (whole_program_function_and_variable_visibility):
1257         Likewise.
1258         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
1259         ...
1260         (ipa_discover_variable_flags): ... this.  Discover
1261         non-addressable variables only with the newly added flag.
1262         * opts.c: Enable the newly added flag with -O1 and higher
1263         optimization level.
1265 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
1267         * json.cc (selftest::test_writing_literals): Fix comment.
1269 2018-11-09  Martin Liska  <mliska@suse.cz>
1271         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
1272         string to a stack buffer.
1273         (aarch64_parse_cpu): Likewise.
1274         (aarch64_parse_tune): Likewise.
1276 2018-11-09  Richard Biener  <rguenther@suse.de>
1278         PR tree-optimization/87953
1279         * tree-vect-loop.c (vectorizable_reduction): For analysis
1280         always pass ops[0] to vectorizable_condition.
1282 2018-11-09  Stafford Horne  <shorne@gmail.com>
1283             Richard Henderson  <rth@twiddle.net>
1284             Joel Sherrill  <joel@rtems.org>
1286         * common/config/or1k/or1k-common.c: New file.
1287         * config/or1k/*: New.
1288         * config.gcc (or1k*-*-*): New.
1289         * configure.ac (or1k*-*-*): New test for openrisc tls.
1290         * configure: Regenerated.
1291         * doc/install.texi: Document OpenRISC triplets.
1292         * doc/invoke.texi: Document OpenRISC arguments.
1293         * doc/md.texi: Document OpenRISC.
1295 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
1297         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
1298         (arm7tdmi-s): Delete CPU.
1299         (arm710t): Add aliases for arm720t and arm740t.
1300         (arm720t, arm740t): Delete CPUs.
1301         (arm920t): Add aliases for arm920, arm922t and arm940t.
1302         (arm920, arm922t, arm940t): Delete CPUs.
1303         (arm10tdmi): Add alias for arm1020t.
1304         (arm1020t): Delete CPU.
1305         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
1306         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
1307         (arm10e): Add aliases for arm1020e and arm1022e.
1308         (arm1020e, arm1022e): Delete CPU.
1309         * config/arm/arm.md (generic_sched): Remove entries that are now
1310         handled by aliases.
1311         (generic_vfp): Likewise.
1312         * config/arm/arm1020e.md: Simplify tuning selection based on alias
1313         changes.
1314         * config/arm/arm-tune.md: Regenerated.
1315         * config/arm/arm-tables.opt: Regenerated.
1317 2018-11-09  Richard Biener  <rguenther@suse.de>
1319         PR tree-optimization/87621
1320         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
1321         op with only phi inputs.
1322         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
1323         (ch_base::copy_headers): Run CSE on copied loop headers.
1324         (pass_ch_vect::process_loop_p): Simplify.
1326 2018-11-09  Alexandre Oliva <oliva@adacore.com>
1328         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
1329         for non-w64 x86_64 biarch.
1331 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
1333         PR rtl-optimization/86438
1334         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
1335         of in_b for the compare if in_b is SET_DEST.
1337         PR target/87793
1338         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
1339         non-toplevel UNSPEC.
1341 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1343         * tree-vrp.c (value_range::check): Do not access internals
1344         directly.
1345         (value_range::singleton_p): Same.
1346         (value_range::type): Same.
1347         (vrp_finalize): Use value_range API.
1349 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1351         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
1353 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1355         * vr-values.c (vr_values::get_value_range): Use value_range API
1356         instead of piecing together ranges.
1357         (vr_values::update_value_range): Same.
1359 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1361         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
1362         of performing ad-hoc calculations.
1363         * tree-ssanames.c (set_range_info): New overloaded function
1364         accepting value_range &.
1365         (get_range_info): Same.
1366         * tree-ssanames.h (set_range_info_raw): Remove.
1367         (set_range_info): New prototype.
1368         (get_range_info): Same.
1369         * tree-vrp.h (value_range::null_p): Rename to zero_p.
1370         * tree-vrp.c (value_range::null_p): Same.
1372 2018-11-09  Jan Hubicka  <jh@suse.cz>
1374         * tree.c (fld_type_variant_equal_p): Test user align flag.
1375         (flt_type_variant): Copy user align flag.
1376         (fld_incomplete_type_of): Clear it.
1378 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1380         * config/arm/neon.md (div<mode>3): New pattern.
1382 2018-11-08  Andi Kleen  <ak@linux.intel.com>
1384         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
1385         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
1386         (ix86_handle_option): Handle OPT_mptwrite.
1387         * config/i386/cpuid.h (bit_PTWRITE): Add.
1388         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
1389         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
1390         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
1391         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
1392         (ix86_option_override_internal): Handle PTA_PTWRITE.
1393         (ix86_valid_target_attribute_inner_p): Define ptwrite.
1394         (def_builtin2): Force UINT64 to be 64bit only.
1395         * config/i386/i386.h (TARGET_PTWRITE): Add.
1396         (TARGET_PTWRITE_P): Add.
1397         (PTA_PTWRITE): Add.
1398         * config/i386/i386.md: Define ptwrite.
1399         * config/i386/i386.opt: Add -mptwrite.
1400         * config/i386/immintrin.h (_ptwrite64): Add.
1401         (_ptwrite32): Add
1402         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
1403         * doc/invoke.texi: Document -mptwrite.
1405 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
1407         PR rtl-optimization/87600
1408         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
1409         * lra-constraints.c (process_alt_operands): Skip illegal hard
1410         register usage.  Prefer reloading non hard register operands.
1412 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
1414         PR other/36572
1415         * doc/invoke.texi (Optimize Options): Clarify default behavior
1416         for -fno-sched-interblock and -fno-sched-spec.
1418 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
1420         * collect2.c (linker_select):  Add USE_LLD_LD.
1421         (ld_suffixes): Add ld.lld.
1422         (main): Handle -fuse-ld=lld.
1423         * common.opt (-fuse-ld=lld): New option.
1424         * doc/invoke.texi (-fuse-ld=lld): Document.
1425         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
1427 2018-11-08  Paul Koning  <ni1d@arrl.net>
1429         * config/pdp11/constraints.md: Add "Z" series constraints for use
1430         with pre-dec and post-inc addressing.
1431         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
1432         (pdp11_expand_operands): Add int argument (word count).
1433         (pdp11_sp_frame_offset): Delete.
1434         (pdp11_cmp_length): New function.
1435         (pushpop_regeq): New function.
1436         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
1437         Add hook.
1438         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
1439         frame layout.
1440         (pdp11_initial_elimination_offset): Ditto.
1441         (pdp11_expand_operands): Add word count argument.  Bugfixes.
1442         (output_move_multiple): Change how pointer adjustment is done.
1443         (pdp11_gen_int_label): Correct format.
1444         (output_ascii): Ditto.
1445         (pdp11_asm_output_var): Add code for DEC assembler case.
1446         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
1447         value.
1448         (legitimate_const_double_p): Ditto.
1449         (pdp11_register_move_cost): Adjust for new register classes.
1450         (pdp11_regno_reg_class): Ditto.
1451         (expand_block_move): Delete.
1452         (pushpop_regeq): New function.
1453         (pdp11_legitimate_address_p): Bugfix in check for constant
1454         offset.
1455         (pdp11_sp_frame_offset): Delete.
1456         (pdp11_reg_save_size): New helper function for new frame layout.
1457         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
1458         (pdp11_expand_shift): Bugfix in check for constant shift count.
1459         (pdp11_shift_length): Ditto.
1460         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
1461         (pdp11_cmp_length): New function.
1462         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
1463         some compile options.
1464         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
1465         (CALL_USED_REGISTERS): Ditto.
1466         (ELIMINABLE_REGS): Ditto.
1467         (REGISTER_NAMES): Ditto.
1468         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
1469         constraints.
1470         (REG_CLASS_NAMES): Ditto.
1471         (REG_CLASS_CONTENTS): Ditto.  Also remove
1472         HARD_FRAME_POINTER_REGNUM.
1473         (CPU_REG_CLASS): New macro.
1474         (CLASS_MAX_NREGS): Adjust for new register classes.
1475         (FUNCTION_PROFILER): Make no-op.
1476         (may_call_alloca): Remove unused declaration.
1477         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
1478         (ASM_OUTPUT_SKIP): Fix format.
1479         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
1480         (HARD_FRAME_POINTER_REGNUM): Remove.
1481         (return): Delete.
1482         (*rts): Rename.  Remove epilogue related checks.
1483         (cmpsi, cmpdi): New insn.
1484         (cbranch<mode>4): Change to apply to SI and DI modes as well.
1485         (mov<mode>): Change constraints to enforce that push/pop
1486         destination cannot use the same register as source.
1487         (*mov<mode><cc_cc>): Ditto.
1488         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
1489         at assembly output rather than as RTL expander.
1490         (zero_extendqihi2): Bugfix in check for same registers.
1491         (adddi3_nocc): Bugfix in check for constant operand.
1492         (addsi3_nocc): Ditto.
1493         (subdi3_nocc): Ditto.
1494         (subsi3_nocc): Ditto.
1495         (negdi2_nocc): Copy input to pdp11_expand_operands.
1496         (negsi2_nocc): Ditto.
1497         (bswap2_nocc): Ditto.
1498         * config/pdp11/pdp11.opt (mlra): Fix documentation.
1499         * config/pdp11/t-pdp11: Use -Os.
1501 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
1503         * config/arm/parsecpu.awk (/alias/): New parsing rule.
1504         (/begin cpu/): Check that the cpu name hasn't been previously defined.
1505         (gen_comm_data): Print out CPU alias tables.
1506         (check_cpu): Match aliases when checking the CPU name.
1507         * config/arm/arm-protos.h (cpu_alias): New structure.
1508         (cpu_option): Add entry for aliases.
1509         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
1510         strongarm1100 and strongarm1110.
1511         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
1512         (config/arm/arm-generic.md): Remove redundant references to
1513         strongarm110, strongarm1100 and strongarm1110.
1514         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1515         Scan aliases for additional hints.
1516         (arm_parse_cpu_option_name): Also match a cpu name against the list
1517         of aliases.
1518         * config/arm/arm-tables.opt: Regenerated.
1519         * config/arm/arm-tune.md: Regenerated.
1521 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
1523         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
1524         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
1525         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1526         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
1527         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1528         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
1529         * gengtype.c (open_base_files): Add omp-general.h.
1530         * gimple.c (gimple_build_omp_critical):
1531         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
1532         gimple_omp_taskgroup_set_clauses.
1533         (gimple_build_omp_atomic_load): Add mo argument, call
1534         gimple_omp_atomic_set_memory_order.
1535         (gimple_build_omp_atomic_store): Likewise.
1536         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
1537         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
1538         instead of GSS_OMP.
1539         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
1540         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
1541         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
1542         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
1543         different value for GF_OMP_ATOMIC_NEED_VALUE.
1544         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
1545         comments.
1546         (struct gimple_statement_omp_single_layout): And remove here.
1547         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
1548         than gimple_statement_omp_single_layout.
1549         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
1550         GIMPLE_OMP_TEAMS.
1551         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
1552         (gimple_omp_subcode): Formatting fix.
1553         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
1554         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
1555         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
1556         gimple_omp_teams_host, gimple_omp_teams_set_host,
1557         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
1558         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
1559         gimple_omp_taskgroup_set_clauses): New inline functions.
1560         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
1561         (gimple_build_omp_atomic_store): Likewise.
1562         (gimple_omp_atomic_seq_cst_p): Remove.
1563         (gimple_omp_atomic_memory_order): New function.
1564         (gimple_omp_atomic_set_seq_cst): Remove.
1565         (gimple_omp_atomic_set_memory_order): New function.
1566         (gimple_build_omp_taskgroup): Add clauses argument.
1567         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
1568         (dump_gimple_omp_task): Print taskwait with depend clauses.
1569         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
1570         dump_omp_atomic_memory_order.
1571         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
1572         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
1573         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
1574         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
1575         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
1576         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
1577         ORT_UNTIED_TASKLOOP enumerators.
1578         (enum gimplify_defaultmap_kind): New.
1579         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
1580         target_map_pointers_as_0len_arrays members, add defaultmap.
1581         (new_omp_context): Initialize defaultmap member.
1582         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1583         (maybe_fold_stmt): Don't fold even in host teams regions.
1584         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
1585         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
1586         ctx->omp_firstprivatize_variable.
1587         (omp_add_variable): Don't add private/firstprivate for VLAs in
1588         ORT_TASKGROUP.
1589         (omp_default_clause): Print "taskloop" rather than "task" if
1590         ORT_*TASKLOOP.
1591         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1592         Handle new defaultmap clause kinds.
1593         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
1594         iterator to be lastprivate or private.  Fix up diagnostics if linear
1595         is used on collapse>1 simd iterator.
1596         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1597         (gimplify_omp_depend): New function.
1598         (gimplify_scan_omp_clauses): Add shared clause on parallel for
1599         combined parallel master taskloop{, simd} if taskloop has
1600         firstprivate, lastprivate or reduction clause.  Handle
1601         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
1602         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
1603         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
1604         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
1605         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
1606         lastprivate.
1607         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
1608         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.  
1609         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
1610         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1611         (gimplify_omp_task): Handle taskwait with depend clauses.
1612         (gimplify_omp_for): Add shared clause on parallel for combined
1613         parallel master taskloop{, simd} if taskloop has firstprivate,
1614         lastprivate or reduction clause.  Use ORT_TASKLOOP or
1615         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
1616         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
1617         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
1618         __for_end and __for_range temporaries on OMP_PARALLEL for
1619         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
1620         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
1621         sandwiched in between two taskloops.
1622         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
1623         instead of ctx->omp_firstprivatize_variable.
1624         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
1625         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
1626         host teams, use gimplify_and_return_first etc. for body like
1627         for target or target data constructs, and at the end call
1628         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
1629         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
1630         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
1631         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
1632         gimple_omp_atomic_set_seq_cst calls.
1633         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
1634         case, handle taskgroup clauses.
1635         * lto-streamer-out.c (hash_tree): Handle
1636         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1637         * Makefile.in (GTFILES): Add omp-general.h.
1638         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
1639         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
1640         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
1641         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
1642         BUILT_IN_GOMP_LOOP_DOACROSS_START,
1643         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
1644         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1645         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
1646         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
1647         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
1648         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
1649         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
1650         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1651         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
1652         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
1653         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
1654         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
1655         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
1656         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
1657         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
1658         * omp-expand.c (workshare_safe_to_combine_p): Return false for
1659         non-worksharing loops.
1660         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
1661         (determine_parallel_type): Don't combine parallel with worksharing
1662         which has _reductemp_ clause.
1663         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
1664         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
1665         if there is nonmonotonic modifier or if there is no modifier and no
1666         ordered clause.  For dynamic and guided schedule without monotonic
1667         and nonmonotonic modifier, default to nonmonotonic.
1668         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
1669         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
1670         task reductions.
1671         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
1672         there are any reduction clauses.
1673         (expand_taskwait_call): New function.
1674         (expand_teams_call): New function.
1675         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
1676         expand_teams_call for it.  Formatting fix.  Handle taskwait with
1677         depend clauses.
1678         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
1679         of worksharing loops with task reductions.
1680         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
1681         expansion of worksharing loops with task reductions.
1682         (expand_omp_sections): Handle expansion of sections with task
1683         reductions.
1684         (expand_omp_synch): For host teams call expand_omp_taskreg.
1685         (omp_memory_order_to_memmodel): New function.
1686         (expand_omp_atomic_load, expand_omp_atomic_store,
1687         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
1688         instead of gimple_omp_atomic_seq_cst_p.
1689         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
1690         depend clauses as a standalone directive.
1691         * omp-general.c (enum omp_requires): New variable.
1692         (omp_extract_for_data): Initialize have_reductemp member.  Allow
1693         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
1694         GT_EXPR loops depending on incr sign.  Formatting fixes.
1695         * omp-general.h (struct omp_for_data): Add have_reductemp member.
1696         (enum omp_requires): New enum.
1697         (omp_requires_mask): Declare.
1698         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
1699         Fix comment typos.
1700         * omp-low.c (struct omp_context): Add task_reductions and
1701         task_reduction_map fields.
1702         (is_host_teams_ctx): New function.
1703         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
1704         (use_pointer_for_field): Use is_global_var instead of
1705         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
1706         in outer contexts.
1707         (build_outer_var_ref): Ignore taskgroup outer contexts.
1708         (delete_omp_context): Release task_reductions and task_reduction_map.
1709         (scan_sharing_clauses): Don't add any fields for reduction clause on
1710         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
1711         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
1712         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
1713         Handle OMP_CLAUSE_NONTEMPORAL.
1714         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
1715         needed.
1716         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
1717         clauses with task modifier.
1718         (scan_omp_task): Handle taskwait with depend clauses.
1719         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
1720         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
1721         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
1722         (scan_omp_for): Fix comment formatting.
1723         (scan_omp_teams): Handle host teams constructs.
1724         (check_omp_nesting_restrictions): Allow teams with no outer
1725         OpenMP context.  Adjust diagnostics for teams strictly nested into
1726         some explicit OpenMP construct other than target.  Allow OpenMP atomics
1727         inside of simd regions.
1728         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
1729         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
1730         taskreg_nesting_level while scanning host teams construct.
1731         (task_reduction_read): New function.
1732         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
1733         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
1734         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
1735         second argument create_tmp_var if it is NULL.  Don't ignore shared
1736         clauses in is_host_teams_ctx contexts.  Handle
1737         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
1738         clauses.
1739         (lower_reduction_clauses): Ignore reduction clauses with task
1740         modifier.  Remove second argument create_tmp_var if it is NULL.
1741         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
1742         (lower_send_clauses): Ignore reduction clauses with task modifier.
1743         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
1744         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
1745         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
1746         rely that it is the last stmt in body so far.  Ignore outer taskgroup
1747         contexts.
1748         (omp_task_reductions_find_first, omp_task_reduction_iterate,
1749         lower_omp_task_reductions): New functions.
1750         (lower_omp_sections): Handle reduction clauses with taskgroup
1751         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
1752         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
1753         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
1754         modifiers.
1755         (lower_omp_taskgroup): Handle taskgroup reductions.
1756         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
1757         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
1758         (lower_depend_clauses): If there are any
1759         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1760         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
1761         seen, assume lowering is done already and return early.  Set kind
1762         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
1763         (lower_omp_taskreg): Handle reduction clauses with task modifier on
1764         parallel construct.  Handle reduction clause on taskloop construct.
1765         Handle taskwait with depend clauses.
1766         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
1767         for host teams constructs.
1768         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
1769         nontemporal and _reductemp_ clause entries.
1770         (omp_clause_code_name): Likewise.
1771         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
1772         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
1773         * tree-core.h (enum omp_clause_code): Add
1774         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
1775         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
1776         (struct tree_base): Add omp_atomic_memory_order field into union.
1777         Remove OMP_ATOMIC_SEQ_CST comment.
1778         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1779         and OMP_CLAUSE_DEPEND_DEPOBJ.
1780         (struct tree_omp_clause): Add subcode.defaultmap_kind.
1781         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
1782         OpenMP constructs with body and clauses operands.
1783         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
1784         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
1785         (OMP_TASKGROUP_CLAUSES): Define.
1786         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
1787         OMP_CLAUSE__LOOPTEMP_.
1788         (OMP_ATOMIC_SEQ_CST): Remove.
1789         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
1790         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
1791         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
1792         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
1793         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
1794         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
1795         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
1796         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
1797         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
1798         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
1799         Define.
1800         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
1801         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1802         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
1803         (convert_local_omp_clauses): Likewise.  Remove useless test.
1804         * tree-parloops.c (create_call_for_reduction_1): Pass
1805         OMP_MEMORY_ORDER_RELAXED as new argument to
1806         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
1807         * tree-pretty-print.c (dump_omp_iterators): New function.
1808         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
1809         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
1810         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
1811         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
1812         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
1813         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
1814         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
1815         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
1816         (dump_omp_atomic_memory_order): New function.
1817         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
1818         taskwait with depend clauses.
1819         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
1820         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
1821         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
1822         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
1823         write_ts_omp_clause_tree_pointers): Likewise.
1825 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1827         PR ipa/86395
1828         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
1829         "-missed", "-note", and "-all" sub-options.
1830         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
1831         API.
1832         (can_early_inline_edge_p): Likewise.
1833         (want_early_inline_function_p): Likewise.
1834         (want_inline_self_recursive_call_p): Likewise.
1835         (recursive_inlining): Likewise.
1836         (inline_small_functions): Likewise.
1837         (flatten_function): Likewise.
1838         (ipa_inline): Likewise.
1839         (inline_always_inline_functions): Likewise.
1840         (early_inline_small_functions): Likewise.
1841         (early_inliner): Likewise.
1842         * tree-inline.c (expand_call_inline): Likewise.
1844 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1846         * pretty-print.c (pp_format): Handle %f.
1847         (selftest::test_pp_format): Add test of %f.
1848         * pretty-print.h (pp_double): New macro.
1850 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1852         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
1853         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
1854         (dump_pretty_printer::decode_format): Implement "%C" for
1855         cgraph_node *.
1856         (selftest::test_capture_of_dump_calls): Rename "where" to
1857         "stmt_loc".  Convert test_decl to a function decl and set its
1858         location.  Add a symbol_table_test RAII instance and a
1859         cgraph_node, using it to test "%C" and dump_symtab_node.
1861 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
1863         PR middle-end/87916
1864         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
1866 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1868         * cgraph.c: Include "selftest.h".
1869         (saved_symtab): New variable.
1870         (selftest::symbol_table_test::symbol_table_test): New ctor.
1871         (selftest::symbol_table_test::~symbol_table_test): New dtor.
1872         (selftest::test_symbol_table_test): New test.
1873         (selftest::cgraph_c_tests): New.
1874         * cgraph.h (saved_symtab): New decl.
1875         (selftest::symbol_table_test): New class.
1876         * selftest-run-tests.c (selftest::run_tests): Call
1877         selftest::cgraph_c_tests.
1878         * selftest.h (selftest::cgraph_c_tests): New decl.
1880 2018-11-08  Richard Biener  <rguenther@suse.de>
1882         * tree-data-ref.h (lambda_int): New typedef.
1883         (lambda_vector_gcd): Adjust.
1884         (lambda_vector_new): Likewise.
1885         (lambda_matrix_new): Likewise.
1886         * tree-data-ref.c  (print_lambda_vector): Adjust.
1888 2018-11-08  Richard Biener  <rguenther@suse.de>
1890         PR tree-optimization/87929
1891         * tree-complex.c (expand_complex_comparison): Clean EH.
1893 2018-11-08  Martin Liska  <mliska@suse.cz>
1895         * doc/extend.texi: Reword.
1896         * predict.c (expr_expected_value_1): Likewise.
1898 2018-11-08  Richard Biener  <rguenther@suse.de>
1900         PR tree-optimization/87913
1901         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
1902         of extreme values to ordered comparisons.
1904 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1906         PR middle-end/42726
1907         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
1909 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1911         * doc/invoke.texi: Remove leading dash from @opindex entries
1912         throughout the file.
1914 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1916         PR driver/80828
1917         * doc/invoke.texi (Option Summary): Add -e and --entry.
1918         (Link Options): Likewise.
1920 2018-11-07  Nathan Sidwell  <nathan@acm.org>
1922         PR 87926
1923         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
1924         --disable-checking bootstrap.
1926 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
1928         * configure: Regenerated.
1930 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1932         PR c/87691
1933         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
1934         to the mode of the widest field iff the widest field has mode class
1935         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
1936         reference.
1938 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
1940         * common.opt: Add -Wattribute-warning.
1941         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
1942         * expr.c (expand_expr_real_1): Add new attribute to warning_at
1943         call to allow user configure behavior of "warning" attribute.
1945 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
1947         * target.def: Put @: after every vs., e.g., and i.e. where it is
1948         followed by whitespace.
1949         * doc/extend.texi: Ditto.
1950         * doc/fragments.texi: Ditto.
1951         * doc/gimple.texi: Ditto.
1952         * doc/implement-c.texi: Ditto.
1953         * doc/install.texi: Ditto.
1954         * doc/invoke.texi: Ditto.
1955         * doc/md.texi: Ditto.
1956         * doc/plugins.texi: Ditto.
1957         * doc/rtl.texi: Ditto.
1958         * doc/sourcebuild.texi: Ditto.
1959         * doc/tm.texi.in: Ditto.
1960         * doc/ux.texi: Ditto.
1961         * doc/tm.texi: Regenerate.
1963 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1965         * config/arm/arm-cpus.in (ares): New entry.
1966         * config/arm/arm-tables.opt: Regenerate.
1967         * config/arm/arm-tune.md: Likewise.
1968         * doc/invoke.texi (ARM Options): Document ares.
1970 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1972         * config/aarch64/aarch64-cores.def (ares): Define.
1973         * config/aarch64/aarch64-tune.md: Regenerate.
1974         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
1976 2018-11-07  Jan Hubicka  <jh@suse.cz>
1978         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
1979         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
1980         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
1981         TYPE_DECL.
1983 2018-11-07  Richard Biener  <rguenther@suse.de>
1985         PR tree-optimization/87914
1986         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
1987         of nested cycles.
1988         (vectorizable_reduction): Handle shifts and rotates by dispatching
1989         to vectorizable_shift.
1990         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
1991         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
1992         def cases.
1993         (vectorizable_shift): Export and handle being called as
1994         vect_nested_cycle.
1995         (vect_analyze_stmt): Call vectorizable_shift after
1996         vectorizable_reduction.
1997         * tree-vectorizer.h (vectorizable_shift): Declare.
1999 2018-11-07  Jan Hubicka  <jh@suse.cz>
2001         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
2002         than const decls in TREE_VALUE of enum.
2003         (dump_type_inheritance_graph): Improve duplicate dumping.
2004         (free_enum_values): New.
2005         (build_type_inheritance_graph): Use it.
2006         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
2007         which are not main variants or not ODR types.
2008         (verify_type_variant): Expect variants to have no TYPE_VALUES.
2010 2018-11-07  Richard Biener  <rguenther@suse.de>
2012         * ipa-inline.c (want_inline_small_function_p): Compute
2013         big_speedup_p lazily and last.
2015 2018-11-07  Jan Hubicka  <jh@suse.cz>
2017         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
2018         building incomplete variant of complete type.
2019         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
2020         variant of complete type.
2022 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2024         * config/mips/mips.c: Fix typo in documentation of
2025         mips_loongson_ext2_prefetch_cookie.
2026         (mips_option_override): fix brain twister logical.
2027         * config/mips/mips.h: Fix typo in documentation of
2028         ISA_HAS_CTZ_CTO and define pattern.
2029         * config/mips/mips.md (prefetch): Hoist EXT2 above
2030         the 2EF/EXT block.
2031         (prefetch_indexed): Hoist EXT2 above the EXT block.
2033 2018-11-07  Jan Hubicka  <jh@suse.cz>
2035         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
2036         return and parameter types of function and method types.
2037         (free_lang_data_in_cgraph): Update.
2039 2018-11-07  Martin Liska  <mliska@suse.cz>
2041         PR rtl-optimization/87868
2042         * postreload-gcse.c (eliminate_partially_redundant_load): Set
2043         threshold to max_count if we would overflow.
2044         * profile-count.h: Make max_count a public constant.
2046 2018-11-07  Martin Liska  <mliska@suse.cz>
2048         * mem-stats.h: Fix GNU coding style.
2050 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2052         * config/mips/gs264e.md: New.
2053         * config/mips/mips-cpus.def: Define gs264e.
2054         * config/mips/mips-tables.opt: Regenerate.
2055         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
2056         gs264e.
2057         (mips_issue_rate): Add support for gs264e.
2058         (mips_multipass_dfa_lookahead): Likewise.
2059         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
2060         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
2061         (MIPS_ASE_MSA_SPEC): New.
2062         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
2063         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
2064         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
2065         * config/mips/mips.md: Include gs264e.md.
2066         (processor): Add gs264e.
2067         * config/mips/mips.opt (MSA): Use Mask instead of Var.
2068         * doc/invoke.texi: Add gs264e to supported architectures.
2070 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2072         * config/mips/gs464e.md: New.
2073         * config/mips/mips-cpus.def: Define gs464e.
2074         * config/mips/mips-tables.opt: Regenerate.
2075         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
2076         gs464e.
2077         (mips_issue_rate): Add support for gs464e.
2078         (mips_multipass_dfa_lookahead): Likewise.
2079         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
2080         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
2081         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
2082         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
2083         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
2084         * config/mips/mips.md: Include gs464e.md.
2085         (processor): Add gs464e.
2086         * doc/invoke.texi: Add gs464e to supported architectures.
2088 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2090         * config/mips/loongson3a.md: Rename to ...
2091         * config/mips/gs464.md: ... here.
2092         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
2093         as an alias of gs464 processor.
2094         * config/mips/mips-tables.opt: Regenerate.
2095         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
2096         instead of PROCESSOR_LOONGSON_3A.
2097         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
2098         TUNE_LOONGSON_3A.
2099         (mips_option_override): Enable MMI and EXT for gs464.
2100         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
2101         Rename TUNE_LOONGSON_3A to TUNE_GS464.
2102         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
2103         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
2104         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
2105         TARGET_LOONGSON_3A.
2106         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
2107         (processor): Add gs464;
2108         * doc/invoke.texi: Add gs464 to supported architectures.
2110 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2112         * config/mips/mips-protos.h
2113         (mips_loongson_ext2_prefetch_cookie): New prototype.
2114         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
2115         (mips_option_override): Enable TARGET_LOONGSON_EXT when
2116         TARGET_LOONGSON_EXT2 is true.
2117         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
2118         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
2119         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
2120         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
2121         TARGET_LOONGSON_EXT2.
2122         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
2123         (define_insn "ctz<mode>2"): New insn pattern.
2124         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
2125         (define_insn "prefetch_indexed_<mode>"): Include
2126         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
2127         * config/mips/mips.opt (-mloongson-ext2): Add option.
2128         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
2130 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2132         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
2133         __mips_loongson_ext.
2134         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
2135         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
2136         -mloongson-ext.
2137         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
2138         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
2139         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
2140         instead of TARGET_LOONGSON_3A.
2141         * config/mips/mips.opt (-mloongson-ext): Add option.
2142         * gcc/doc/invoke.texi (-mloongson-ext): Document.
2144 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2146         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
2147         * config/mips/loongson.md: Move to ...
2148         * config/mips/loongson-mmi.md: here; Adjustment.
2149         * config/mips/loongson.h: Move to ...
2150         State as deprecated. Include loongson-mmiintrin.h for back
2151         compatibility and warning.
2152         * config/mips/loongson-mmiintrin.h: ... here.
2153         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
2154         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
2155         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
2156         (mips_option_override): Make sure MMI use hard float;
2157         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
2158         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
2159         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
2160         TARGET_LOONGSON_VECTORS.
2161         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
2162         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
2163         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
2164         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
2165         -mloongson-mmi.
2166         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
2167         TARGET_LOONGSON_VECTORS.
2168         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
2169         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
2170         (Loongson MMI patterns): Include loongson-mmi.md instead of
2171         loongson.md.
2172         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
2173         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
2175 2018-11-07  Richard Biener  <rguenther@suse.de>
2177         PR lto/87906
2178         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
2179         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
2181 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
2183         PR rtl-optimization/87874
2184         * lra.c (lra_substitute_pseudo): Do not create a subreg for
2185         const wide ints.
2187 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
2189         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
2190         if not in indexed or indirect form.
2191         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
2192         (bswap<mode>2_store): Ditto.
2194 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
2196         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
2197         the UNSPEC.
2199 2018-11-06  Richard Biener  <rguenther@suse.de>
2201         PR tree-optimization/86850
2202         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
2203         instead of src.m_vec.
2205 2018-11-06  Jan Hubicka  <jh@suse.cz>
2207         * tree.c (fld_simplified_type_name): Break out form ...
2208         (free_lang_data_in_type): ... here.
2209         (fld_type_variant_equal_p): Use it.
2211 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2213         * config/default-d.c: Include memmodel.h.
2215         * config/sol2-d.c: New file.
2216         * config/t-sol2 (sol2-d.o): New rule.
2217         * config.gcc <*-*-solaris2*>: Set d_target_objs,
2218         target_has_targetdm.
2220 2018-11-06  Jan Hubicka  <jh@suse.cz>
2222         * tree.c (fld_type_variant): Also copy alignment; be sure that
2223         new variant is equal.
2225 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2227         PR target/87762
2228         * config/s390/s390.md: Add relative_long attribute.
2230 2018-11-06  Jan Hubicka  <jh@suse.cz>
2232         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
2233         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
2234         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
2235         stream TYPE_NEEDS_CONSTRUCTING.
2236         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
2237         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
2239 2018-11-06  Richard Biener  <rguenther@suse.de>
2241         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
2242         dump-scope ...
2243         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
2245 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2247         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
2248         REG_CLASS_CONTENTS[GEN_REGS].
2249         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
2251 2018-11-06  Jan Hubicka  <jh@suse.cz>
2253          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
2255 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2257         PR sanitizer/80953
2258         * config/sol2.h (ASAN_CC1_SPEC): Define.
2259         (LD_WHOLE_ARCHIVE_OPTION): Define.
2260         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
2261         (ASAN_REJECT_SPEC): Provide default.
2262         (LIBASAN_EARLY_SPEC): Define.
2263         (LIBTSAN_EARLY_SPEC): Define.
2264         (LIBLSAN_EARLY_SPEC): Define.
2265         * config/i386/sol2.h (CC1_SPEC): Redefine.
2266         (ASAN_REJECT_SPEC): Define.
2268         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
2269         (TARGET_ASAN_SHADOW_OFFSET): Define.
2270         (sparc_asan_shadow_offset): New function.
2271         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
2272         (ASAN_REJECT_SPEC): Define.
2274 2018-11-06  Jan Hubicka  <jh@suse.cz>
2276         * tree.c (fld_type_variant): Copy canonical type.
2277         (fld_incomplete_type_of): Check that canonical types looks sane;
2278         copy canonical type.
2279         (verify_type): Accept when incomplete type has complete canonical type.
2281 2018-11-06  Jan Hubicka  <jh@suse.cz>
2283         * tree.c (free_lang_data): Reset overwite_assembler_name,
2284         print_xnode, print_decl, print_type and print_identifier of
2285         langhooks.
2287 2018-11-06  Richard Biener  <rguenther@suse.de>
2289         PR tree-optimization/87889
2290         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2291         Do nothing if old and new arg are the same
2293 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
2295         PR target/87723
2296         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
2297         attributes for operands 3 and 4.
2299 2018-11-06  Richard Biener  <rguenther@suse.de>
2301         PR middle-end/18041
2302         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
2303         matching bitfield insertion.
2305 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
2307         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
2308         (reg_next_debug_use): New.
2309         (attempt_change): Propagate adjusted expression into affected
2310         debug insns.
2311         (merge_in_block): Track uses in debug insns.
2312         (pass_inc_dec::execute): Allocate and release
2313         reg_next_debug_use.
2315 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
2317         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
2318         (_mm512_fixupimm_round_pd): Update parameters and builtin.
2319         (_mm512_maskz_fixupimm_round_pd): Ditto.
2320         (_mm512_fixupimm_round_ps): Ditto.
2321         (_mm512_maskz_fixupimm_round_ps): Ditto.
2322         (_mm_fixupimm_round_sd): Ditto.
2323         (_mm_maskz_fixupimm_round_sd): Ditto.
2324         (_mm_fixupimm_round_ss): Ditto.
2325         (_mm_maskz_fixupimm_round_ss): Ditto.
2326         (_mm512_fixupimm_pd): Ditto.
2327         (_mm512_maskz_fixupimm_pd): Ditto.
2328         (_mm512_fixupimm_ps): Ditto.
2329         (_mm512_maskz_fixupimm_ps): Ditto.
2330         (_mm_fixupimm_sd): Ditto.
2331         (_mm_maskz_fixupimm_sd): Ditto.
2332         (_mm_fixupimm_ss): Ditto.
2333         (_mm_maskz_fixupimm_ss): Ditto.
2334         (_mm512_mask_fixupimm_round_pd): Update builtin.
2335         (_mm512_mask_fixupimm_round_ps): Ditto.
2336         (_mm_mask_fixupimm_round_sd): Ditto.
2337         (_mm_mask_fixupimm_round_ss): Ditto.
2338         (_mm512_mask_fixupimm_pd): Ditto.
2339         (_mm512_mask_fixupimm_ps): Ditto.
2340         (_mm_mask_fixupimm_sd): Ditto.
2341         (_mm_mask_fixupimm_ss): Ditto.
2342         * config/i386/avx512vlintrin.h:
2343         (_mm256_fixupimm_pd): Update parameters and builtin.
2344         (_mm256_maskz_fixupimm_pd): Ditto.
2345         (_mm256_fixupimm_ps): Ditto.
2346         (_mm256_maskz_fixupimm_ps): Ditto.
2347         (_mm_fixupimm_pd): Ditto.
2348         (_mm_maskz_fixupimm_pd): Ditto.
2349         (_mm_fixupimm_ps): Ditto.
2350         (_mm_maskz_fixupimm_ps): Ditto.
2351         (_mm256_mask_fixupimm_pd): Update builtin.
2352         (_mm256_mask_fixupimm_ps): Ditto.
2353         (_mm_mask_fixupimm_pd): Ditto.
2354         (_mm_mask_fixupimm_ps): Ditto.
2355         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
2356         * config/i386/i386-builtin.def: Update builtin definitions.
2357         * config/i386/i386.c: Handle new builtin types and remove useless ones.
2358         * config/i386/sse.md: Update VFIXUPIMM* patterns.
2359         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2360         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2361         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
2362         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2363         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2364         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
2365         * config/i386/subst.md:
2366         (round_saeonly_sd_mask_operand4): Add new subst_attr.
2367         (round_saeonly_sd_mask_op4): Ditto.
2368         (round_saeonly_expand_operand5): Ditto.
2369         (round_saeonly_expand): Update.
2371 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
2373         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
2375 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
2377         PR rtl-optimization/87871
2378         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
2380 2018-11-05  Paul Koning  <ni1d@arrl.net>
2382         * doc/sourcebuild.texi (target attributes): Document new "inf"
2383         effective target keyword.
2385 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2387         * config/s390/s390.c (s390_register_move_cost): Increase costs for
2388         moves involving the CC reg.
2390 2018-11-05  Richard Biener  <rguenther@suse.de>
2392         PR tree-optimization/87873
2393         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
2394         argument.
2395         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
2396         * tree-vect-loop.c (vect_transform_loop): When splitting the
2397         loop exit also create forwarder PHIs for constants.
2398         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2399         Handle constant to_arg, add extra checking we match up the correct
2400         PHIs.
2402 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2404         * config/s390/s390.md: QImode and HImode for load on condition.
2406 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2408         * config/s390/predicates.md: Fix typo.
2409         * config/s390/s390.md: Allow immediates for load on condition.
2411 2018-11-05  Martin Liska  <mliska@suse.cz>
2413         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
2414         * bitmap.h (struct bitmap_usage): Likewise.
2415         * ggc-common.c (SCALE): Remove.
2416         (LABEL): Likewise.
2417         (struct ggc_usage): Use SIZE_AMOUNT. And update
2418         compare method.
2419         * ggc-page.c (SCALE): Remove.
2420         (STAT_LABEL): Remove.
2421         (ggc_print_statistics): Use SIZE_AMOUNT.
2422         * gimple.h (SCALE): Remove.
2423         (LABEL): Likewise.
2424         * input.c (ONE_K): Remove.
2425         (ONE_M): Likewise.
2426         (SCALE): Likewise.
2427         (STAT_LABEL): Likewise.
2428         (FORMAT_AMOUNT): Likewise.
2429         (dump_line_table_statistics): Use SIZE_AMOUNT.
2430         * mem-stats.h (struct mem_usage): Likewise.
2431         * rtl.c (dump_rtx_statistics): Likewise.
2432         (rtx_alloc_counts): Change type to size_t.
2433         (rtx_alloc_sizes): Likewise.
2434         (rtx_count_cmp): New.
2435         (dump_rtx_statistics): Sort first based on counts.
2436         * tree.c (tree_nodes_cmp): New.
2437         (tree_codes_cmp): New.
2438         (dump_tree_statistics): Sort first based on counts.
2439         * system.h (ONE_K): New.
2440         (ONE_M): Likewise.
2441         (SIZE_SCALE): Likewise.
2442         (SIZE_LABEL): Likewise.
2443         (SIZE_AMOUNT): Likewise.
2444         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
2445         * tree-dfa.c (dump_dfa_stats): Likewise.
2446         * tree-phinodes.c (phinodes_print_statistics): Likewise.
2447         * tree-ssanames.c (ssanames_print_statistics): Likewise.
2448         * tree.c (dump_tree_statistics): Likewise.
2449         * vec.c (struct vec_usage): Likewise.
2450         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
2451         trigger a -Werror=format-overflow with
2452         --enable-gather-detailed-stats.
2454 2018-11-05  Martin Liska  <mliska@suse.cz>
2456         * mem-stats.h (mem_alloc_description::release_instance_overhead):
2457         Return T *.
2458         * vec.c (struct vec_usage): Register m_element_size.
2459         (vec_prefix::register_overhead): New arguments: elements and
2460         element_size.
2461         (vec_prefix::release_overhead): Subtract elements.
2462         * vec.h (struct vec_prefix): Change signature.
2463         (va_heap::reserve): Pass proper arguments.
2464         (va_heap::release): Likewise.
2466 2018-11-05  Martin Liska  <mliska@suse.cz>
2468         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
2469         style.
2470         * vec.c: Likewise.
2472 2018-11-05  Richard Biener  <rguenther@suse.de>
2474         * tree-scalar-evolution.h (final_value_replacement_loop): Update
2475         prototype.
2476         * tree-scalar-evolution.c (final_value_replacement_loop): Return
2477         whether anything was done.
2478         (scev_const_prop): Remove constant propagation part, fold
2479         remains into ...
2480         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
2481         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
2482         conditionally.
2484 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
2486         PR tree-optimization/87859
2487         * gimple-ssa-store-merging.c (struct merged_store_group): Add
2488         only_constants and first_nonmergeable_order members.
2489         (merged_store_group::merged_store_group): Initialize them.
2490         (merged_store_group::do_merge): Clear only_constants member if
2491         adding something other than INTEGER_CST store.
2492         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
2493         stores with order >= first_nonmergeable_order.  Use
2494         merged_store->only_constants instead of always recomputing it.
2495         Set merged_store->first_nonmergeable_order if we've skipped any
2496         stores.  Attempt to merge overlapping INTEGER_CST stores that
2497         we would otherwise skip.
2499         PR sanitizer/87837
2500         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
2502 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
2504         PR target/87853
2505         * config/i386/emmintrin.h (__v16qs): New to cope with option
2506         -funsigned-char.
2507         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
2508         (_mm_cmplt_epi8): Likewise.
2509         (_mm_cmpgt_epi8): Likewise.
2511 2018-11-05  Richard Biener  <rguenther@suse.de>
2513         PR rtl-optimization/87852
2514         * fwprop.c (use_killed_between): Only consider single-defs of the
2515         use whose definition statement dominates the use.
2517 2018-11-05  Martin Liska  <mliska@suse.cz>
2519         PR web/87829
2520         * doc/invoke.texi: Remove options that are
2521         not disabled with -Os.
2523 2018-11-05  Martin Liska  <mliska@suse.cz>
2525         PR c/87811
2526         * doc/extend.texi: Update constrain about the last argument
2527         of __builtin_expect_with_probability.
2529 2018-11-05  Martin Liska  <mliska@suse.cz>
2531         PR c/87811
2532         * predict.c (expr_expected_value_1): Verify
2533         that last argument is a real constants and emit
2534         error.
2536 2018-11-05  Martin Liska  <mliska@suse.cz>
2538         PR gcov-profile/77698
2539         * ipa-profile.c (ipa_profile): Adjust hotness threshold
2540         only in LTO mode.
2542 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2544         PR tree-optimization/86572
2545         * builtins.c (c_strlen): Handle negative offsets in a safe way.
2547 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2549         PR tree-optimization/87672
2550         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
2551         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
2553 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
2555         PR middle-end/58372
2556         * cfgexpand.c (pass_expand::execute): Move the call to
2557         finish_eh_generation in front of the call to expand_stack_alignment.
2559 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
2561         * common/config/i386/i386-common.c (processor_alias_table): Add
2562         znver2 entry.
2563         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
2564         (case ${target}): Add znver2.
2565         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
2566         -march=native recognize znver2 processors.
2567         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
2568         * config/i386/i386.c (m_znver2): New definition.
2569         (m_ZNVER): New definition.
2570         (m_AMD_MULTIPLE): Includes m_znver2.
2571         (processor_cost_table): Add znver2 entry.
2572         (processor_target_table): Add znver2 entry.
2573         (get_builtin_code_for_version): Set priority for
2574         PROCESSOR_ZNVER2.
2575         (processor_model): Add M_AMDFAM17H_ZNVER2.
2576         (arch_names_table): Ditto.
2577         (ix86_reassociation_width): Include znver2.
2578         * config/i386/i386.h (TARGET_znver2): New definition.
2579         (struct ix86_size_cost): Add TARGET_ZNVER2.
2580         (enum processor_type): Add PROCESSOR_ZNVER2.
2581         * config/i386/i386.md (define_attr "cpu"): Add znver2.
2582         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
2583         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
2584         (ix86_adjust_cost): Add znver2.
2585         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
2586         * gcc/doc/extend.texi: Add details about znver2.
2587         * gcc/doc/invoke.texi: Add details about znver2.
2589 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
2591         PR target/87079
2593         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
2594         pattern.
2596 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
2598         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
2599         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
2601 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
2603         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
2604         shrink-wrapping when -mtrack-speculation.
2606 2018-11-02  Richard Biener  <rguenther@suse.de>
2608         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
2609         (pop_cost_one_pair): Do not free pair.
2610         (pop_best_coalesce): Likewise.
2611         (create_coalesce_list): Initialize obstack.
2612         (delete_coalesce_list): Free obstack.
2613         (find_coalesce_pair): Obstack-allocate coalesce pairs.
2614         (add_cost_one_coalesce): Likewise.
2615         (struct live_track): Remove bitmap pointer indirections.
2616         (new_live_track): Adjust.
2617         (delete_live_track): Likewise.
2618         (live_track_remove_partition): Likewise.
2619         (live_track_add_partition): Likewise.
2620         (live_track_live_p): Likewise.
2621         (live_track_process_def): Likewise.
2622         (live_track_clear_base_vars): Likewise.
2624 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2626         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
2627         * configure: Regenerate.
2628         * config.in: Regenerate.
2629         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
2630         HAVE_AS_REGISTER_PSEUDO_OP guard.
2631         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
2632         !HAVE_AS_REGISTER_PSEUDO_OP support.
2633         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
2634         guard.
2636 2018-11-02  Richard Biener  <rguenther@suse.de>
2638         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
2639         with is_type_die.
2641 2018-11-02  Richard Biener  <rguenther@suse.de>
2643         PR tree-optimization/87776
2644         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
2645         executable when iterating but running into rpo-vn-max-loop-depth
2646         and not eliding the iteration.
2648 2018-11-30  Jan Hubicka  <jh@suse.cz>
2650         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
2651         TYPE_DECL.
2653 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
2655         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
2656         prototype.
2657         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
2658         Combine with rs6000_address_for_fpconvert.
2659         (rs6000_address_for_fpconvert): Combine with
2660         rs6000_force_indexed_or_indirect_mem.
2661         (rs6000_expand_vector_init): Change function call from
2662         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2663         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
2664         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2665         (floatsi<mode>2_lfiwax_mem): Ditto.
2666         (floatunssi<mode>2_lfiwzx): Ditto.
2667         (floatunssi<mode>2_lfiwzx_mem): Ditto.
2668         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
2669         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
2670         (fix_trunc<mode>si2_stfiwx): Ditto.
2671         (fixuns_trunc<mode>si2_stfiwx): Ditto.
2672         (float_<mode>si2_hw): Ditto.
2673         (floatuns_<mode>si2_hw): Ditto.
2674         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
2675         (vsx_splat_<mode>): Ditto.
2677 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
2679         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
2680         at end of assembler input text.
2681         * configure: Regenerate.
2683 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
2685         PR tree-optimization/87826
2686         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
2687         negative or larger or equal to type's precision.
2689 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
2691         * opts.c (default_options_table): Do not enable
2692         OPT_fdelayed_branch at -Og.
2693         * doc/invoke.texi (-fdelayed-branch): Document it.
2695 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2697         * optabs-libfuncs.c (build_libfunc_function_visibility):
2698         New, split out from...
2699         (build_libfunc_function): ... here.
2700         (init_one_libfunc_visibility): New, split out from ...
2701         (init_one_libfunc): ... here.
2703         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
2704         scratch register need not be early-clobber.  Document the reason
2705         why we cannot use ST<OP>.
2707 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
2709         PR bootstrap/82856
2710         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
2711         line for second argument of AC_DEFINE_UNQUOTED.
2712         * doc/install.texi (Tools/packages necessary for modifying GCC):
2713         Update to autoconf 2.69 and automake 1.15.1.
2714         * aclocal.m4, config.in, configure: Regenerate.
2716 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
2718         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
2719         initialization.
2721 2018-10-31  Martin Liska  <mliska@suse.cz>
2723         PR driver/83193
2724         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
2725         Add new argument invalid_extension.
2726         (aarch64_get_all_extension_candidates): New function.
2727         (aarch64_rewrite_selected_cpu): Add NULL to function call.
2728         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
2729         new argument.
2730         (aarch64_get_all_extension_candidates): New function.
2731         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
2732         argument invalid_extension.
2733         (aarch64_parse_cpu): Likewise.
2734         (aarch64_print_hint_for_extensions): New function.
2735         (aarch64_validate_mcpu): Provide hint about invalid extension.
2736         (aarch64_validate_march): Likewise.
2737         (aarch64_handle_attr_arch): Pass new argument.
2738         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
2739         (aarch64_handle_attr_isa_flags): Likewise.
2741 2018-10-31  Richard Biener  <rguenther@suse.de>
2743         PR middle-end/70359
2744         PR middle-end/86270
2745         * tree-outof-ssa.c (insert_backedge_copies): Restrict
2746         copy generation to useful cases.  Place the copy before
2747         the definition of the backedge value when possible.
2749 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2751         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
2752         * config/arc/arc.c (arc_active_insn): New function.
2753         (check_store_cacheline_hazard): Likewise.
2754         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
2755         (arc_override_options): Disable delay slot scheduler for older
2756         A7.
2757         (arc_store_addr_hazard_p): New implementation, old one renamed to
2758         ...
2759         (arc_store_addr_hazard_internal_p): Renamed.
2760         (arc_reorg): Don't combine into brcc instructions which are part
2761         of hardware hazard solution.
2762         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
2763         (tune_arc700): Likewise.
2764         * config/arc/arc.opt (arc7xx): New tune value.
2765         * config/arc/arc700.md: Improve A7 scheduler.
2767 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2769         * config/arc/arc.c (arc_override_options): Remove
2770         TARGET_COMPACT_CASESI.
2771         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
2772         (CASE_VECTOR_MODE): Likewise.
2773         (CASE_VECTOR_PC_RELATIVE): Likewise.
2774         (CASE_VECTOR_SHORTEN_MODE): Likewise.
2775         (CASE_VECTOR_SHORTEN_MODE1): Delete.
2776         (ADDR_VEC_ALIGN): Update.
2777         (ASM_OUTPUT_CASE_LABEL): Undefine.
2778         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
2779         (TARGET_BI_BIH): Define.
2780         (DEFAULT_BRANCH_INDEX): Likewise.
2781         * config/arc/arc.md (casesi): Rework to accept BI/BIH
2782         instructions, remove compact_casesi use case.
2783         (casesi_compact_jump): Remove.
2784         (casesi_dispatch): New pattern.
2785         * config/arc/arc.opt: Add mbranch-index option. Deprecate
2786         compact_casesi option.
2787         * doc/invoke.texi: Document mbranch-index option.
2789 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2791         * config/arc/arc.c (arc_get_tp): Remove function.
2792         (arc_emit_call_tls_get_addr): Likewise.
2793         (arc_call_tls_get_addr): New function.
2794         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
2795         * config/arc/arc.md (tls_load_tp_soft): Remove.
2796         (tls_gd_get_addr): Likewise.
2798 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2800         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
2801         (mulsi3_600_lib): Remove pattern.
2802         (umulsi3_highpart_600_lib_le): Likewise.
2803         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
2804         (umulsidi3): Remove call to umulsidi3_600_lib.
2805         (umulsidi3_600_lib): Remove pattern.
2806         (peephole2): Remove peephole using the above deprecated patterns.
2808 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
2810         PR target/87374
2811         * config/arm/arm.c (arm_option_check_internal): Disable the combined
2812         use of -mslow-flash-data and -mword-relocations.
2813         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
2814         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
2815         flag_pic.
2816         * doc/invoke.texi (-mword-relocations): Mention conflict with
2817         -mslow-flash-data.
2818         (-mslow-flash-data): Reciprocally.
2820 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2822         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
2823         16-byte modes held in GP registers to use an even regno.
2825         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
2826         (aarch64_atomic_ldop_supported_p): Remove.
2827         (aarch64_gen_atomic_ldop): Remove.
2828         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
2829         Fully expand LSE operations here.
2830         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
2831         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
2832         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
2833         and use ATOMIC_LDOP instead; use register_operand for the input;
2834         drop the split and emit insns directly.
2835         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
2836         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
2837         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
2839         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
2840         (aarch64_gen_atomic_ldop): Don't call it.
2841         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
2842         Use aarch64_reg_or_zero.
2843         (aarch64_atomic_exchange<ALLI>): Likewise.
2844         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
2845         operand 0; use aarch64_reg_or_zero for input; merge ...
2846         (@aarch64_atomic_swp<ALLI>): ... this and remove.
2848         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
2849         (aarch64_split_compare_and_swap): Use it.
2850         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
2851         test oldval against the proper predicate.
2852         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
2853         Use nonmemory_operand for expected.
2854         (cas_short_expected_pred): New.
2855         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
2856         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
2857         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
2858         (aarch64_plushi_operand): New.
2860         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
2861         Force oldval into the rval register for TARGET_LSE; emit the compare
2862         during initial expansion so that it may be deleted if unused.
2863         (aarch64_gen_atomic_cas): Remove.
2864         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
2865         Change =&r to +r for operand 0; use match_dup for operand 2;
2866         remove is_weak and mod_f operands as unused.  Drop the split
2867         and merge with...
2868         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
2869         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
2870         (@aarch64_atomic_cas<GPI>): Similarly.
2872 2018-10-31  Richard Biener  <rguenther@suse.de>
2874         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
2875         using ABSU_EXPR.
2877 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2879         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
2880         * config/aarch64/aarch64.md: Include saphira.md
2881         * config/aarch64/saphira.md: New file for pipeline description.
2883 2018-10-30  Martin Sebor  <msebor@redhat.com>
2885         PR middle-end/87041
2886         * gimple-ssa-sprintf.c (format_directive): Use %G to include
2887         inlining context.
2888         (sprintf_dom_walker::compute_format_length):
2889         Avoid setting POSUNDER4K here.
2890         (get_destination_size): Handle null argument values.
2891         (get_user_idx_format): New function.
2892         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
2893         functions, including user-defined with attribute format printf.
2894         Use %G to include inlining context.
2895         Set POSUNDER4K here.
2897 2018-10-30  Jan Hubicka  <jh@suse.cz>
2899         * params.def (lto-partitions): Bump from 32 to 128.
2901 2018-10-30  Jan Hubicka  <jh@suse.cz>
2903         * tree.c
2904         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
2905         head in file.
2906         (free_lang_data_in_type): Forward declare.
2907         (fld_type_variant_equal_p): New function.
2908         (fld_type_variant): New function
2909         (fld_incomplete_types): New hash.
2910         (fld_incomplete_type_of): New function
2911         (fld_simplfied-type): New function.
2912         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
2913         (free_lang_data): Allocate and free fld_incomplete_type; update call
2914         of free_lang_data_in_decl.
2916 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
2918         * gcov.c (output_lines): Remove duplicate line.
2920 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
2922         * config/rs6000/rs6000.md (bswapdi2): Force address into register
2923         if not in indexed or indirect form.
2924         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
2925         (bswapdi2_store): Ditto.
2926         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
2927         helper function.
2928         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
2929         Prototype for helper function.
2931 2018-10-30  Martin Sebor  <msebor@redhat.com>
2933         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
2934         (target, pragma GCC optimize, pragma GCC target): Ditto.
2936 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
2938         * doc/extend.texi: Fix prototype and description of
2939         __builtin_expect_with_probability.
2941 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2943         * cgraph.h (clone_function_name_1): Replaced by new
2944           clone_function_name_numbered that takes name as string; for
2945           privatize_symbol_name_1 use only.
2946           (clone_function_name): Renamed to
2947           clone_function_name_numbered to be explicit about numbering.
2948           (clone_function_name): New two-argument function that does
2949           not number its output.
2950           (clone_function_name): New three-argument function that
2951           takes a number to append to its output.
2952         * cgraphclones.c (duplicate_thunk_for_node):
2953           (clone_function_name_1): Renamed.
2954           (clone_function_name_numbered): Two new functions.
2955           (clone_function_name): Improved documentation.
2956           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
2957         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
2958         * final.c (final_scan_insn_1): Use the new clone_function_name
2959           without numbering.
2960         * multiple_target.c (create_dispatcher_calls): Ditto.
2961           (create_target_clone): Ditto.
2962         * omp-expand.c (grid_expand_target_grid_body): Ditto.
2963         * omp-low.c (create_omp_child_function_name): Ditto.
2964         * omp-simd-clone.c (simd_clone_create): Ditto.
2965         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
2966           new clone_function_name without numbering.
2968 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
2970         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
2971         Assert that the allocation size is not zero.
2973 2018-10-30  Richard Biener  <rguenther@suse.de>
2975         PR tree-optimization/87800
2976         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
2977         non-induction or reduction PHIs.
2979 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
2981         * config/aarch64/falkor-tag-collision-avoidance.c
2982         (execute_tag_collision_avoidance): Call df_note_add_problem.
2984 2018-10-30  Martin Liska  <mliska@suse.cz>
2986         * doc/extend.texi: Fix typo in documentation
2987         of __builtin_expect_with_probability.
2989 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2991         PR c++/87721
2992         * input.c (get_substring_ranges_for_loc): Detect if
2993         linemap_resolve_location gives us a NULL map, and reject
2994         this case.
2996 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
2998         * config.gcc (xstormy16-*-elf): Set tm_d_file.
3000 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
3001             Martin Sebor  <msebor@redhat.com>
3002             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3004         * doc/ux.texi (Quoting): New subsection, adapted from material at
3005         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
3006         MartinSebor and ManuelLopezIbanez.
3007         (Fix-it hints): Note that fix-it hints shouldn't be marked for
3008         translation.
3010 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
3012         PR middle-end/87469
3013         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
3014         max value.
3016 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
3018         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
3020 2018-10-29  Paul Koning  <ni1d@arrl.net>
3022         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
3024 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
3026         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
3027         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
3028         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
3029         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
3030         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
3031         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
3032         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
3033         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
3034         Change 'vector' to '__vector'.
3035         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
3036         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
3037         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
3038         _mm_avg_pu16): Likewise.  And, whitespace corrections.
3040 2018-10-29  Richard Biener  <rguenther@suse.de>
3042         PR tree-optimization/87785
3043         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
3044         internal defs.
3046 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3048         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
3050 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3052         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
3053         ports configuration macro, defaults to "gnu".
3054         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
3055         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
3057 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3059         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
3061 2018-10-29  Richard Biener  <rguenther@suse.de>
3063         PR tree-optimization/87790
3064         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
3065         (vect_make_slp_decision): Adjust.
3066         (vect_slp_analyze_bb_1): Likewise.
3067         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
3068         edges.
3070 2018-10-29  Richard Biener  <rguenther@suse.de>
3072         PR tree-optimization/87785
3073         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
3074         and processing.
3075         (vect_build_slp_tree): Likewise.
3076         (vect_gather_slp_loads): New function.
3077         (vect_analyze_slp_instance): Gather loads separately from the
3078         SLP tree build.
3080 2018-10-29  Martin Liska  <mliska@suse.cz>
3082         * Makefile.in: Make dependency to json.o.
3083         * doc/gcov.texi: Document new JSON format, remove
3084         old intermediate format documentation.
3085         * gcov.c (struct function_info): Come up with m_name and
3086         m_demangled_name.
3087         (function_info::function_info): Initialize it.
3088         (function_info::~function_info): Release it.
3089         (main): Rename flag_intermediate_format to flag_json_format.
3090         (print_usage): Describe --json-format.
3091         (process_args): Set flag_json_format.
3092         (output_intermediate_line): Remove.
3093         (output_intermediate_json_line): Likewise.
3094         (get_gcov_intermediate_filename): Return new extension
3095         ".gcov.json.gz".
3096         (output_intermediate_file): Implement JSON emission.
3097         (output_json_intermediate_file): Implement JSON emission.
3098         (generate_results): Use ::get_name for function name.
3099         Handle JSON output file.
3100         (read_graph_file): Use ::get_name instead of cplus_demangle.
3101         (read_count_file): Likewise.
3102         (solve_flow_graph): Likewise.
3103         (add_line_counts): Likewise.
3104         (accumulate_line_counts): Use new flag_json_format.
3105         (output_function_details): Use ::get_name instead of cplus_demangle.
3106         (output_lines): Likewise.
3107         * json.cc (test_writing_literals): Add new tests.
3108         * json.h (class literal): Add new boolean constructor.
3110 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
3112         PR rtl-optimization/87701
3113         PR rtl-optimization/87780
3114         * combine.c (make_more_copies): Rewrite.
3116 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
3118         * doc/generic.texi (ABSU_EXPR): Document.
3119         * match.pd (absu(x)*absu(x) -> x*x): Handle.
3120         (absu(absu(X)) -> absu(X)): Likewise.
3121         (absu(-X) -> absu(X)): Likewise.
3122         (absu(X)  where X is nonnegative -> X): Likewise.
3124 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
3126         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
3127         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
3128         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
3129         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
3130         (s-tm-texi): Also check timestamp on d-target.def.
3131         (generated_files): Add TM_D_H and d-target-hooks-def.h.
3132         (build/genhooks.o): Also depend on D_TARGET_DEF.
3133         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
3134         variables.
3135         * config/aarch64/aarch64-d.c: New file.
3136         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
3137         Define.
3138         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
3139         prototype.
3140         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
3141         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
3142         * config/arm/arm-d.c: New file.
3143         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
3144         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
3145         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3146         * config/arm/t-arm (arm-d.o): New rule.
3147         * config/default-d.c: New file.
3148         * config/glibc-d.c: New file.
3149         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3150         * config/i386/i386-d.c: New file.
3151         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
3152         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
3153         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3154         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
3155         * config/i386/t-i386 (i386-d.o): New rule.
3156         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3157         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3158         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
3159         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3160         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3161         * config/mips/mips-d.c: New file.
3162         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
3163         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
3164         * config/mips/t-mips (mips-d.o): New rule.
3165         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3166         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3167         * config/powerpcspe/powerpcspe-d.c: New file.
3168         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
3169         New prototype.
3170         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3171         Support GNU D by using 0 as the language type.
3172         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
3173         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
3174         * config/riscv/riscv-d.c: New file.
3175         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
3176         prototype.
3177         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
3178         * config/riscv/t-riscv (riscv-d.o): New rule.
3179         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3180         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3181         * config/rs6000/rs6000-d.c: New file.
3182         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
3183         prototype.
3184         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3185         Support GNU D by using 0 as the language type.
3186         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
3187         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
3188         * config/s390/s390-d.c: New file.
3189         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
3190         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
3191         * config/s390/t-s390 (s390-d.o): New rule.
3192         * config/sparc/sparc-d.c: New file.
3193         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
3194         prototype.
3195         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
3196         * config/sparc/t-sparc (sparc-d.o): New rule.
3197         * config/t-glibc (glibc-d.o): New rule.
3198         * configure: Regenerated.
3199         * configure.ac (tm_d_file): New variable.
3200         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
3201         * doc/contrib.texi (Contributors): Add self for the D frontend.
3202         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
3203         * doc/install.texi (Configuration): Mention libphobos as an option for
3204         --enable-shared.  Mention d as an option for --enable-languages.
3205         (Testing): Mention check-d as a target.
3206         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
3207         name suffixes.  Mention d as a -x option.
3208         * doc/sourcebuild.texi (Top Level): Mention libphobos.
3209         * doc/standards.texi (Standards): Add section on D language.
3210         * doc/tm.texi: Regenerated.
3211         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
3212         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
3213         * dwarf2out.c (is_dlang): New function.
3214         (gen_compile_unit_die): Use DW_LANG_D for D.
3215         (declare_in_namespace): Return module die for D, instead of adding
3216         extra declarations into the namespace.
3217         (gen_namespace_die): Generate DW_TAG_module for D.
3218         (gen_decl_die): Handle CONST_DECLSs for D.
3219         (dwarf2out_decl): Likewise.
3220         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
3221         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
3222         kinds of aggregates.
3223         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
3224         * genhooks.c: Include d/d-target.def.
3226 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
3228         PR target/85669
3229         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
3230         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
3231         (STACK_DYNAMIC_OFFSET): Likewise.
3233 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
3235         PR target/80024
3236         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
3237         error message.
3239 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
3241         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
3242         return type and other typos.
3244 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
3246         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
3247         a shorter sequence with fewer branches.
3248         (emit_final_str_compare_gpr): Ditto.
3250 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3252         * config/rs6000/tmmintrin.h: New file.
3253         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
3255 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3257         * config/rs6000/mmintrin.h: Enable 32bit compilation.
3258         * config/rs6000/xmmintrin.h: Likewise.
3260 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3262         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
3264 2018-10-26  Richard Biener  <rguenther@suse.de>
3266         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
3267         and wrapper.
3268         (vect_mark_slp_stmts_relevant): Likewise.
3269         (vect_detect_hybrid_slp_stmts): Likewise.
3270         (vect_bb_slp_scalar_cost): Likewise.
3271         (vect_remove_slp_scalar_calls): Likewise.
3273 2018-10-26  Jan Hubicka  <jh@suse.cz>
3275         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
3276         (warn_types_mismatch): Fix walk of DECL_NAME.
3277         (odr_types_equivalent_p): Fix overactive assert.
3279 2018-10-26  Richard Biener  <rguenther@suse.de>
3281         PR tree-optimization/87105
3282         * tree-vectorizer.h (_slp_tree::refcnt): New member.
3283         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
3284         refcnt.
3285         (vect_create_new_slp_node): Initialize refcnt to one.
3286         (bst_traits): Move.
3287         (scalar_stmts_set_t, bst_fail): Remove.
3288         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
3289         (vect_build_slp_tree): Add bst_map argument and lookup
3290         already created SLP nodes.
3291         (vect_print_slp_tree): Handle a SLP graph, print SLP node
3292         addresses.
3293         (vect_slp_rearrange_stmts): Handle a SLP graph.
3294         (vect_analyze_slp_instance): Adjust and free SLP nodes from
3295         the CSE map.  Fix indenting.
3296         (vect_schedule_slp_instance): Add short-cut.
3298 2018-10-26  Martin Liska  <mliska@suse.cz>
3300         PR testsuite/86158
3301         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
3302         addr_expr and not with pointers.
3304 2018-10-26  Jan Hubicka  <jh@suse.cz>
3306         * tree.c (free_lang_data_in_type): Only check main variants.
3307         * ipa-devirt.c (warn_odr): Make static.
3308         (types_same_for_odr): Drop strict variant.
3309         (types_odr_comparable): Likewise.
3310         (odr_or_derived_type_p): Look for main variants.
3311         (odr_name_hasher::equal): Cleanup comment.
3312         (odr_subtypes_equivalent): Add warn and warned arguments; check main
3313         variants.
3314         (type_variants_equivalent_p): break out from ...
3315         (odr_types_equivalent): ... here; go for main variants where needed.
3316         (warn_odr): ... here; turn static.
3317         (warn_types_mismatch): Compare mangled names of main variants.
3318         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
3319         (type_with_linkage_p): Sanity check that we look at main variant.
3320         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
3321         * tree.h (types_same_for_odr): Drop strict argument.
3323 2018-10-26  Richard Biener  <rguenther@suse.de>
3325         PR tree-optimization/87746
3326         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
3327         Simplify and fix WRT strided store groups with size not
3328         equal to step in element count.
3329         (vect_analyze_group_access_1): Dump the whole group.
3331 2018-10-25  Carl Love  <cel@us.ibm.com>
3333         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
3334         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
3335         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
3336         precicion entry for each overloaded builtin.
3337         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
3338         VSCEDPUO): Rename overloaded name.
3339         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
3340         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
3341         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
3342         define_expand for xscmpexqp instruction.
3343         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
3345 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3346             Jinsong Ji <jji@us.ibm.com>
3348         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
3349         function with vec_sl.
3350         (_mm_slli_epi32): Likewise.
3351         (_mm_slli_epi64): Likewise.
3352         (_mm_srai_epi16): Replace deprecated function with vec_sra.
3353         (_mm_srai_epi32): Likewise.
3354         (_mm_srli_epi16): Replace deprecated function with vec_sr.
3355         (_mm_srli_epi32): Likewise.
3356         (_mm_srli_epi64): Likewise.
3357         (_mm_sll_epi16): Replace deprecated function with vec_sl.
3358         (_mm_sll_epi32): Likewise.
3359         (_mm_sll_epi64): Likewise.
3360         (_mm_sra_epi16): Replace deprecated function with vec_sra.
3361         (_mm_sra_epi32): Likewise.
3362         (_mm_srl_epi16): Replace deprecated function with vec_sr.
3363         (_mm_srl_epi32): Likewise.
3364         (_mm_srl_epi64): Likewise.
3366 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3367             Jinsong Ji <jji@us.ibm.com>
3369         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
3370         comparison operators with vec_cmp* for compatibility due to
3371         unfortunate history; clean up formatting and use types more
3372         appropriately.
3373         (_mm_sll_epi32): Likewise.
3374         (_mm_sll_epi64): Likewise.
3375         (_mm_srl_epi16): Likewise.
3376         (_mm_srl_epi32): Likewise.
3377         (_mm_srl_epi64): Likewise.
3379 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3380             Jinsong Ji <jji@us.ibm.com>
3382         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
3383         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
3384         __vector __bool int.  Use vec_cmpgt in preference to deprecated
3385         function vec_vcmpgtfp.
3386         (_mm_max_ps): Likewise.
3388 2018-10-25  Jeff Law  <law@redhat.com>
3390         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
3391         if returning false.
3393 2018-10-25  Martin Sebor  <msebor@redhat.com>
3395         * doc/extend.texi (aligned): Expand attribute description.
3396         (Alignment): Rename section.  Discuss function arguments.
3398 2018-10-25  Jan Hubicka  <jh@suse.cz>
3400         * ipa-devirt.c (main_odr_variant): Remove.
3401         (hash_odr_name, types_same_for_odr, types_odr_comparable,
3402         odr_name_hasher::equal, odr_subtypes_equivalent_p):
3403         Drop use of main_odr_variant.
3404         (add_type_duplicate): Silence confused warnings on integer types.
3405         (get_odr_type): Always look for main variant.
3406         (register_odr_type): Simplify.
3408 2018-10-25  Richard Biener  <rguenther@suse.de>
3410         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
3411         Initialize ng to silence error with release checking bootstrap.
3413 2018-10-25  Richard Biener  <rguenther@suse.de>
3415         * tree-if-conv.c: Include tree-ssa-sccvn.h.
3416         (tree_if_conversion): Run CSE on the if-converted loop body.
3418 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3420         * config/s390/constraints.md (ZL): New constraint.
3421         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
3422         operands.
3423         * config/s390/s390.md (movdi_larl): Remove.
3424         (movdi_64): Add the LARL alternative.
3426 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3428         PR bootstrap/87747
3429         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
3430         (RTX_CODE_HWINT_P): New macro.
3431         (rtx_code_size): Use RTX_CODE_HWINT_P ().
3433 2018-10-25  Jan Hubicka  <jh@suse.cz>
3435         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
3436         is anonymous.
3438 2018-10-25  Richard Biener  <rguenther@suse.de>
3440         PR tree-optimization/87665
3441         PR tree-optimization/87745
3442         * tree-vectorizer.h (get_earlier_stmt): Remove.
3443         (get_later_stmt): Pick up UID from the original non-pattern stmt.
3445 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
3447         * options.texi (Deprecated): Move list to Var section.
3449 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
3450             Jinsong Ji <jji@us.ibm.com>
3452         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
3453         __vector long to __vector long long.
3454         (_mm_cvtpd_ps): Likewise.
3455         (_mm_cvttpd_epi32): Likewise.
3456         (_mm_cvtpi32_pd): Likewise.
3457         (_mm_unpackhi_epi64): Likewise.
3458         (_mm_unpacklo_epi64): Likewise.
3460 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
3462         PR rtl-optimization/87720
3463         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
3465 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
3467         * gimple-ssa-isolate-paths.c
3468         (find_implicit_erroneous_behavior): Do not change code if the
3469         pass is running for warnings only.
3470         (find_explicit_erroneous_behavior): Likewise.
3472 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
3474         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
3475         Define as rs6000_mangle_decl_assembler_name.
3476         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
3477         long double to IEEE long double, switch the names of the long
3478         double built-in functions to be <func>f128 instead of <func>l.
3480 2018-10-24  Martin Sebor  <msebor@redhat.com>
3482         * doc/extend.texi (nonnull): List no-argument form.  Reference
3483         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
3485 2018-10-24  Richard Biener  <rguenther@suse.de>
3487         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
3489 2018-10-24  Martin Liska  <mliska@suse.cz>
3491         PR tree-optimization/84436
3492         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
3493         Remove.
3494         (switch_conversion::contains_linear_function_p): New.
3495         (switch_conversion::build_one_array): Support linear
3496         transformation on input.
3497         * tree-switch-conversion.h (struct switch_conversion): Add
3498         contains_linear_function_p declaration.
3500 2018-10-24  Richard Biener  <rguenther@suse.de>
3502         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
3503         if its argument is CONSTANT_CLASS_P.
3505 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
3507         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
3508         it is wrong for forward declarations.
3510 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3512         * config/s390/s390.c (s390_check_qrst_address): Add the missing
3513         SYMBOL_REF_P () check.
3515 2018-10-24  Richard Biener  <rguenther@suse.de>
3517         PR tree-optimization/87105
3518         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
3519         dump classification.
3520         (vect_analyze_data_ref_accesses): Handle duplicate loads and
3521         stores by splitting the affected group after the fact.
3522         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
3523         fail the SLP build because of size constraints.
3525 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3527         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
3528         * configure: Regenerate.
3529         * config.in: Regenerate.
3530         * varasm.c (mergeable_string_section): Use readonly_data_section
3531         if linker doesn't support SHF_MERGE with alignment > 8.
3532         (mergeable_constant_section): Likewise.
3534 2018-10-24  Richard Biener  <rguenther@suse.de>
3536         PR tree-optimization/84013
3537         * tree-ssa-structalias.c (struct msdi_data): New struct for
3538         marshalling data to walk_stmt_load_store_ops.
3539         (maybe_set_dependence_info): Refactor as callback for
3540         walk_stmt_load_store_ops.
3541         (compute_dependence_clique): Set restrict info on all stmt kinds.
3543 2018-10-24  Martin Liska  <mliska@suse.cz>
3545         * cgraph.c (cgraph_node::dump):
3546         Remove reduntant dumps and make tp_first_run dump more compact.
3548 2018-10-24  Richard Biener  <rguenther@suse.de>
3550         PR tree-optimization/87665
3551         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
3552         to reflect reality.
3554 2018-10-12  Jeff Law  <law@redhat.com>
3556         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
3557         for H8/S.
3559 2018-10-23  Richard Biener  <rguenther@suse.de>
3561         * tree-vrp.c (add_assert_info): Guard dump_printf with
3562         dump_enabled_p.
3563         * gimple-ssa-evrp-analyze.c
3564         (evrp_range_analyzer::record_ranges_from_incoming_edge):
3565         Use value_range::ignore_equivs_equal_p.
3567 2018-10-23  Richard Biener  <rguenther@suse.de>
3569         PR tree-optimization/87105
3570         PR tree-optimization/87608
3571         * passes.def (pass_all_early_optimizations): Add early phi-opt
3572         after dce.
3573         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
3574         addition to debug stmts.
3575         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
3576         and abs replacement early.
3577         * tree-cfg.c (gimple_empty_block_p): Likewise.
3579 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
3581         PR target/86383
3582         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
3583         specified to configure.
3584         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
3586 2018-10-23  Richard Biener  <rguenther@suse.de>
3588         PR tree-optimization/87700
3589         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
3591 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
3593         PR target/87674
3594         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
3595         second argument from __mmask16 to __mmask8.
3596         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
3597         _mm_mask_packs_epi32): Likewise.
3598         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
3599         Likewise.
3600         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
3602 2018-10-23  Richard Biener  <rguenther@suse.de>
3604         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
3606 2018-10-23  Richard Biener  <rguenther@suse.de>
3608         PR tree-optimization/86144
3609         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
3610         over simd attribute.
3612 2018-10-23  Richard Biener  <rguenther@suse.de>
3614         PR tree-optimization/87693
3615         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
3616         the case we do not find the taken edge.
3618 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
3619             Jinsong Ji  <jji@us.ibm.com>
3621         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
3622         (_mm_store_pd): Use unaligned vector type for pointer cast.
3623         (_mm_maskmoveu_si128): Likewise.
3624         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
3625         (_mm_store_ps): Use unaligned vector type for pointer cast.
3627 2018-10-22  Paul Koning  <ni1d@arrl.net>
3629         * symtab.c (symtab_node::increase_alignment): Correct max
3630         alignment check.
3632 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
3634         PR tree-optimization/87633
3635         * match.pd: Do not generate unordered integer comparisons.
3637 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3639         PR rtl-optimization/87600
3640         * combine.c: Add include of expr.h.
3641         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
3642         register to a pseudo.
3643         (make_more_copies): New function, add a copy to a new pseudo after
3644         the moves from hard registers into pseudos.
3645         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
3646         later.  Call make_more_copies.
3648 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
3650         * lra-constraints.c (process_alt_operands): New local array,
3651         matching_early_clobber.  Check matching_early_clobber before
3652         decrementing reject, and set matching_early_clobber after.
3654 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3656         PR target/87598
3657         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
3658         call output_addr_const and hope for the best.
3660 2018-10-22  Richard Biener  <rguenther@suse.de>
3662         * gimple-ssa-evrp-analyze.c
3663         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
3664         smarter about what ranges to use.
3665         * tree-vrp.c (add_assert_info): Dump here.
3666         (register_edge_assert_for_2): Instead of here at multiple but
3667         not all places.
3669         * gcc.dg/tree-ssa/evrp12.c: New testcase.
3670         * gcc.dg/predict-6.c: Adjust.
3671         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
3672         * gcc.dg/tree-ssa/vrp02.c: Likewise.
3673         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
3675 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
3676         Richard Biener  <rguenther@suse.de>
3678         * bitmap.h: Update data structure documentation, including a
3679         description of bitmap views as either linked-lists or splay trees.
3680         (struct bitmap_element_def): Update comments for splay tree bitmaps.
3681         (struct bitmap_head_def): Likewise.
3682         (bitmap_list_view, bitmap_tree_view): New prototypes.
3683         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
3684         tree_form fields.
3685         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
3686         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
3687         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
3688         released bitmap element here.
3689         (bitmap_element_free): Remove.
3690         (bitmap_elt_clear_from): Work on splay tree bitmaps.
3691         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
3692         this function similar ones such that linked-list bitmap implementation
3693         functions are grouped.
3694         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
3695         and moved for grouping.
3696         (bitmap_list_insert_element_after): Renamed from
3697         bitmap_elt_insert_after, and moved for grouping.
3698         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
3699         (bitmap_tree_link_left, bitmap_tree_link_right,
3700         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
3701         bitmap_tree_link_element, bitmap_tree_unlink_element,
3702         bitmap_tree_find_element): New functions for splay-tree bitmap
3703         implementation.
3704         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
3705         Renamed and moved, see above entries.
3706         (bitmap_tree_listify_from): New function to convert part of a splay
3707         tree bitmap to a linked-list bitmap.
3708         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
3709         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
3710         (bitmap_find_bit): Remove.
3711         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
3712         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
3713         Handle splay tree bitmaps.
3714         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
3715         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
3716         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
3717         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
3718         bitmap_intersect_compl_p, bitmap_ior_and_compl,
3719         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
3720         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
3721         corresponding changes to use linked-list specific bitmap_element
3722         manipulation functions as applicable for efficiency.
3723         (bitmap_tree_to_vec): New function.
3724         (debug_bitmap_elt_file): New function split out from ...
3725         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
3726         (bitmap_print): Likewise.
3728         PR tree-optimization/63155
3729         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
3730         SSA edge worklists.
3731         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
3732         in tree-view.
3734 2018-10-22  Martin Liska  <mliska@suse.cz>
3736         PR tree-optimization/87686
3737         Revert
3738         2018-08-29  Martin Liska  <mliska@suse.cz>
3740         * tree-switch-conversion.c (switch_conversion::expand):
3741         Strenghten assumption about gswitch statements.
3743 2018-10-22  Martin Liska  <mliska@suse.cz>
3745         * ipa-icf.c (sem_item::compare_attributes): Remove.
3746         (sem_item::compare_referenced_symbol_properties): Use
3747         attribute_list_equal instead.
3748         (sem_function::equals_wpa): Likewise.
3749         * ipa-icf.h: Remove compare_attributes.
3751 2018-10-22  Richard Biener  <rguenther@suse.de>
3753         PR middle-end/87682
3754         * mem-stats.h (mem_usage::operator==): Fix pasto.
3756 2018-10-22  Richard Biener  <rguenther@suse.de>
3758         PR tree-optimization/87640
3759         * tree-vrp.c (set_value_range_with_overflow): Decompose
3760         incomplete result.
3761         (extract_range_from_binary_expr_1): Adjust.
3763 2018-10-22  Martin Jambor  <mjambor@suse.cz>
3765         * tree-eh.h (stmt_could_throw_p): Add function parameter.
3766         (stmt_can_throw_external): Likewise.
3767         (stmt_can_throw_internal): Likewise.
3768         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
3769         (lower_eh_constructs_2): Likewise.
3770         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
3771         (stmt_can_throw_external): Likewise.
3772         (stmt_can_throw_internal): Likewise.
3773         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
3774         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
3775         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
3776         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
3777         (pass_lower_eh_dispatch::execute): Pass cfun to
3778         stmt_can_throw_external.
3779         (cleanup_empty_eh): Likewise.
3780         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
3781         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
3782         stmt_can_throw_external instead of pushing it to cfun.
3783         (symbol_table::create_edge): Likewise.
3784         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
3785         stmt_can_throw_internal.
3786         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
3787         to stmt_could_throw_p.
3788         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
3789         stmt_can_throw_internal.
3790         (pass_store_merging::execute): Likewise.
3791         * gimple-ssa-strength-reduction.c
3792         (find_candidates_dom_walker::before_dom_children): Pass cfun to
3793         stmt_could_throw_p.
3794         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
3795         stmt_can_throw_internal.
3796         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
3797         to stmt_can_throw_external.
3798         (check_stmt): Pass cfun to stmt_could_throw_p.
3799         (check_stmt): Pass cfun to stmt_can_throw_external.
3800         (pass_nothrow::execute): Likewise.
3801         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
3802         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
3803         stmt_can_throw_internal.
3804         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
3805         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
3806         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
3807         * tree-complex.c (expand_complex_libcall): Pass cfun to
3808         stmt_could_throw_p and to stmt_can_throw_internal.
3809         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
3810         * tree-inline.c (copy_edges_for_bb): Likewise.
3811         (maybe_move_debug_stmts_to_successors): Likewise.
3812         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
3813         stmt_could_throw_p.
3814         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
3815         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
3816         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
3817         stmt_can_throw_internal.
3818         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
3819         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
3820         stmt_could_throw_p.
3821         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
3822         stmt_can_throw_internal.
3823         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
3824         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
3825         stmt_could_throw_p.
3826         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
3827         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
3828         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3829         (convert_mult_to_fma_1): Likewise.
3830         (convert_to_divmod): Likewise.
3831         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
3832         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
3833         * tree-ssa-propagate.c
3834         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
3835         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
3836         (maybe_optimize_range_tests): Likewise.
3837         (linearize_expr_tree): Likewise.
3838         (reassociate_bb): Likewise.
3839         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3840         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
3841         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
3842         (handle_char_store): Likewise.
3843         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
3844         stmt_can_throw_internal.
3845         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
3846         stmt_could_throw_p.
3847         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3848         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
3849         (vectorizable_simd_clone_call): Likewise.
3850         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
3851         (gimple_stringop_fixed_value): Likewise.
3853 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
3855         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
3856         literal pool references.
3857         (s390_check_qrst_address): Adapt to the new behavior of
3858         s390_loadrelative_operand_p ().
3860 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3862         PR target/72782
3863         * config/i386/sse.md (*andnot<mode>3_bcst): New.
3865 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3867         PR target/72782
3868         * config/i386/sse.md (*<code><mode>3_bcst): New.
3870 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3872         PR target/72782
3873         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
3874         V4DI, V16SI and V8DI.
3875         (*sub<mode>3<mask_name>_bcst): New.
3876         (*add<mode>3<mask_name>_bcst): Likewise.
3878 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
3879             Jinsong Ji  <jji@us.ibm.com>
3881         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
3882         __m64 with __vector unsigned long long for compatibility.
3883         (_mm_movemask_epi8): Likewise.
3884         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
3885         (_mm_cvttps_pi32): Likewise.
3886         (_mm_cvtpi32_ps): Likewise.
3887         (_mm_cvtps_pi16): Likewise.
3888         (_mm_loadh_pi): Likewise.
3889         (_mm_storeh_pi): Likewise.
3890         (_mm_movehl_ps): Likewise.
3891         (_mm_movelh_ps): Likewise.
3892         (_mm_loadl_pi): Likewise.
3893         (_mm_storel_pi): Likewise.
3894         (_mm_movemask_ps): Likewise.
3895         (_mm_shuffle_pi16): Likewise.
3897 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3899         PR target/72782
3900         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
3901         __builtin_ia32_vfnmsubpd512_mask.
3902         (_mm512_mask_fnmsub_round_pd): Likewise.
3903         (_mm512_fnmsub_pd): Likewise.
3904         (_mm512_mask_fnmsub_pd): Likewise.
3905         (_mm512_maskz_fnmsub_round_pd): Use
3906         __builtin_ia32_vfnmsubpd512_maskz.
3907         (_mm512_maskz_fnmsub_pd): Likewise.
3908         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
3909         (_mm512_mask_fnmsub_round_ps): Likewise.
3910         (_mm512_fnmsub_ps): Likewise.
3911         (_mm512_mask_fnmsub_ps): Likewise.
3912         (_mm512_maskz_fnmsub_round_ps): Use
3913         __builtin_ia32_vfnmsubps512_maskz.
3914         (_mm512_maskz_fnmsub_ps): Likewise.
3915         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
3916         __builtin_ia32_vfnmsubpd256_mask.
3917         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
3918         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3919         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
3920         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3921         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3922         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
3923         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
3924         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
3925         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
3926         __builtin_ia32_vfnmsubpd.
3927         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
3928         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
3929         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
3930         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
3931         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
3932         * config/i386/i386-builtin.def: Add
3933         __builtin_ia32_vfnmsubpd256_mask,
3934         __builtin_ia32_vfnmsubpd256_maskz,
3935         __builtin_ia32_vfnmsubpd128_mask,
3936         __builtin_ia32_vfnmsubpd128_maskz,
3937         __builtin_ia32_vfnmsubps256_mask,
3938         __builtin_ia32_vfnmsubps256_maskz,
3939         __builtin_ia32_vfnmsubps128_mask,
3940         __builtin_ia32_vfnmsubps128_maskz,
3941         __builtin_ia32_vfnmsubpd512_mask,
3942         __builtin_ia32_vfnmsubpd512_maskz,
3943         __builtin_ia32_vfnmsubps512_mask,
3944         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
3945         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
3946         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
3947         __builtin_ia32_vfnmsubpd256.
3948         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
3949         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
3950         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
3951         Likewise.
3952         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
3953         Likewise.
3954         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
3955         Likewise.
3956         (fmai_vmfnmsub_<mode><round_name>): Likewise.
3958 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3960         PR target/72782
3961         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
3962         __builtin_ia32_vfnmaddpd512_mask.
3963         (_mm512_mask_fnmadd_round_pd): Likewise.
3964         (_mm512_fnmadd_pd): Likewise.
3965         (_mm512_mask_fnmadd_pd): Likewise.
3966         (_mm512_maskz_fnmadd_round_pd): Use
3967         __builtin_ia32_vfnmaddpd512_maskz.
3968         (_mm512_maskz_fnmadd_pd): Likewise.
3969         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
3970         (_mm512_mask_fnmadd_round_ps): Likewise.
3971         (_mm512_fnmadd_ps): Likewise.
3972         (_mm512_mask_fnmadd_ps): Likewise.
3973         (_mm512_maskz_fnmadd_round_ps): Use
3974         __builtin_ia32_vfnmaddps512_maskz.
3975         (_mm512_maskz_fnmadd_ps): Likewise.
3976         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
3977         __builtin_ia32_vfnmaddpd256_mask.
3978         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
3979         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
3980         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
3981         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3982         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3983         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
3984         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
3985         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
3986         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
3987         __builtin_ia32_vfnmaddpd.
3988         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
3989         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
3990         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
3991         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
3992         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
3993         * config/i386/i386-builtin.def: Add
3994         __builtin_ia32_vfnmaddpd256_mask,
3995         __builtin_ia32_vfnmaddpd256_maskz,
3996         __builtin_ia32_vfnmaddpd128_mask,
3997         __builtin_ia32_vfnmaddpd128_maskz,
3998         __builtin_ia32_vfnmaddps256_mask,
3999         __builtin_ia32_vfnmaddps256_maskz,
4000         __builtin_ia32_vfnmaddps128_mask,
4001         __builtin_ia32_vfnmaddps128_maskz,
4002         __builtin_ia32_vfnmaddpd512_mask,
4003         __builtin_ia32_vfnmaddpd512_maskz,
4004         __builtin_ia32_vfnmaddps512_mask,
4005         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
4006         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
4007         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
4008         __builtin_ia32_vfnmaddpd256.
4009         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
4010         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
4011         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
4012         Likewise.
4013         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
4014         Likewise.
4015         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
4016         Likewise.
4017         (fmai_vmfnmadd_<mode><round_name>): Likewise.
4019 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4021         PR target/72782
4022         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
4023         __builtin_ia32_vfmsubpd512_mask.
4024         (_mm512_mask_fmsub_round_pd): Likewise.
4025         (_mm512_fmsub_pd): Likewise.
4026         (_mm512_mask_fmsub_pd): Likewise.
4027         (_mm512_maskz_fmsub_round_pd): Use
4028         __builtin_ia32_vfmsubpd512_maskz.
4029         (_mm512_maskz_fmsub_pd): Likewise.
4030         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
4031         (_mm512_mask_fmsub_round_ps): Likewise.
4032         (_mm512_fmsub_ps): Likewise.
4033         (_mm512_mask_fmsub_ps): Likewise.
4034         (_mm512_maskz_fmsub_round_ps): Use
4035         __builtin_ia32_vfmsubps512_maskz.
4036         (_mm512_maskz_fmsub_ps): Likewise.
4037         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
4038         __builtin_ia32_vfmsubpd256_mask.
4039         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
4040         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
4041         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
4042         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
4043         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
4044         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
4045         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
4046         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
4047         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
4048         __builtin_ia32_vfmsubpd.
4049         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
4050         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
4051         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
4052         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
4053         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
4054         * config/i386/i386-builtin.def: Add
4055         __builtin_ia32_vfmsubpd256_mask,
4056         __builtin_ia32_vfmsubpd256_maskz,
4057         __builtin_ia32_vfmsubpd128_mask,
4058         __builtin_ia32_vfmsubpd128_maskz,
4059         __builtin_ia32_vfmsubps256_mask,
4060         __builtin_ia32_vfmsubps256_maskz,
4061         __builtin_ia32_vfmsubps128_mask,
4062         __builtin_ia32_vfmsubps128_maskz,
4063         __builtin_ia32_vfmsubpd512_mask,
4064         __builtin_ia32_vfmsubpd512_maskz,
4065         __builtin_ia32_vfmsubps512_mask,
4066         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
4067         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
4068         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
4069         __builtin_ia32_vfmsubpd256.
4070         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
4071         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
4072         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
4073         Likewise.
4074         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
4075         Likewise.
4076         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
4077         Likewise.
4078         (fmai_vmfmsub_<mode><round_name>): Likewise.
4080 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4082         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
4083         Remove plus.  Renamed to ...
4084         (*sub<mode>3<mask_name>_bcst): This.
4085         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
4086         (*add<mode>3<mask_name>_bcst): This.
4088 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4090         PR target/72782
4091         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
4093 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4095         PR target/87662
4096         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
4097         (_mm_or_epi32): Likewise.
4098         (_mm256_xor_epi32): Likewise.
4099         (_mm_xor_epi32): Likewise.
4100         (_mm256_or_epi64): Likewise.
4101         (_mm_or_epi64): Likewise.
4102         (_mm256_xor_epi64): Likewise.
4103         (_mm_xor_epi64): Likewise.
4105 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
4107         PR target/72782
4108         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
4110 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
4112         PR middle-end/87647
4113         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
4115 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
4117         * doc/ux.texi: Move @section directly after @node.
4119 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
4121         PR middle-end/85488
4122         PR middle-end/87649
4123         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
4124         depend closely nested inside of loop with ordered clause with
4125         a parameter.
4127 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
4129         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
4130         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
4131         * doc/ux.texi: New file.
4133 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
4135         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
4136         be the first CR field allocated.
4138 2018-10-19  Richard Biener  <rguenther@suse.de>
4140         PR target/87657
4141         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
4142         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
4144 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
4146         PR target/72782
4147         * config/i386/sse.md
4148         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
4149         (*add<mode>3<mask_name>_bcst_2): Likewise.
4151 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
4153         * config/i386/sse.md
4154         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
4155         Replace nonimmediate_operand with register_operand.
4156         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
4157         Likewise.
4158         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
4159         Likewise.
4161 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
4163         PR rtl-optimization/87596
4164         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
4165         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
4166         for instructions in FROM..TO range.
4168 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
4170         * cfgexpand.c (expand_one_var): Use specific wording in error message
4171         for non-local frame variables.
4172         * stor-layout.c (layout_decl): Do not issue a warning for them.
4174 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
4176         * haifa-sched.c (priority): Add force_recompute parameter.
4177         (apply_replacement): Call priority () with force_recompute = true.
4178         (restore_pattern): Likewise.
4180 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4182         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
4183         HOST_BITS_PER_WIDE_INT.
4184         (test_vector_ops_duplicate): Likewise.
4186 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4188         PR target/72782
4189         * config/i386/sse.md (VF_AVX512): New.
4190         (avx512bcst): Likewise.
4191         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
4192         Likewise.
4193         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
4194         Likewise.
4195         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
4196         Likewise.
4198 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
4200         * doc/invoke.texi (-dumpversion): Improve grammar.
4201         (-dumpfullversion): Make more consistent with -dumpversion.
4203 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
4205         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
4206         Set JUMP_LABEL to the jump insn.
4207         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
4208         Predict emitted jump and add label to jump insn.
4210 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
4212         PR tree-optimization/87562
4213         * input.c (get_substring_ranges_for_loc): Use
4214         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
4215         getting the linemap for the endpoint.  Verify that it's either
4216         in the same linemap as the start point's spelling location, or
4217         at least in the same file.
4219 2018-10-18  Richard Biener  <rguenther@suse.de>
4221         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
4222         feed width-specific load/store costs through ix86_vec_cost.
4223         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
4224         (k8_cost): Likewise.
4225         (bdver_cost): Likewise.
4226         (znver1_cost): Likewise.
4227         (btver1_cost): Likewise.
4228         (btver2_cost): Likewise.
4230 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4232         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
4233         to simplify subreg of vec_merge.
4235 2018-10-18  Richard Biener  <rguenther@suse.de>
4237         * config/i386/i386.c: Fix costing of vector FMA.
4239 2018-10-18  Richard Biener  <rguenther@suse.de>
4241         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
4242         and argument.
4243         (ix86_builtin_vectorization_cost): For vec_construct properly
4244         cost insertion into SSE regs.
4245         (...): Adjust calls to ix86_vec_cost.
4247 2018-10-18  Richard Biener  <rguenther@suse.de>
4249         PR middle-end/87087
4250         Revert
4251         2018-02-07  Richard Biener  <rguenther@suse.de>
4253         PR tree-optimization/84204
4254         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4255         this place.
4257 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4259         PR target/87537
4260         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
4261         of vec_duplicate.
4262         (test_vector_ops_duplicate): Add test for a scalar subreg of a
4263         VEC_MERGE of a VEC_DUPLICATE.
4265 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
4267         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
4268         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
4269         * doc/standards.texi (C Language): Document C2X.
4270         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
4271         (rl78_option_override): Handle "GNU C2X" language name.
4273 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
4275         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
4276         Document C17 as published in 2018.
4278 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
4280         PR middle-end/87623
4281         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
4282         bail out if both sides do not have the same storage order.
4284 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4286         * bitmap.c (bitmap_head::dump): New.
4287         * bitmap.h (bitmap_head): Add dump().
4288         * gimple-ssa-evrp-analyze.c
4289         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
4290         (evrp_range_analyzer::set_ssa_range_info): Same.
4291         (evrp_range_analyzer::record_ranges_from_phis): Same.
4292         (evrp_range_analyzer::record_ranges_from_stmt): Same.
4293         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
4294         * gimple-ssa-sprintf.c (get_int_range): Same.
4295         (format_integer): Same.
4296         (sprintf_dom_walker::handle_gimple_call): Same.
4297         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
4298         (ipcp_vr_lattice::top_p): Same.
4299         (ipcp_vr_lattice::bottom_p): Same.
4300         (ipcp_vr_lattice::set_to_bottom): Same.
4301         (ipa_vr_operation_and_type_effects): Same.
4302         (propagate_vr_across_jump_function): Same.
4303         (ipcp_store_vr_results): Same.
4304         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
4305         (ipa_print_node_jump_functions_for_edge): Same.
4306         (ipa_get_value_range): Same.
4307         (ipa_compute_jump_functions_for_edge): Same.
4308         (ipa_write_jump_function): Same.
4309         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
4310         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4311         Same.
4312         * vr-values.c (set_value_range_to_nonnegative): Same.
4313         (set_value_range_to_truthvalue): Same.
4314         (vr_values::get_value_range): Same.
4315         (vr_values::set_defs_to_varying): Same.
4316         (vr_values::update_value_range): Same.
4317         (symbolic_range_based_on_p): Same.
4318         (vr_values::op_with_boolean_value_range_p): Same.
4319         (vr_values::extract_range_for_var_from_comparison_expr): Same.
4320         (vr_values::extract_range_from_ssa_name): Same.
4321         (vr_values::extract_range_from_binary_expr): Same.
4322         (vr_values::extract_range_from_unary_expr): Same.
4323         (vr_values::extract_range_from_cond_expr): Same.
4324         (vr_values::extract_range_from_comparison): Same.
4325         (vr_values::check_for_binary_op_overflow): Same.
4326         (vr_values::extract_range_basic): Same.
4327         (vr_values::extract_range_from_assignment): Same.
4328         (compare_ranges): Same.
4329         (compare_range_with_value): Same.
4330         (vr_values::adjust_range_with_scev): Same.
4331         (vrp_valueize): Same.
4332         (vrp_valueize_1): Same.
4333         (vr_values::get_vr_for_comparison): Same.
4334         (vr_values::compare_name_with_value): Same.
4335         (vr_values::compare_names): Same.
4336         (vr_values::vrp_evaluate_conditional): Same.
4337         (find_case_label_ranges): Same.
4338         (vr_values::vrp_visit_switch_stmt): Same.
4339         (vr_values::extract_range_from_phi_node): Same.
4340         (vr_values::simplify_div_or_mod_using_ranges): Same.
4341         (vr_values::simplify_bit_ops_using_ranges): Same.
4342         (test_for_singularity): Same.
4343         (range_fits_type_p): Same.
4344         (vr_values::simplify_cond_using_ranges_1): Same.
4345         (vr_values::simplify_switch_using_ranges): Same.
4346         (vr_values::simplify_float_conversion_using_ranges): Same.
4347         (vr_values::two_valued_val_range_p): Same.
4348         (vr_values::add_equivalence): Move to value_range::equiv_add.
4349         * vr-values.h (vr_values::add_equivalence): Remove.
4350         (VR_INITIALIZER): Remove.
4351         * tree-vrp.c (value_range::set): New.
4352         (value_range::equiv_add): New.
4353         (value_range::value_range): New.
4354         (value_range::deep_copy): New.
4355         (value_range::check): New.
4356         (value_range::equal_p): New.
4357         (value_range::ignore_equivs_equal_p): New.
4358         (value_range::operator==): New.
4359         (value_range::operator!=): New.
4360         (value_range::symbolic_p): New.
4361         (value_range::numeric_p): New.
4362         (value_range::set_undefined): New.
4363         (value_range::set_varying): New.
4364         (value_range::may_contain_p): New.
4365         (value_range::equiv_clear): New.
4366         (value_range::singleton_p): New.
4367         (value_range::intersect): New.
4368         (value_range::dump): New.
4369         (value_range::set_and_canonicalize): New.
4370         (set_value_range): Adjust for value_range API.
4371         (set_value_range_to_undefined): Same.
4372         (set_value_range_to_varying): Same.
4373         (set_and_canonicalize_value_range): Same.
4374         (set_value_range_to_nonnull): Same.
4375         (set_value_range_to_null): Same.
4376         (range_is_null): Same.
4377         (range_is_nonnull): Same.
4378         (range_int_cst_p): Same.
4379         (range_int_cst_singleton_p): Same.
4380         (symbolic_range_p): Same.
4381         (range_includes_zero_p): Same.
4382         (value_range_constant_singleton): Same.
4383         (vrp_set_zero_nonzero_bits): Same.
4384         (ranges_from_anti_range): Same.
4385         (extract_range_into_wide_ints): Same.
4386         (extract_range_from_multiplicative_op): Same.
4387         (set_value_range_with_overflow): Same.
4388         (extract_range_from_binary_expr_1): Same.
4389         (extract_range_from_unary_expr): Same.
4390         (dump_value_range): Same.
4391         (debug_value_range): Same.
4392         (vrp_prop::check_array_ref): Same.
4393         (vrp_prop::check_mem_ref): Same.
4394         (vrp_prop::vrp_initialize): Same.
4395         (vrp_prop::visit_stmt): Same.
4396         (intersect_ranges): Same.
4397         (vrp_prop::visit_phi): Same.
4398         (vrp_prop::vrp_finalize): Same.
4399         (determine_value_range_1): Same.
4400         (determine_value_range): Same.
4401         (vrp_intersect_ranges_1): Rename to...
4402         (vrp_intersect_1): this.
4403         (vrp_intersect_ranges): Rename to...
4404         (value_range::intersect_helper): ...this.
4405         (vrp_meet_1): Rename to...
4406         (value_range::union_helper): ...this.
4407         (vrp_meet): Rename to...
4408         (value_range::union_): ...this.
4409         (copy_value_range): Remove.
4410         * tree-vrp.h (struct value_range): Rewrite into a proper class.
4411         (value_range::vrtype): New.
4412         (value_range::type): New.
4413         (value_range::equiv): New.
4414         (value_range::min): New.
4415         (value_range::max): New.
4416         (value_range::varying_p): New.
4417         (value_range::undefined_p): New.
4418         (value_range::null_p): New.
4419         (value_range::equiv_add): New.
4420         (copy_value_range): Remove.
4422 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
4424         * Makefile.in (SELFTEST_TARGETS): New.
4425         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
4426         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
4427         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
4428         c/Make-lang.in.
4429         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
4430         (selftest-c++-gdb, selftest-c++-valgrind): Move to
4431         cp/Make-lang.in.
4432         * configure: Regenerate.
4433         * configure.ac (selftest_languages): New.
4435 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4437         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
4438         overflow wraps argument.
4439         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
4440         wide_int_range_multiplicative_op.
4441         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
4442         overflow wraps argument.
4443         (wide_int_range_multiplicative_op): Same.
4444         (wide_int_range_lshift): Same.
4445         (wide_int_range_div): Same.
4446         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
4447         (wide_int_range_lshift): Same.
4448         (wide_int_range_div): Same.
4450 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4452         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
4453         use sign as argument.
4454         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
4455         wide_int_range_shift_undefined_p.
4457 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4459         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
4460         Rename to...
4461         (@despeculate_copy<ALLI_TI:mode>): ... This.
4462         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
4463         switch statement.
4465 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4467         * config.gcc: Obsolete *-*-solaris2.10*.
4468         * doc/install.texi (Specific, *-*-solaris2*): Document it.
4470 2018-10-12  Jeff Law  <law@redhat.com>
4472         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
4473         reg + sym +- const_int addressing modes.
4475 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4477         * common.opt (fdiagnostics-minimum-margin-width=): New option.
4478         * diagnostic-show-locus.c (layout::layout): Apply the minimum
4479         margin width.
4480         (layout::start_annotation_line): Only print up to 3 of the
4481         margin character, to avoid touching the left-hand side.
4482         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
4483         minimum margin width, as set by test_diagnostic_context's ctor.
4484         (selftest::test_fixit_insert_containing_newline): Likewise.
4485         (selftest::test_fixit_insert_containing_newline_2): Likewise.
4486         (selftest::test_line_numbers_multiline_range): Clear
4487         dc.min_margin_width.
4488         * diagnostic.c (diagnostic_initialize): Initialize
4489         min_margin_width.
4490         * diagnostic.h (struct diagnostic_context): Add field
4491         "min_margin_width".
4492         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
4493         * opts.c (common_handle_option): Handle
4494         OPT_fdiagnostics_minimum_margin_width_.
4495         * selftest-diagnostic.c
4496         (selftest::test_diagnostic_context::test_diagnostic_context):
4497         Initialize min_margin_width to 6.
4498         * toplev.c (general_init): Initialize global_dc->min_margin_width.
4500 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4502         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4503         Fix usage of "error_at_rich_loc" in the comment.
4505 2018-10-15  Renlin Li  <renlin.li@arm.com>
4507         PR target/87563
4508         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
4509         if-conversioned loop when it contains ifn with types not
4510         supported by backend.
4511         * internal-fn.c (expand_direct_optab_fn): Add an assert.
4512         (direct_internal_fn_supported_p): New helper function.
4513         * internal-fn.h (direct_internal_fn_supported_p): Declare.
4515 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
4517         PR target/87572
4518         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4519         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4520         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4522 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
4524         PR tree-optimization/87022
4525         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
4526         bits in dist vector rather than the first one.
4528 2018-10-15  Richard Biener  <rguenther@suse.de>
4530         PR middle-end/87610
4531         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
4532         (visit_loadstore): When a used restrict tag escaped verify that
4533         the points-to solution of "other" pointers do not include
4534         escaped.
4535         (compute_dependence_clique): If a used restrict tag escaped
4536         communicated that down to visit_loadstore.
4538 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
4540         * config/s390/s390.c (s390_expand_vec_init): Force vector element
4541         into reg if it isn't a general operand.
4543 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4545         PR target/87599
4546         * config/i386/sse.md (*vec_dupv2di): Add register source to
4547         movddup.
4549 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4551         PR target/87572
4552         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4553         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
4554         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4555         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4557 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
4559         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
4560         (notice_args_size): Set it in the current trace if no insn that can
4561         throw internally has been seen yet.
4562         (connect_traces): When connecting args_size between traces, allow the
4563         incoming values not to match if there is an insn setting it before the
4564         first insn that can throw internally; in that case, force the creation
4565         of a CFI note on this latter insn.
4567 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
4569         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
4570         the base class, because using the injected-class-name was not clearly
4571         specified until DR 176.
4573 2018-10-12  Paul Koning  <ni1d@arrl.net>
4575         * config/pdp11/pdp11.md (doloop_end): New expander.
4576         (doloop_end_insn): renamed from "doloop_end".
4577         (addqi3): New pattern.
4578         (subqi3): New pattern.
4579         * config/pdp11/predicates.md (incdec_operand): New predicate.
4581 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
4583         PR middle-end/81376
4584         * real.c (format_helper::can_represent_integral_type_p): New function
4585         * real.h (format_helper::can_represent_integral_type_p): Ditto.
4586         * match.pd: New pattern.
4588 2018-10-12  Alexandre Oliva <oliva@adacore.com>
4590         * configure.ac: Introduce --enable-large-address-aware
4591         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
4592         * doc/install.texi: Document it.
4593         * configure, config.in: Rebuilt.
4594         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
4595         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
4596         (LINK_SPEC): Insert it.
4597         * config/i386/mingw-w64.h: Likewise.
4599         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
4601 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
4603         PR rtl-optimization/87600
4604         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
4606 2018-10-12  Paul Koning  <ni1d@arrl.net>
4608         * doc/md.texi (doloop_end): Document that the pattern code may
4609         need to check operand mode.
4611 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
4613         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
4614         to zero-extend between int and floating-point registers.
4615         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
4616         ldp into floating-point registers.  Add type and arch attributes.
4617         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
4618         Use f_loads for type attribute.
4620 2018-10-11  Martin Sebor  <msebor@redhat.com>
4622         * doc/extend.texi (attribute packed): Correct typos.
4624 2018-10-11  Martin Sebor  <msebor@redhat.com>
4626         * doc/extend.texi (attribute flatten): Mention interaction with
4627         noinline.
4629 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
4631         PR target/87156
4632         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4633         Set new_decl virtual flag to zero.
4635 2018-10-11  Martin Sebor  <msebor@redhat.com>
4637         PR middle-end/87593
4638         * doc/extend.texi (attribute format_arg): Discuss using multiple
4639         attributes on a single function.
4641 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
4643         PR tree-optimization/86829
4644         * match.pd (sin (atan (x))): New simplification rules.
4645         (cos (atan (x))): Likewise.
4646         * real.c (build_sinatan_real): New function.
4647         * real.h (build_sinatan_real): Prototype.
4649 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
4651         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
4652         function.
4653         (fold_mergeeo_helper): New helper function.
4654         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
4655         intrinsics.  Correct some whitespace indentation issues.
4657 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
4659         PR target/87511
4660         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
4661         Use HOST_WIDE_INT_1U for shift.
4663 2018-10-11  Doug Rupp  <rupp@adacore.com>
4664             Olivier Hainque  <hainque@adacore.com>
4666         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
4667         Pass --relax to the linker for RTPs.
4668         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
4670 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
4671             Jan Hubicka  <jh@suse.cz>
4672             Martin Jambor  <mjambor@suse.cz>
4674         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
4675         the same elements are repeated rather than printing all of them.
4676         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
4677         "repeated" elements.
4678         * read-rtl-function.c (test_loading_repeat): New function.
4679         (read_rtl_function_c_tests): Call test_loading_repeat.
4680         * rtl-tests.c (test_dumping_repeat): New function.
4681         (rtl_tests_c_tests): Call test_dumping_repeat.
4683 2018-10-11  Richard Biener  <rguenther@suse.de>
4685         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
4686         bdver?_cost): Unify to ...
4687         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
4688         * config/i386/i386.c (processor_cost_table): Adjust.
4690 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
4692         PR middle-end/87574
4693         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
4694         the thunk when expanding to GIMPLE.
4696 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4698         * varasm.c (mergeable_string_section): Don't try to move zero-length
4699         strings to the merge section.
4701 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
4703         PR target/87573
4704         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
4706 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
4708         PR target/87550
4709         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
4710         to special_args set.
4712 2018-10-10  Richard Biener  <rguenther@suse.de>
4714         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
4715         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
4716         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
4717         and recursing and pattern terminating the recursion on SSE
4718         vector width using ix86_expand_reduc.
4719         (reduc_sminmax_scal_<mode>): Split into part reducing to half
4720         width and recursing and SSE2 vector variant doing the final
4721         reduction with ix86_expand_reduc.
4722         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
4723         with terminating the recursion at AVX level, splitting that
4724         to SSE there.
4726 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
4728         * genmatch.c (error_cb): Rename to...
4729         (diagnostic_cb): ...this, converting int params to enums.
4730         (fatal_at): Update for renaming.
4731         (warning_at): Likewise.
4732         (main): Likewise.
4733         * input.c (selftest::ebcdic_execution_charset::apply):
4734         Update for renaming of...
4735         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
4736         to...
4737         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
4738         converting level and reason to enums.
4739         (class selftest::lexer_error_sink): Rename to...
4740         (class selftest::lexer_test_options): ...this, renaming field
4741         "m_errors" to "m_diagnostics".
4742         (selftest::lexer_test_options::apply): Update for renaming of...
4743         (selftest::lexer_test_options::on_error): ...this, renaming to...
4744         (selftest::lexer_test_options::on_diagnostic): ...this
4745         converting level and reason to enums.
4746         (selftest::test_lexer_string_locations_raw_string_unterminated):
4747         Update for renamings.
4748         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
4749         "reason".
4751 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
4753         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
4754         * config/rs6000/pmmintrin.h: New file.
4756 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
4758         PR tree-optimization/86659
4759         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
4761 2018-10-09  Richard Biener  <rguenther@suse.de>
4763         PR tree-optimization/63155
4764         * tree-ssa-structalias.c: Include tree-ssa.h.
4765         (get_constraint_for_ssa_var): For undefs return nothing_id.
4766         (find_func_aliases): Cleanup PHI handling.
4768 2018-10-09  Richard Biener  <rguenther@suse.de>
4770         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
4771         replacements.
4773 2018-10-09  Martin Liska  <mliska@suse.cz>
4775         * asan.c (asan_emit_stack_protection): If a stack variable
4776         is located in a same file as current function, then emit
4777         line info into variable definition string.
4779 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4781         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
4782         information.
4784 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4786         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
4787         on the thunk.
4789 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4791         PR tree-optimization/86659
4792         * gimple-match.h (struct gimple_match_op): Add reverse field.
4793         (gimple_match_op::set_op): New overloaded method.
4794         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
4795         the REF_REVERSE_STORAGE_ORDER flag on the value.
4796         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
4797         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
4798         
4799 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
4801         PR middle-end/63155
4802         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
4803         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
4805 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
4807         PR target/87517
4808         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
4809         Defined with __builtin_ia32_vfmaddsubpd512_mask.
4811 2018-10-08  Richard Biener  <rguenther@suse.de>
4813         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
4814         cost the same as AVX128 ones.
4816 2018-10-08  Paul Koning  <ni1d@arrl.net>
4818         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
4819         (expand_block_move): New function.
4820         * config/pdp11/pdp11.c (output_block_move): Remove.
4821         (expand_block_move): New function.
4822         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
4823         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
4824         (*movmemhi1): Remove.
4826 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
4828         * config/s390/2827.md: Increase latencies for some FP instructions.
4830 2018-10-08  Richard Biener  <rguenther@suse.de>
4832         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4833         Open a dump scope.
4834         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
4835         * tree-vectorizer.h (dump_stmt_cost): Adjust.
4836         (add_stmt_cost): Dump return value of the hook.
4838 2018-10-08  Richard Biener  <rguenther@suse.de>
4840         PR tree-optimization/63155
4841         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
4842         (ssa_propagation_engine::ssa_propagate): Remove redundant
4843         bitmap bit clearing.
4845 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
4847         PR rtl-optimization/86939
4848         PR rtl-optimization/87479
4849         * ira.h (non_conflicting_reg_copy_p): New prototype.
4850         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
4851         (make_hard_regno_dead): Don't add conflicts for register
4852         ignore_reg_for_conflicts.
4853         (make_object_dead): Likewise.
4854         (non_conflicting_reg_copy_p): New function.
4855         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
4856         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
4857         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
4858         (make_hard_regno_dead): Don't add conflicts for register
4859         ignore_reg_for_conflicts.  Remove special conflict handling of
4860         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
4861         check_pic_pseudo_p and update callers.
4862         (mark_pseudo_dead): Don't add conflicts for register
4863         ignore_reg_for_conflicts.
4864         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
4866 2018-10-05  Andrew Waterman  <andrew@sifive.com>
4867             Jim Wilson  <jimw@sifive.com>
4869         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
4870         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
4871         new pattern using HONOR_SNANS that emits one extra instruction.
4873 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4875         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
4876         patterns): Merge SI and DI patterns to a GPR pattern.
4877         (unnamed define_insn and define_split for record form of that): Merge
4878         to a single define_insn_and_split pattern.
4880 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
4882         PR c++/56856
4883         * input.c (expand_location_to_spelling_point): Add param "aspect"
4884         and use rather than hardcoding LOCATION_ASPECT_CARET.
4885         (get_substring_ranges_for_loc): Handle the case of a single token
4886         within a macro expansion.
4887         * input.h (expand_location_to_spelling_point): Add "aspect" param,
4888         defaulting to LOCATION_ASPECT_CARET.
4890 2018-10-05  Paul Koning  <ni1d@arrl.net>
4892         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
4893         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
4894         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4895         (pdp11_guard_type): New function.
4897 2018-10-05  Paul Koning  <ni1d@arrl.net>
4899         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
4900         * config/pdp11/pdp11.opt (mfloat32): Remove.
4901         (mfloat64): Remove.
4902         * doc/invoke.texi (pdp11 -mfloat32): Remove:
4903         (pdp11 -mfloat64): Remove.
4905 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4907         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
4908         (*cmp<mode>_cc_i387): Ditto.
4909         (*cmpu<mode>_cc_i387): Ditto.
4910         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4911         * config/i386/i386.c (ix86_expand_fp_compare): Remove
4912         "scratch" argument.
4913         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
4914         Emit x86_sahf_1 pattern.
4915         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
4916         (ix86_expand_carry_flag_compare): Ditto.
4918 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4920         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
4921         to reg_or_0_operand.  Add "C" constraint.
4922         (*cmpxf_cc_i387): Ditto.
4923         (*cmp<mode>_i387): Change operand 2 predicate
4924         to nonimm_or_0_operand.  Add "C" constraint.
4925         (*cmp<mode>_cc_i387): Ditto.
4926         (*cmp<mode>_0_i387): Remove insn pattern.
4927         (*cmp<mode>_0_cc_i387): Ditto.
4929 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4931         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
4932         * config/i386/predicates.md (nonimm_or_0_operand): Rename
4933         from vector_move_operand.  Update all uses.
4935 2018-10-05  Martin Sebor  <msebor@redhat.com>
4937         PR tree-optimization/87490
4938         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
4939         consistently.
4941 2018-10-05  Richard Biener  <rguenther@suse.de>
4943         PR tree-optimization/63155
4944         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
4945         vertical space in dumpfiles.
4946         * tree-ssa-propagate.h
4947         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
4948         * tree-ssa-propagate.c (cfg_blocks_back): New global.
4949         (ssa_edge_worklist_back): Likewise.
4950         (curr_order): Likewise.
4951         (cfg_blocks_get): Remove abstraction.
4952         (cfg_blocks_add): Likewise.
4953         (cfg_blocks_empty_p): Likewise.
4954         (add_ssa_edge): Add to current or next worklist based on
4955         RPO index.
4956         (add_control_edge): Likewise.
4957         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
4958         into ...
4959         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
4960         iteration from CFG and SSA edge worklist so we process
4961         everything in RPO order, prioritizing forward progress
4962         over iteration.
4963         (ssa_prop_init): Allocate new worklists, do not dump
4964         immediate uses.
4965         (ssa_prop_fini): Free new worklists.
4967 2018-10-05  Richard Biener  <rguenther@suse.de>
4969         * tree-core.h (tree_block::abstract_flag): Remove.
4970         (tree_block::block_num): Make full 32bits.
4971         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
4972         * tree.h (BLOCK_ABSTRACT): Remove.
4973         * dwarf2out.c (gen_lexical_block_die): Remove dead code
4974         resulting from BLOCK_ABSTRACT being always false.
4975         (gen_inlined_subroutine_die): Likewise.
4976         (gen_block_die): Likewise.
4977         * tree.c (block_ultimate_origin): Likewise.
4978         * tree-pretty-print.c (dump_block_node): Remove code dealing
4979         with BLOCK_ABSTRACT.
4980         * tree-ssa-live.c (dump_scope_block): Likewise.
4981         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
4982         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
4984 2018-10-05   Richard Biener  <rguenther@suse.de>
4986         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
4987         is asked for initialize mode to the component mode of the
4988         vector type.
4990 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
4992         PR target/87522
4993         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
4994         assembler for -mavx.
4995         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
4997 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4999         PR target/87509
5000         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
5001         RS6000_BTM_DFP.
5002         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
5003         to be DImode.  When using mffscrn, force the operand to a register.
5005 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
5007         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
5008         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
5009         X87MODEF mode iterator.
5010         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
5011         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
5012         X87MODEF mode iterator.
5014 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
5016         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
5017         -Wno-prio-ctor-dtor.
5019 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5021         * Makefile.in (OBJS): Add opt-problem.o.
5022         * dump-context.h: Include "selftest.h.
5023         (selftest::temp_dump_context): New forward decl.
5024         (class dump_context): Make friend of class
5025         selftest::temp_dump_context.
5026         (dump_context::dump_loc_immediate): New decl.
5027         (class dump_pretty_printer): Move here from dumpfile.c.
5028         (class temp_dump_context): Move to namespace selftest.
5029         (temp_dump_context::temp_dump_context): Add param
5030         "forcibly_enable_dumping".
5031         (selftest::verify_dumped_text):
5032         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
5033         (selftest::verify_item):
5034         (ASSERT_IS_TEXT): Move here from dumpfile.c.
5035         (ASSERT_IS_TREE): Likewise.
5036         (ASSERT_IS_GIMPLE): Likewise.
5037         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
5038         to...
5039         (dump_context::dump_loc_immediate): ...this new function.
5040         (class dump_pretty_printer): Move to dump-context.h.
5041         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
5042         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
5043         (temp_dump_context::temp_dump_context): Move to "selftest"
5044         namespace.  Add param "forcibly_enable_dumping", and use it to
5045         conditionalize the use of m_pp;
5046         (selftest::verify_dumped_text): Make non-static.
5047         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
5048         (selftest::verify_item): Make non-static.
5049         (ASSERT_IS_TEXT): Move to dump-context.h.
5050         (ASSERT_IS_TREE): Likewise.
5051         (ASSERT_IS_GIMPLE): Likewise.
5052         (selftest::test_capture_of_dump_calls): Pass "true" for new
5053         param of temp_dump_context.
5054         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
5055         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
5056         TDF_COMPARE_DEBUG.
5057         * opt-problem.cc: New file.
5058         * opt-problem.h: New file.
5059         * optinfo-emit-json.cc
5060         (selftest::test_building_json_from_dump_calls): Pass "true" for
5061         new param of temp_dump_context.
5062         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
5063         (optinfo::emit_for_opt_problem): New function.
5064         (optinfo::emit): Clarity which emit_item is used.
5065         * optinfo.h (optinfo::get_dump_location): New accessor.
5066         (optinfo::emit_for_opt_problem): New decl.
5067         (optinfo::emit): Make const.
5068         * selftest-run-tests.c (selftest::run_tests): Call
5069         selftest::opt_problem_cc_tests.
5070         * selftest.h (selftest::opt_problem_cc_tests): New decl.
5071         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
5072         bool to opt_result, converting fprintf messages to
5073         opt_result::failure_at calls.  Add "stmt" param for use by the
5074         failure_at calls.
5075         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
5076         (runtime_alias_check_p): Convert return type from bool to
5077         opt_result, converting dump_printf calls to
5078         opt_result::failure_at, using the statement DDR_A for their
5079         location.
5080         (find_data_references_in_stmt): Convert return type from bool to
5081         opt_result, converting "return false" to opt_result::failure_at
5082         with a new message.
5083         * tree-data-ref.h: Include "opt-problem.h".
5084         (dr_analyze_innermost): Convert return type from bool to opt_result,
5085         and add a const gimple * param.
5086         (find_data_references_in_stmt): Convert return type from bool to
5087         opt_result.
5088         (runtime_alias_check_p): Likewise.
5089         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
5090         dr_analyze_innermost.
5091         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
5092         Convert return type from bool to opt_result, adding a message for
5093         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
5094         (vect_analyze_data_ref_dependence): Convert return type from bool
5095         to opt_result.  Change sense of return type from "false"
5096         effectively meaning "no problems" to "false" meaning a problem,
5097         so that "return false" becomes "return opt_result::success".
5098         Convert "return true" calls to opt_result::failure_at, using
5099         the location of statement A rather than vect_location.
5100         (vect_analyze_data_ref_dependences): Convert return type from bool
5101         to opt_result.
5102         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
5103         calls to opt_result::failure_at, using the stmt location rather
5104         than vect_location.
5105         (vect_verify_datarefs_alignment): Convert return type from bool
5106         to opt_result.
5107         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
5108         into multiple more-tightly-scoped copies.
5109         (vect_analyze_data_refs_alignment): Convert return type from bool
5110         to opt_result.
5111         (vect_analyze_data_ref_accesses): Likewise, converting a
5112         "return false" to a "return opt_result::failure_at", adding a
5113         new message.
5114         (vect_prune_runtime_alias_test_list): Convert return type from
5115         bool to opt_result, converting dump_printf_loc to
5116         opt_result::failure_at.  Add a %G to show the pertinent statement,
5117         and use the stmt's location rather than vect_location.
5118         (vect_find_stmt_data_reference): Convert return type from
5119         bool to opt_result, converting dump_printf_loc to
5120         opt_result::failure_at, using stmt's location.
5121         (vect_analyze_data_refs):  Convert return type from bool to
5122         opt_result.  Convert "return false" to "return
5123         opt_result::failure_at", adding messages as needed.
5124         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
5125         type from bool to opt_result.
5126         (vect_determine_vf_for_stmt): Likewise.
5127         (vect_determine_vectorization_factor): Likewise, converting
5128         dump_printf_loc to opt_result::failure_at, using location of phi
5129         rather than vect_location.
5130         (vect_analyze_loop_form_1): Convert return type from bool to
5131         opt_result, converting dump_printf_loc calls, retaining the use of
5132         vect_location.
5133         (vect_analyze_loop_form): Convert return type from loop_vec_info
5134         to opt_loop_vec_info.
5135         (vect_analyze_loop_operations): Convert return type from bool to
5136         opt_result, converting dump_printf_loc calls, using the location
5137         of phi/stmt rather than vect_location where available.  Convert
5138         various "return false" to "return opt_result::failure_at" with
5139         "unsupported phi" messages.
5140         (vect_get_datarefs_in_loop): Convert return type from bool to
5141         opt_result.  Add a message for the
5142         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
5143         (vect_analyze_loop_2): Convert return type from bool to
5144         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
5145         each "goto again;", adding new messages where needed.
5146         Add "unsupported grouped {store|load}" messages.
5147         (vect_analyze_loop): Convert return type from loop_vec_info to
5148         opt_loop_vec_info.
5149         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
5150         bool to opt_result.
5151         * tree-vect-stmts.c (process_use): Likewise, converting
5152         dump_printf_loc call and using stmt location, rather than
5153         vect_location.
5154         (vect_mark_stmts_to_be_vectorized): Likeise.
5155         (vect_analyze_stmt): Likewise, adding a %G.
5156         (vect_get_vector_types_for_stmt): Convert return type from bool to
5157         opt_result, converting dump_printf_loc calls and using stmt
5158         location, rather than vect_location.
5159         (vect_get_mask_type_for_stmt): Convert return type from tree to
5160         opt_tree, converting dump_printf_loc calls and using stmt location.
5161         * tree-vectorizer.c: Include "opt-problem.h.
5162         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
5163         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
5164         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
5165         enabled, use it to report at the top level "couldn't vectorize
5166         loop" followed by the problem.
5167         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
5168         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
5169         to opt_result.
5170         (vect_analyze_stmt): Likewise.
5171         (vect_get_vector_types_for_stmt): Likewise.
5172         (tree vect_get_mask_type_for_stmt): Likewise.
5173         (vect_analyze_data_ref_dependences): Likewise.
5174         (vect_enhance_data_refs_alignment): Likewise.
5175         (vect_analyze_data_refs_alignment): Likewise.
5176         (vect_verify_datarefs_alignment): Likewise.
5177         (vect_analyze_data_ref_accesses): Likewise.
5178         (vect_prune_runtime_alias_test_list): Likewise.
5179         (vect_find_stmt_data_reference): Likewise.
5180         (vect_analyze_data_refs): Likewise.
5181         (vect_analyze_loop): Convert return type from loop_vec_info to
5182         opt_loop_vec_info.
5183         (vect_analyze_loop_form): Likewise.
5184         (vect_analyze_slp): Convert return type from bool to opt_result.
5186 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5188         * doc/invoke.texi (-fopt-info): Document new "internals"
5189         sub-option.
5190         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
5191         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
5192         MSG_ALL_KINDS.
5193         (optinfo_verbosity_options): Add "internals".
5194         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
5195         (dump_context::apply_dump_filter_p): New member function.
5196         (dump_context::dump_loc): Use apply_dump_filter_p rather than
5197         explicitly masking the dump_kind.
5198         (dump_context::begin_scope): Increment the scope depth first.  Use
5199         apply_dump_filter_p rather than explicitly masking the dump_kind.
5200         (dump_context::emit_item): Use apply_dump_filter_p rather than
5201         explicitly masking the dump_kind.
5202         (dump_dec): Likewise.
5203         (dump_hex): Likewise.
5204         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
5205         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
5206         (opt_info_switch_p): Update handling of default
5207         MSG_OPTIMIZED_LOCATIONS to cope with default of
5208         MSG_PRIORITY_USER_FACING.
5209         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
5210         masking the dump_kind.
5211         (selftest::test_capture_of_dump_calls): Update test_dump_context
5212         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
5213         than MSG_ALL.  Generalize scope test to be run at all four
5214         combinations of with/without MSG_PRIORITY_USER_FACING and
5215         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
5216         for each of the two values.
5217         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
5218         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
5219         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
5220         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
5221         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
5222         with MSG_PRIORITY_*.
5223         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
5224         dump messages as MSG_PRIORITY_USER_FACING.
5225         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
5226         about the interaction with MSG_PRIORITY_*.
5228 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5230         * varasm.c (output_constant): Add new parameter merge_strings.
5231         Make strings properly zero terminated in merge string sections.
5232         (mergeable_string_section): Don't fail if the last char is non-zero.
5233         (assemble_variable_contents): Handle merge string sections.
5234         (assemble_variable): Likewise.
5235         (assemble_constant_contents): Likewise.
5236         (output_constant_def_contents): Likewise.
5237         (output_constructor_array_range,
5238         output_constructor_regular_field): Adjust call to output_constant.
5239         (output_object_block): Adjust call to assemble_constant_contents
5240         and assemble_variable_contents.
5242 2018-10-04  Martin Liska  <mliska@suse.cz>
5244         PR c/87483
5245         * cgraphunit.c (process_function_and_variable_attributes):
5246         Warn about a function with alias attribute and a body.
5248 2018-10-04  Martin Liska  <mliska@suse.cz>
5250         PR ipa/82625
5251         * multiple_target.c (redirect_to_specific_clone): New function.
5252         (ipa_target_clone): Use it.
5253         * tree-inline.c: Fix comment.
5255 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5257         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
5258         fields.
5259         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
5260         (gcc::dump_manager::register_pass): New member function, adapted
5261         from loop body in gcc::pass_manager::register_pass, adding a
5262         call to update_dfi_for_opt_info.
5263         (gcc::dump_manager::opt_info_enable_passes): Store the
5264         -fopt-info options into the new fields.  Move the loop
5265         bodies into...
5266         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
5267         function.
5268         * dumpfile.h (struct opt_pass): New forward decl.
5269         (gcc::dump_manager::register_pass): New decl.
5270         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
5271         (class gcc::dump_manager): Add fields "m_optgroup_flags",
5272         "m_optinfo_flags", and "m_optinfo_filename".
5273         * passes.c (gcc::pass_manager::register_pass): Move all of the
5274         dump-handling code to gcc::dump_manager::register_pass.
5276 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
5278         PR rtl-optimization/87466
5279         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
5280         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
5281         * doc/tm.texi: Regenerate.
5282         * ira-lives.c (process_bb_node_lives): Use the new target hook.
5283         * lra-lives.c (process_bb_lives): Likewise.
5284         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
5285         Define.
5287 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
5289         * params.c (add_params): Fix initialization.
5291 2018-10-04  Martin Liska  <mliska@suse.cz>
5293         PR gcov-profile/84107
5294         * tree-profile.c (init_ic_make_global_vars):
5295         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
5296         Come up with new ic_tuple* variables.  Emit
5297         __gcov_indirect_call{,_topn} variables.
5298         (gimple_gen_ic_profiler): Access the variable
5299         and emit gimple.
5300         (gimple_gen_ic_func_profiler): Access
5301         __gcov_indirect_call.callee field.
5302         (gimple_init_gcov_profiler): Use ptr_type_node.
5303         * value-prof.c (gimple_ic): Use ptr_type_node.
5305 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5307         PR tree-optimization/85787
5308         * ipa-pure-const.c (malloc_candidate_p_1): Move most of
5309         malloc_candidate_p into this function and add support for
5310         detecting multiple phis.
5311         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
5313 2018-10-04  Martin Liska  <mliska@suse.cz>
5315         PR ipa/87491
5316         * ipa-inline.c (inline_to_all_callers_1):
5317         Call ultimate_alias_target for node being inlined.
5319 2018-10-03  Jeff Law  <law@redhat.com>
5321         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
5322         target's wchar_t.
5323         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
5324         * tree.h (get_typenode_from_name): Prototype.
5326 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
5328         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
5329         Change operand 2 predicate to nonimmediate_operand.
5330         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
5332 2018-10-03  Martin Sebor  <msebor@redhat.com>
5333             Jeff Law  <law@redhat.com>
5335         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
5336         initialize it.
5337         (get_string_length): Detect unterminated arrays.
5338         (format_string): Same.
5339         (format_directive): Warn about unterminated arrays.
5340         (handle_gimple_call): Mark statements with no_warning as needed.
5342 2018-10-03  Jim Wilson  <jimw@sifive.com>
5344         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
5345         also define __riscv_abi_rve.  Delete trailing white space.
5347 2018-10-03  Paul Koning  <ni1d@arrl.net>
5349         Enable LRA register allocator for PDP11.
5350         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
5351         (R): Likewise.
5352         (D): Likewise.
5353         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
5354         * config/pdp11/pdp11.opt (-mlra): New option.
5355         * doc/invoke.texi (PDP-11 Options): Document -mlra.
5357 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
5359         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
5360         (*<absneg:code>extend<mode>xf2): Ditto.
5362 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
5364         PR tree-optimization/87415
5365         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
5366         precision fields.
5368 2018-10-02  Jeff Law  <law@redhat.com>
5370         * gimple-fold.c (get_range_strlen): Only set *nonstr when
5371         an unterminated string is discovered.  Bubble up range
5372         even for unterminated strings.
5373         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
5374         indicates the string was not terminated via NONSTR.
5376 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
5378         * tree-vrp.c (extract_range_from_unary_expr): Special case all
5379         pointer conversions.
5380         Do not do anything special for anti-ranges.
5382 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
5384         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
5385         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
5387 2018-10-03  Martin Liska  <mliska@suse.cz>
5389         PR gcov-profile/86109
5390         * coverage.c (coverage_begin_function): Do not
5391         mark lambdas as artificial.
5392         * tree-core.h (struct GTY): Remove tm_clone_flag
5393         and introduce new lambda_function.
5394         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
5396 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
5398         PR target/87474
5399         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
5400         P8_VECTOR and VSX are enabled.
5402 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
5404         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
5405         0x3907 as CPU model number.
5407 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
5409         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
5410         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
5411         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
5412         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
5413         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
5414         * config/s390/s390.md: Likewise. Rename also the cpu attribute
5415         value from arch12 to z14.
5417 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
5419         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
5420         (isinfxf2): Ditto.
5421         (isinf<mode>2): Ditto.
5423 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
5425         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
5426         before emitting fxam.  Perform calculations in XFmode.
5428 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5430         * match.pd (((X /[ex] A) +- B) * A): New transformation.
5432 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5434         PR middle-end/87319
5435         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
5436         * tree.c (signed_or_unsigned_type_for): Handle complex.
5438 2018-10-02  Jeff Law  <law@redhat.com>
5440         * gimple-fold.c (get_range_strlen): Remove dead code.
5442 2018-10-02  Martin Sebor  <msebor@redhat.com>
5443             Jeff Law  <law@redhat.com>
5445         * builtins.c (unterminated_array): Add new arguments.
5446         If argument is not terminated, bubble up size and exact
5447         state to callers.
5448         (expand_builtin_strnlen): Detect, avoid expanding
5449         and diagnose unterminated arrays.
5450         (c_strlen): Fill in offset of start of unterminated strings.
5451         * builtins.h (unterminated_array): Update prototype.
5453 2018-10-02  Richard Biener  <rguenther@suse.de>
5455         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
5456         of haddv4df, first reduce to SSE width and exploit the fact
5457         that we only need element zero with the reduction result.
5458         (reduc_plus_scal_v2df): Likewise.
5460 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5462         * dojump.h (do_jump): Delete.
5463         (do_jump_1): Likewise.
5464         (split_comparison): Move around.
5465         * dojump.c (do_jump): Make static.
5466         (do_jump_1): Likewise.
5467         (jumpifnot): Move around.
5468         (jumpifnot_1): Likewise.
5469         (jumpif): Likewise.
5470         (jumpif_1): Likewise.
5471         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
5473 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5475         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
5476         insns in the delay slot and add_insn_after for the jump insn.
5478 2018-10-02  Richard Biener  <rguenther@suse.de>
5480         * tree-inline.c (expand_call_inline): Use the location of
5481         the callee declaration for the inline-entry marker.
5482         * final.c (notice_source_line): Remove special-casing of
5483         NOTE_INSN_INLINE_ENTRY.
5485 2018-10-01  Carl Love  <cel@us.ibm.com>
5487         PR 69431
5488         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
5489         (__builtin_mtfsb0): New.
5490         (__builtin_mtfsb1): New.
5491         ( __builtin_set_fpscr_rn): New.
5492         (__builtin_set_fpscr_drn): New.
5493         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
5494         (rs6000_expand_set_fpscr_rn_builtin): Add.
5495         (rs6000_expand_set_fpscr_drn_builtin): Add.
5496         (rs6000_expand_builtin): Add case statement entries for
5497         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
5498         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
5499         RS6000_BUILTIN_MFFSL.
5500         (rs6000_init_builtins): Add ftype initialization and def_builtin
5501         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
5502         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
5503         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
5504         rs6000_mffscdrn): Add define_insn.
5505         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
5506         * doc/extend.texi: Add documentation for the builtins.
5508 2018-10-01  Richard Biener  <rguenther@suse.de>
5510         PR tree-optimization/87465
5511         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
5512         causing branch miscounts.
5514 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5516         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
5517         aarch64_option_default_param):  New.
5518         (params.h): Include.
5519         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
5520         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
5521         stack-clash protection validation code.
5523 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5525         * params.c (validate_param): New.
5526         (add_params): Use it.
5527         (set_param_value): Refactor param validation into validate_param.
5528         (diagnostic.h): Include.
5529         * diagnostic.h (diagnostic_ready_p): New.
5531 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5533         * params.c (set_param_value):
5534         Add index of parameter being validated.
5535         * common/common-target.def (option_validate_param): New.
5536         * common/common-targhooks.h (default_option_validate_param): New.
5537         * common/common-targhooks.c (default_option_validate_param): New.
5538         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
5539         * doc/tm.texi: Regenerate.
5541 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5543         PR target/86486
5544         * config/aarch64/aarch64.c (aarch64_override_options_internal):
5545         Add validation for stack-clash parameters and set defaults.
5547 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5549         PR target/86486
5550         * configure.ac: Add stack-clash-protection-guard-size.
5551         * doc/install.texi: Document it.
5552         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
5553         * params.def: Update comment for guard-size.
5554         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
5555         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
5556         * configure: Regenerate.
5558 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5560         PR target/86486
5561         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
5562         STACK_DYNAMIC_OFFSET): New.
5563         * config/aarch64/aarch64.c (aarch64_layout_frame):
5564         Update outgoing args size.
5565         (aarch64_stack_clash_protection_alloca_probe_range,
5566         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5568 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5570         PR target/86486
5571         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
5572         probe ranges.
5573         * target.def (stack_clash_protection_alloca_probe_range): New.
5574         (stack_clash_protection_final_dynamic_probe): Remove.
5575         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
5576         (default_stack_clash_protection_final_dynamic_probe): Remove.
5577         * targhooks.c: Likewise.
5578         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5579         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
5580         * doc/tm.texi: Regenerate.
5582 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5584         PR target/86486
5585         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
5586         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
5587         aarch64_clamp_to_uimm12_shift): New.
5588         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
5589         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
5591 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5593         PR target/86486
5594         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
5596 2018-10-01  Jeff Law  <law@redhat.com>
5597             Richard Sandiford <richard.sandiford@linaro.org>
5598             Tamar Christina  <tamar.christina@arm.com>
5600         PR target/86486
5601         * config/aarch64/aarch64.md
5602         (probe_stack_range): Add k (SP) constraint.
5603         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
5604         STACK_CLASH_MAX_UNROLL_PAGES): New.
5605         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
5606         stack probes for stack clash.
5607         (aarch64_allocate_and_probe_stack_space): New.
5608         (aarch64_expand_prologue): Use it.
5609         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
5610         (aarch64_sub_sp): Add emit_move_imm optional param.
5612 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
5614         PR tree-optimization/87261
5615         * match.pd: Remove trailing whitespace.
5616         Add (x & y) | ~(x | y) -> ~(x ^ y),
5617         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
5619 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5621         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
5622         constraints.
5623         (ashlsi3_insn): Update instruction constraints.
5624         (ashrsi3_insn): Likewise.
5625         (rotrsi3): Likewise.
5626         (add_shift): Likewise.
5627         * config/arc/constraints.md (Csz): New 32 bit constraint. It
5628         avoids placing in the limm field small constants which, otherwise,
5629         could end into a small instruction.
5631 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5633         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
5634         destination register is not odd-even.
5635         (umaddsidi4_split): Likewise.
5637 2018-10-01  Richard Biener  <rguenther@suse.de>
5639         * tree-inline.c (expand_call_inline): Store origin of fn
5640         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
5641         * tree.c (block_ultimate_origin): Simplify and do some
5642         checking.
5644 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
5646         * config/i386/mmx.md (EMMS): New int iterator.
5647         (emms): New int attribute.
5648         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
5649         EMMS int iterator.  Explicitly declare clobbers.
5650         (mmx_emms): Remove expander.
5651         (mmx_femms): Ditto.
5652         * config/i386/predicates.md (emms_operation): Remove predicate.
5653         (vzeroall_pattern): New predicate.
5654         (vzeroupper_pattern): Rename from vzeroupper_operation.
5655         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
5656         vzeroupper_pattern and vzeroall_pattern predicates.
5658 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
5660         PR rtl-optimization/86939
5661         * ira-lives.c (make_hard_regno_born): Rename from this...
5662         (make_hard_regno_live): ... to this.  Remove update to conflict
5663         information.  Update function comment.
5664         (make_hard_regno_dead): Add conflict information update.  Update
5665         function comment.
5666         (make_object_born): Rename from this...
5667         (make_object_live): ... to this.  Remove update to conflict information.
5668         Update function comment.
5669         (make_object_dead):  Add conflict information update.  Update function
5670         comment.
5671         (mark_pseudo_regno_live): Call make_object_live.
5672         (mark_pseudo_regno_subword_live): Likewise.
5673         (mark_hard_reg_dead): Update function comment.
5674         (mark_hard_reg_live): Call make_hard_regno_live.
5675         (process_bb_node_lives): Likewise.
5676         * lra-lives.c (make_hard_regno_born): Rename from this...
5677         (make_hard_regno_live): ... to this.  Remove update to conflict
5678         information.  Remove now uneeded check_pic_pseudo_p argument.
5679         Update function comment.
5680         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
5681         to conflict information.  Update function comment.
5682         (mark_pseudo_live): Remove update to conflict information.  Update
5683         function comment.
5684         (mark_pseudo_dead): Add conflict information update.
5685         (mark_regno_live): Call make_hard_regno_live.
5686         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
5687         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
5689 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
5691         PR target/87370
5692         * config/i386/i386.c (construct_container): Use TImode for
5693         BLKmode values in 2 integer registers.
5695 2018-09-29  Jeff Law  <law@redhat.com>
5697         * builtins.c (unterminated_array): Pass in c_strlen_data * to
5698         c_strlen rather than just a tree *.
5699         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
5700         Update recursive calls appropriately.  If caller did not provide a
5701         suitable data pointer, create a local one.  When a non-terminated
5702         string is discovered, bubble up information about the string via the
5703         c_strlen_data object.
5704         * builtins.h (c_strlen): Update prototype.
5705         (c_strlen_data): New structure.
5706         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
5707         For a type 2 call, if c_strlen indicates a non-terminated string
5708         use the length of the non-terminated string.
5709         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
5711 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
5713         PR target/87467
5714         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
5715         __m512d type for __A argument rather than __m512.
5717 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
5719         * match.pd (simple_comparison): Don't optimize if either operand is
5720         a function pointer when target needs function pointer canonicalization.
5722 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5724         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
5725         power5 .. power9 to remove indirection.
5726         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
5727         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
5728         ASM_CPU_476_SPEC): Delete.
5729         (ASM_CPU_SPEC): Adjust.
5730         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
5731         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
5733 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5735         * config.in: Delete HAVE_AS_DCI.
5736         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
5737         * config/rs6000/rs6000.h: Ditto.
5738         * configure.ac: Delete HAVE_AS_DCI.
5739         * configure: Regenerate.
5741 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5743         * config.in (HAVE_AS_LWSYNC): Delete.
5744         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5745         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
5746         do it as a .long .
5747         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5748         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
5749         as a .long .
5750         * configure.ac: Delete HAVE_AS_LWSYNC.
5751         * configure: Regenerate.
5753 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5754             Pierre-Marie de Rodat  <derodat@adacore.com>
5756         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
5757         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
5758         (cgraph_node::create_thunk): Add indirect_offset parameter.
5759         (thunk_adjust): Likewise.
5760         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
5761         and initialize the corresponding field with it.
5762         (cgraph_node::dump): Dump indirect_offset field.
5763         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
5764         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
5765         (thunk_adjust): Add indirect_offset parameter and deal with it.
5766         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
5767         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
5768         or if the thunk is external or local.  Fix formatting.  Do not chain
5769         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
5770         if any, in the GIMPLE representation.
5771         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
5772         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
5773         (input_node): Read indirect_offset field.
5774         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
5775         call to thunk_adjust.
5776         * tree-nested.c (struct nesting_info): Add thunk_p field.
5777         (create_nesting_tree): Set it.
5778         (convert_all_function_calls): Copy static chain from targets to thunks.
5779         (finalize_nesting_tree_1): Return early for thunks.
5780         (unnest_nesting_tree_1): Do not finalize thunks.
5781         (gimplify_all_functions): Do not gimplify thunks.
5783 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5785         * opt-suggestions.c (option_proposer::build_option_suggestions):
5786         Release "option_values".
5788 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5790         * coverage.c (get_coverage_counts): Convert problem-reporting dump
5791         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
5792         * dumpfile.c (kind_as_string): New function.
5793         (dump_loc): Rather than a hardcoded prefix of "note: ", use
5794         kind_as_string to vary the prefix based on dump_kind.
5795         (selftest::test_capture_of_dump_calls): Update for above.
5797 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5799         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
5800         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
5802 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5804         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
5805         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
5806         INVALID_REGNUM instead of FPSR_REG.
5807         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
5808         * config/i386/i386.md: Update comment of FP compares.
5809         (fldenv): Do not clobber FPSR_REG.
5811 2018-09-28  Richard Biener  <rguenther@suse.de>
5813         * tree.h (BLOCK_ORIGIN): New.
5814         * omp-expand.c (grid_expand_target_grid_body): Assign
5815         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
5816         * tree-inline.c (remap_block): Likewise.
5817         * auto-profile.c (get_function_decl_from_block): Simplify
5818         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
5819         * langhooks.c (lhd_print_error_function): Likewise.
5820         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
5821         Likewise.
5822         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5823         * tree.c (block_nonartificial_location): Likewise.
5824         (block_ultimate_origin): Likewise.
5825         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
5826         no longer needed LTO case.
5828 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
5829             Jan Hubicka  <jh@suse.cz>
5830             Martin Jambor  <mjambor@suse.cz>
5832         * simplify-rtx.c (simplify_merge_mask): New function.
5833         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
5834         same masks are used in op1 or op2.
5835         (test_vec_merge): New function.
5836         (test_vector_ops): Call test_vec_merge.
5838 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5840         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
5841         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
5842         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
5843         (cypress_costs): Set it.
5844         (supersparc_costs): Likewise.
5845         (hypersparc_costs): Likewise.
5846         (leon_cost): Likewise.
5847         (leon3_costs): Likewise.
5848         (sparclet_costs): Likewise.
5849         (ultrasparc_costs): Likewise.
5850         (ultrasparc_costs): Likewise.
5851         (niagara_costs): Likewise.
5852         (niagara2_costs): Likewise.
5853         (niagara3_costs): Likewise.
5854         (niagara4_costs): Likewise.
5855         (niagara7_costs): Likewise.
5856         (m8_costs): Likewise.
5857         (TARGET_CAN_FOLLOW_JUMP): Define.
5858         (pass_work_around_errata::gate): Minor tweak.
5859         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
5860         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
5861         Automaitcally clear MASK_FSMULD mask for V7 processors.
5862         (sparc_can_follow_jump): New static function.
5863         (output_ubranch): Deal with CROSSING_JUMP_P.
5864         (sparc_use_sched_lookahead): Rewrite using switch statement.
5865         (sparc_issue_rate): Reorder.
5866         (sparc_branch_cost): New function.
5868 2018-09-27  Martin Sebor  <msebor@redhat.com>
5870         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
5871         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
5872         (int_fits_type_p): Same.
5874 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
5876         * config/i386/i386.md (FPCR_REG): Remove.
5877         (UNSPEC_FLDCW): Remove.
5878         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
5879         (x86_fldcw_1): Remove insn pattern.
5880         (fnstenv): Do not clobber FPCR_REG.
5881         (fldenv): Ditto.
5882         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
5883         (CALL_USED_REGISTERS): Ditto.
5884         (REG_ALLOC_ORDER): Ditto.
5885         (REG_CLASS_CONTENTS): Ditto.
5886         (HI_REGISTER_NAMES): Ditto.
5887         (ADDITIONAL_REGISTER_NAMES): Use defines instead
5888         of numerical constants.
5889         * config/i386/i386.c (regclass_map): Remove fpsr register.
5890         (dbx_register_map): Ditto.
5891         (dbx64_register_map): Ditto.
5892         (svr4_dbx_register_map): Ditto.
5893         (print_reg): Do not handle FPCR_REG.
5895 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
5897         PR target/87149
5898         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
5899         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
5900         Delete, always treat as true.
5901         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5902         Ditto.  Simplify remaining code.
5903         * config/powerpcspe/powerpcspe.h: Ditto.
5904         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
5905         Simplify remaining code.
5906         (rs6000_expand_builtin): Ditto.
5907         * config/rs6000/rs6000.h: Ditto.
5908         * configure.ac: Ditto.
5909         * configure: Regenerate.
5911 2018-09-27  Martin Liska  <mliska@suse.cz>
5913         * coverage.c (get_coverage_counts): Revert the formatting
5914         of missing profile opt info.
5916 2018-09-27  Richard Biener  <rguenther@suse.de>
5918         PR debug/37801
5919         PR debug/87440
5920         * dwarf2out.c (set_block_origin_self): Do not mark outermost
5921         block as we do not output that.
5922         (gen_inlined_subroutine_die): Elide the originally outermost
5923         block, matching what we do for concrete instances.
5924         (decls_for_scope): Add parameter specifying whether to recurse
5925         to subblocks.
5927 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
5928             Tom de Vries  <tom@codesourcery.com>
5930         PR 82089
5932         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
5933         STORE_FLAG_VALUE == 1.
5935 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
5937         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
5938         constant definitions.
5939         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
5940         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
5941         ("speculation_barrier"): New expander definition.
5943 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
5945         PR gcov-profile/86957
5946         * common.opt: New warning option -Wmissing-profile.
5947         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
5948         * doc/invoke.texi: Document -Wmissing-profile.
5950 2018-09-26  Jim Wilson  <jimw@sifive.com>
5952         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
5953         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
5955 2018-09-26  Martin Sebor  <msebor@redhat.com>
5957         * tree.c (zerop): Change return type to bool.
5958         (integer_zerop, integer_onep, integer_each_onep): Same.
5959         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5960         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5961         (real_onep, real_minus_onep, chain_index): Same.
5962         (print_type_hash_statistics, type_list_equal): Same.
5963         * tree.h (zerop): Same.
5964         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
5965         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5966         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5967         (real_onep, real_minus_onep, chain_index): Same.
5968         (print_type_hash_statistics, type_list_equal): Same.
5970 2018-09-26  Jim Wilson  <jimw@sifive.com>
5972         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
5974 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
5976         PR target/87414
5977         * config/i386/i386.c: Include debug.h and dwarf2out.h.
5978         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
5979         call.
5981 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
5983         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
5985 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5987         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
5988         and FP_SECOND_SSE_REGS.
5989         (REG_CLASS_NAMES): Ditto.
5990         (REG_CLASS_CONTENTS): Ditto.
5991         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
5992         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
5993         (ix86_preferred_output_reload_class): Ditto.
5994         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
5995         clobber constraint to "=&f".
5996         (fix_truncdi_i387): Ditto.
5997         (lrintxfdi2): Ditto.
5998         (fistdi2_<rounding>): Ditto.
5999         (fpremxf4_i387): Change "=u" constraint to "=f".
6000         (fprem1xf4_i387): Ditto.
6001         (sincosxf3): Ditto.
6002         (fptanxf4_i387): Ditto.
6003         (fxtractxf3_i387): Ditto.
6004         (fscalexf4_i387): Ditto.
6005         (atan2xf3): Change "u" constraint to "f".
6006         (fyl2xxf3_i387): Ditto.
6007         (fyl2xp1xf3_i387): Ditto.
6009 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
6011         PR target/87439
6012         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
6013         for removed I387_MASK_PM entity.
6016 2018-09-26  Jeff Law  <law@redhat.com>
6017         Revert
6018         2018-09-26  Alexey Neyman  <stilor@att.net>
6020         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
6021         headers are no longer pulled in by <isl/val.h>.
6023 2018-09-26  Richard Biener  <rguenther@suse.de>
6025         PR debug/87443
6026         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
6027         or concrete instance DIE to the tree.  Create abstract origin
6028         attributes also for concrete instances.
6030 2018-09-26  Alexey Neyman  <stilor@att.net>
6032         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
6033         headers are no longer pulled in by <isl/val.h>.
6035 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
6037         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
6038         Use new helper functions.
6039         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
6040         Use new helper functions.
6041         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
6042         aarch_mm_needs_release): New declarations.
6043         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
6044         aarch_mm_needs_release): New.
6046 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
6048         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
6049         (arm32_output_mi_thunk): Deal with long calls.
6051 2018-09-26  Richard Biener  <rguenther@suse.de>
6053         PR debug/87428
6054         PR debug/87362
6055         * tree-inline.c (expand_call_inline): When the location
6056         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
6057         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
6058         the inserted BLOCK to make inlined_function_outer_scope_p
6059         recognize it.
6060         * dwarf2out.c (add_call_src_coords_attributes): Do not add
6061         coords for reserved locations.
6063 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6065         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
6066         (*call_indirect_nonlocal_sysv<mode>): Ditto.
6067         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
6068         (*sibcall_nonlocal_sysv<mode>): Ditto.
6069         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
6070         (<bd>_<mode>): Ditto.
6071         (<bd>tf_<mode>): Ditto.
6073 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6075         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
6076         control string as a list of templates instead of as C code.
6077         (*altivec_movti): Ditto.
6078         * config/rs6000/darwin.md (movdf_low_di): Ditto.
6080 2018-09-25  Jim Wilson  <jimw@sifive.com>
6082         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
6083         when target symbol is weak.
6085 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6087         PR c/87387
6088         * builtins.c (unterminated_array): Simplify.
6089         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
6090         where pointer arithmetic is safe.
6092 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6094         PR target/86987
6095         * config/rs6000/altivec.md (altivec_vspltb): Use
6096         const_0_to_15_operand instead of u5bit_cint_operand.
6097         (*altivec_vspltb_internal): Ditto.
6098         (altivec_vspltb_direct): Ditto.
6099         (altivec_vsplth): Use const_0_to_7_operand instead of
6100         u5bit_cint_operand.
6101         (*altivec_vsplth_internal): Ditto.
6102         (altivec_vsplth_direct): Ditto.
6103         (altivec_vspltw): Use const_0_to_3_operand instead of
6104         u5bit_cint_operand.
6105         (*altivec_vspltw_internal): Ditto.
6106         (altivec_vspltw_direct): Ditto.
6107         (altivec_vspltsf): Ditto.
6108         (*altivec_vspltsf_internal): Ditto.
6109         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
6110         various splats with the proper size immediate.  Reorder the various
6111         cases by ascending size of immediate, and put all such together.
6113 2018-09-25  Richard Biener  <rguenther@suse.de>
6115         PR debug/83941
6116         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
6117         GC-ification.
6118         (maybe_create_die_with_external_ref): Do not create
6119         DW_TAG_imported_unit here.
6120         (add_abstract_origin_attribute): Handle external BLOCK refs.
6121         (dwarf2out_abstract_function): Simplify LTO case.
6122         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
6123         rather than using maybe_create_die_with_external_ref.
6125 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
6127         PR target/71278
6128         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
6129         (frndintxf2_mask_pm_i387): Ditto.
6130         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
6131         Enable for !flag_trapping_math.
6132         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
6133         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
6134         Change operand 1 predicate to nonimmediate_operand.
6135         (attr "i387_cw"): Remove mask_pm.
6136         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
6137         (enum ix86_entity): Remove I387_MASK_PM.
6138         * config/i386/i386.c (ix86_i387_mode_needed): Do not
6139         handle I387_MASK_PM.
6140         (ix86_mode_needed): Ditto.
6141         (ix86_mode_after): Ditto.
6142         (ix86_mode_entry): Ditto.
6143         (ix86_mode_exit): Ditto.
6144         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
6146 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
6148         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
6149         to_update_switch_stmts to vNULL instead of calling create on them
6150         immediately.
6152 2018-09-25  Richard Biener  <rguenther@suse.de>
6154         PR tree-optimization/87402
6155         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
6156         (visit_phi): Re-instantiate handling of supposed to be VARYING
6157         but non-VARYING backedge value.
6159 2018-09-25  Richard Biener  <rguenther@suse.de>
6161         PR debug/83941
6162         * dwarf2out.c (struct sym_off_pair): New.
6163         (external_die_map): New global.
6164         (lookup_decl_die): When in LTO create DIEs lazily from the
6165         external_die_map.
6166         (lookup_block_die): New function, create DIEs lazily in LTO.
6167         (equate_block_to_die): New function.
6168         (dwarf2out_die_ref_for_decl): During WPA get the association
6169         from the external DIE map.
6170         (dwarf2out_register_external_die): Record mapping into the
6171         external DIE map.
6172         (maybe_create_die_with_external_ref): New function split out from
6173         DIE generation part of old dwarf2out_register_external_die.
6174         (add_abstract_origin_attribute): Do not return the DIE.  When
6175         in LTO reference externals directly.
6176         (dwarf2out_abstract_function): When in LTO ignore calls for
6177         decls with external DIEs (already present abstract instances).
6178         (gen_call_site_die): Adjust.
6179         (add_high_low_attributes): Likewise.
6180         (gen_lexical_block_die): Likewise.
6181         (gen_inlined_subroutine_die): Likewie.
6182         (gen_block_die): Likewise.
6183         (dwarf2out_inline_entry): Likewise.
6184         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
6185         DIEs.
6187 2018-09-25  Martin Liska  <mliska@suse.cz>
6189         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
6190         integers and not by a float value.
6192 2018-09-25  Martin Liska  <mliska@suse.cz>
6194         PR fortran/87394
6195         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
6196         instead of NULL.
6197         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
6198         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
6199         * godump.c (go_format_type): Remove extra parenthesis.
6201 2018-09-25  Martin Liska  <mliska@suse.cz>
6203         * alias.c (set_dest_equal_p): Remove unused function.
6204         * config/i386/i386.c (def_builtin_pure2): Likewise.
6205         * diagnostic-show-locus.c (class layout): Remove
6206         unused field.
6207         (layout::layout): Likewise here.
6208         * dump-context.h (class temp_dump_context): Likewise.
6209         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
6210         (add_AT_loclistsptr): Likewise.
6211         (add_AT_offset): Likewise.
6212         (get_AT_hi_pc): Likewise.
6213         (is_comdat_die): Likewise.
6214         (type_is_enum): Likewise.
6215         (ceiling): Likewise.
6216         (add_AT_vms_delta): Likewise.
6217         (is_class_die): Likewise.
6218         * edit-context.c (class line_event): Remove unused field.
6219         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
6220         unused function.
6221         * ipa-cp.c (ipa_get_vr_lat): Likewise.
6222         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
6223         (ok_for_base_p_nonstrict): Likewise.
6224         * tree-chrec.c (is_not_constant_evolution): Likewise.
6225         (chrec_fold_poly_cst): Likewise.
6226         * tree-if-conv.c (has_pred_critical_p): Likewise.
6227         * tree-ssa-coalesce.c (print_exprs): Likewise.
6228         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
6229         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
6230         * tree-vrp.c (value_ranges_intersect_p): Likewise.
6231         (value_range_nonnegative_p): Likewise.
6233 2018-09-25  Martin Liska  <mliska@suse.cz>
6235         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
6236         Do not handle "GNU Pascal".
6237         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
6238         Likewise.
6239         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
6240         from documentation. Likewise.
6241         * dbxout.c (dbxout_range_type): Likewise.
6242         * doc/cpp.texi: Likewise.
6243         * doc/extend.texi: Likewise.
6244         * doc/frontends.texi: Likewise.
6245         * doc/invoke.texi: Remove Pascal entry.
6246         * tree.def (CLEANUP_POINT_EXPR): Likewise.
6247         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
6249 2018-09-25  Martin Liska  <mliska@suse.cz>
6251         PR middle-end/86078
6252         * doc/invoke.texi: Document all parameters and remove default
6253         of the parameters.
6255 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
6257         PR bootstrap/87417
6258         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
6259         contains HOST_WIDE_INTs when computing its size.
6261 2018-09-24  Jim Wilson  <jimw@sifive.com>
6263         PR target/87391
6264         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
6265         not TARGET_RVE.
6266         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
6268 2018-09-24  Andrew Pinski  <apinski@marvell.com>
6270         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
6271         access prev before checking it for NULLness in the
6272         AARCH64_FUSE_CMP_BRANCH case.
6274 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
6276         PR target/82699
6277         * config/i386/i386.c (rest_of_insert_endbranch): Set
6278         endbr_queued_at_entrance to true and don't insert ENDBR if
6279         x86_function_profiler will be called.
6280         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
6281         is true.
6282         * config/i386/i386.h (machine_function): Add
6283         endbr_queued_at_entrance.
6285 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6287         * genattrtab.c (mk_attr_alt): Use alternative_mask.
6288         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
6289         types.
6290         (check_attr_test): Use alternative_mask.
6291         (get_attr_value): Likewise.
6292         (compute_alternative_mask): Use alternative_mask and XWINT.
6293         (make_alternative_compare): Use alternative_mask.
6294         (attr_alt_subset_p): Use XWINT.
6295         (attr_alt_subset_of_compl_p): Likewise.
6296         (attr_alt_intersection): Use alternative_mask and XWINT.
6297         (attr_alt_union): Likewise.
6298         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
6299         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
6300         (simplify_test_exp): Use alternative_mask and XWINT.
6301         (write_test_expr): Use alternative_mask and XWINT, adjust bit
6302         number calculation to support 64 bits.  Generate code that
6303         checks 64-bit masks.
6304         (main): Use alternative_mask.
6305         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
6307 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6309         PR target/80080
6310         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
6311         RETURN+USE when returning via %r14.
6313 2018-09-24  Martin Liska  <mliska@suse.cz>
6315         * gcov.c (output_lines): Print colorization legend
6316         for both flag_use_colors and flag_use_hotness_colors.
6317         Reword the help.
6319 2018-09-24  Martin Liska  <mliska@suse.cz>
6321         * coverage.c (get_coverage_counts): Use warning_at
6322         with current_function_decl location. Use %qD in warning
6323         message.
6325 2018-09-24  Martin Liska  <mliska@suse.cz>
6327         * memory-block.h (memory_block_pool::release): Annotate with
6328         valgrind that the memory is not accessible.
6330 2018-09-24  Martin Liska  <mliska@suse.cz>
6332         PR sanitizer/85774
6333         * asan.c: Make asan_handled_variables extern.
6334         * asan.h: Likewise.
6335         * cfgexpand.c (expand_stack_vars): Make sure
6336         a representative is unpoison if another
6337         variable in the partition is handled by
6338         use-after-scope sanitization.
6340 2018-09-24  Richard Biener  <rguenther@suse.de>
6342         PR tree-optimization/63155
6343         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
6344         the worklist when the edge of the respective argument isn't
6345         executable.
6347 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6349         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
6350         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
6351         (MASK_CLASS_P): Update for rename.
6352         (MAYBE_MASK_CLASS_P): Ditto.
6353         (REG_CLASS_NAMES): Update.
6354         (REG_CLASS_CONTENT): Update.
6355         * config/i386/i386.c (regclass_map): Update for MASK_REG
6356         and ALL_MASK_REGS rename.
6357         * config/i386/constraints.md (Yk): Update for rename.
6358         (k): Ditto.
6360 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6362         * config/i386/i386.h (enum reg_class): Remove
6363         EVEX_SSE_REGS and MOD4_SSE_REGS.
6364         (REG_CLASS_NAMES): Update.
6365         (REG_CLASS_CONTENT): Update.
6366         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
6367         registers as ALL_SSE_REGS.
6368         (ix86_additional_allocno_class_p): Remove.
6369         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
6370         (ix86_register_priority): Lower priority of EVEX SSE registers.
6371         Use IN_RANGE macro where appropriate.
6372         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
6373         AVX-5124VNNIW checks.
6374         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
6375         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
6376         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
6377         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
6378         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
6379         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
6380         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
6381         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
6382         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
6383         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
6384         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
6385         * config/i386/constraints.md (Yh): Remove.
6387 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6389         * config/i386/i386.c (regclass_map): Declare integer REX registers
6390         as GENERAL_REGS.
6392 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
6394         * doc/service.texi (Service): Switch the fsf.org link to https.
6396 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
6398         PR target/86798
6399         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6400         Define to speculation_safe_value_not_needed.
6402 2018-09-21  Florian Weimer  <fweimer@redhat.com>
6404         PR middle-end/81035
6405         * doc/extend.texi (Common Function Attributes): Mention that
6406         noreturn suppresses tail call optimization.
6408 2018-09-21  Jeff Law  <law@redhat.com>
6410         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
6411         vr_values::cleanup_edges_and_switches.
6412         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
6413         vr_values class.
6414         (identify_jump_threads): Remove EDGE_IGNORE handling.
6415         (execute_vrp): Move handling of to_remove_edges and
6416         to_update_switch_stmts into vr_values class member functions.
6417         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
6418         (to_update_switch_stmts): Likewise.
6419         * vr-values.c: Include cfghooks.h.
6420         (vr_values::vr_values): Initialize to_remove_edges and
6421         to_update_switch_stmts.
6422         (vr_values::~vr_values): Verify to_remove_edges and
6423         to_update_switch_stmts are empty.
6424         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
6425         (vr_values::cleanup_edges_and_switches): New member function.
6426         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
6427         function.  Add new data members.
6429 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
6431         PR tree-optimization/87309
6432         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
6433         calls with pflags and alt_flags.
6434         (selftest::test_capture_of_dump_calls): Add test of interaction of
6435         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
6437 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6439         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
6441 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6443         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
6445 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6447         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
6448         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
6449         (VXWORKS_LIBS_RTP): Minor reordering.
6451 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6453         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
6454         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
6456 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6458         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
6459         (PTRDIFF_TYPE): Likewise.
6461 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6463         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
6464         triplet, similar to support for VxWorks7.
6465         * config/vxworks-dummy.h: Provide a default definition
6466         of TARGET_VXWORKS64 to 0.
6468 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6470         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
6471         * config/vxworks-dummy.h: here.
6473 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6475         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
6477 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
6478             Bo Zhou  <zbo.zhou@hisilicon.com>
6480         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
6481         * config/aarch64/aarch64-tune.md: Regenerated.
6482         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
6483         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
6484         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
6486 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
6487             Julian Brown  <julian@codesourcery.com>
6489         * builtins.c (get_builtin_sync_mem): Handle address spaces.
6491 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6493         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
6494         if the call takes a static chain.
6496 2018-09-21  Martin Liska  <mliska@suse.cz>
6498         * auto-profile.c (autofdo_source_profile::read): Do not
6499         set sum_all.
6500         (read_profile): Do not add working sets.
6501         (read_autofdo_file): Remove sum_all.
6502         (afdo_callsite_hot_enough_for_early_inline): Remove const
6503         qualifier.
6504         * coverage.c (struct counts_entry): Remove gcov_summary.
6505         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
6506         do not support GCOV_TAG_PROGRAM_SUMMARY.
6507         (get_coverage_counts): Remove summary and expected
6508         arguments.
6509         * coverage.h (get_coverage_counts): Likewise.
6510         * doc/gcov-dump.texi: Remove -w option.
6511         * gcov-dump.c (dump_working_sets): Remove.
6512         (main): Do not support '-w' option.
6513         (print_usage): Likewise.
6514         (tag_summary): Likewise.
6515         * gcov-io.c (gcov_write_summary): Do not dump
6516         histogram.
6517         (gcov_read_summary): Likewise.
6518         (gcov_histo_index): Remove.
6519         (gcov_histogram_merge): Likewise.
6520         (compute_working_sets): Likewise.
6521         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
6522         it not obsolete.
6523         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
6524         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
6525         (GCOV_HISTOGRAM_SIZE): Remove.
6526         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
6527         (struct gcov_summary): Simplify rapidly just
6528         to runs and sum_max fields.
6529         (gcov_histo_index): Remove.
6530         (NUM_GCOV_WORKING_SETS): Likewise.
6531         (compute_working_sets): Likewise.
6532         * gcov-tool.c (print_overlap_usage_message): Remove
6533         trailing empty line.
6534         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
6535         (output_lines): Remove program related line.
6536         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
6537         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
6538         histogram.
6539         (input_profile_summary): Do not read it.
6540         (merge_profile_summaries): And do not merge it.
6541         (input_symtab): Do not call removed function.
6542         * modulo-sched.c (sms_schedule): Do not print sum_max.
6543         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
6544         removed when histogram method was invented.
6545         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
6546         mode.
6547         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
6548         GCOV coding style.
6549         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
6550         and dump selected value.
6551         * profile.c (add_working_set): Remove.
6552         (get_working_sets): Likewise.
6553         (find_working_set): Likewise.
6554         (get_exec_counts): Do not work with working sets.
6555         (read_profile_edge_counts): Do not inform as sum_max is removed.
6556         (compute_branch_probabilities): Likewise.
6557         (compute_value_histograms): Remove argument for call of
6558         get_coverage_counts.
6559         * profile.h: Do not make gcov_summary const.
6561 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
6563         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
6565 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6567         PR tree-optimization/86990
6568         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
6569         Check that the entire merged store group is made of constants only for
6570         overlapping stores.
6572 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
6574         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
6575         (VTABLE_VERIFICATION_SPEC): Likewise.
6576         (SANITIZER_EARLY_SPEC): Likewise.
6577         (SANITIZER_SPEC): Likewise.
6578         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
6579         * doc/invoke.texi (Link Options): Document -r.
6581 2018-09-20  Richard Biener <rguenther@suse.de>
6583         PR middle-end/87054
6584         * gimplify.c (gimplify_expr): Retain alignment of
6585         addressable lvalue in dereference.
6587 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
6589         PR bootstrap/87013
6590         * configure.ac: Check for .loc is_stmt support.
6591         * configure, config.in: Rebuilt.
6592         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
6593         if not supported.
6595 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6597         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
6598         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
6599         -misel=no.
6601 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6603         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6604         VECTOR_OTHER.
6605         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
6606         case VECTOR_OTHER.
6608 2018-09-20  Marek Polacek  <polacek@redhat.com>
6610         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
6612 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6614         PR tree-optimization/87288
6615         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
6616         into account when determining PEELING_FOR_NITERS.
6618 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6620         PR tree-optimization/86877
6621         * tree-vect-loop.c (vect_analyze_loop_2): Call
6622         vect_verify_datarefs_alignment.
6624 2018-09-19  Marek Polacek  <polacek@redhat.com>
6626         * doc/invoke.texi: Document -Wclass-conversion.
6628 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
6630         * config/pa/pa.c (pa_adjust_priority): Delete.
6631         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
6633         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
6634         (atomic_storehi): Likewise.
6635         (atomic_storesi): Likewise.
6636         (atomic_loaddi): Restore compare and swap exchange loop code.
6638 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
6640         PR rtl-optimization/86902
6641         * combine.c (try_combine): When changing the CC mode used, don't change
6642         an unrelated mode in other_insn to that new CC mode.
6644 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
6646         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
6647         with %T in place of calls to dump_generic_expr.
6648         (prune_runtime_alias_test_list): Likewise.
6649         (create_runtime_alias_checks): Likewise.
6650         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
6651         (vect_analyze_data_ref_dependence): Likewise.
6652         (vect_slp_analyze_data_ref_dependence): Likewise.
6653         (vect_record_base_alignment): Likewise.  Use %G in place of call
6654         to dump_gimple_stmt.
6655         (vect_compute_data_ref_alignment): Likewise.
6656         (verify_data_ref_alignment): Likewise.
6657         (vect_find_same_alignment_drs): Likewise.
6658         (vect_analyze_group_access_1): Likewise.
6659         (vect_analyze_data_ref_accesses): Likewise.
6660         (dependence_distance_ge_vf): Likewise.
6661         (dump_lower_bound): Likewise.
6662         (vect_prune_runtime_alias_test_list): Likewise.
6663         (vect_find_stmt_data_reference): Likewise.
6664         (vect_analyze_data_refs): Likewise.
6665         (vect_create_addr_base_for_vector_ref): Likewise.
6666         (vect_create_data_ref_ptr): Likewise.
6667         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
6668         (vect_can_advance_ivs_p): Likewise.
6669         (vect_update_ivs_after_vectorizer): Likewise.
6670         (vect_gen_prolog_loop_niters): Likewise.
6671         (vect_prepare_for_masked_peels): Likewise.
6672         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6673         (vect_determine_vectorization_factor): Likewise.
6674         (vect_is_simple_iv_evolution): Likewise.
6675         (vect_analyze_scalar_cycles_1): Likewise.
6676         (vect_analyze_loop_operations): Likewise.
6677         (report_vect_op): Likewise.
6678         (vect_is_slp_reduction): Likewise.
6679         (check_reduction_path): Likewise.
6680         (vect_is_simple_reduction): Likewise.
6681         (vect_create_epilog_for_reduction): Likewise.
6682         (vect_finalize_reduction:): Likewise.
6683         (vectorizable_induction): Likewise.
6684         (vect_transform_loop_stmt): Likewise.
6685         (vect_transform_loop): Likewise.
6686         (optimize_mask_stores): Likewise.
6687         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
6688         (vect_split_statement): Likewise.
6689         (vect_recog_over_widening_pattern): Likewise.
6690         (vect_recog_average_pattern): Likewise.
6691         (vect_determine_min_output_precision_1): Likewise.
6692         (vect_determine_precisions_from_range): Likewise.
6693         (vect_determine_precisions_from_users): Likewise.
6694         (vect_mark_pattern_stmts): Likewise.
6695         (vect_pattern_recog_1): Likewise.
6696         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6697         (vect_record_max_nunits): Likewise.
6698         (vect_build_slp_tree_1): Likewise.
6699         (vect_build_slp_tree_2): Likewise.
6700         (vect_print_slp_tree): Likewise.
6701         (vect_analyze_slp_instance): Likewise.
6702         (vect_detect_hybrid_slp_stmts): Likewise.
6703         (vect_detect_hybrid_slp_1): Likewise.
6704         (vect_slp_analyze_operations): Likewise.
6705         (vect_slp_analyze_bb_1): Likewise.
6706         (vect_transform_slp_perm_load): Likewise.
6707         (vect_schedule_slp_instance): Likewise.
6708         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6709         (vect_mark_stmts_to_be_vectorized): Likewise.
6710         (vect_init_vector_1): Likewise.
6711         (vect_get_vec_def_for_operand): Likewise.
6712         (vect_finish_stmt_generation_1): Likewise.
6713         (vect_check_load_store_mask): Likewise.
6714         (vectorizable_call): Likewise.
6715         (vectorizable_conversion): Likewise.
6716         (vectorizable_operation): Likewise.
6717         (vectorizable_load): Likewise.
6718         (vect_analyze_stmt): Likewise.
6719         (vect_is_simple_use): Likewise.
6720         (vect_get_vector_types_for_stmt): Likewise.
6721         (vect_get_mask_type_for_stmt): Likewise.
6722         * tree-vectorizer.c (increase_alignment): Likewise.
6724 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
6726         * doc/rtl.texi: Adjust vec_select description.
6727         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
6728         non-constant selectors.
6730 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
6732         * config/aarch64/aarch64-protos.h
6733         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
6734         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
6735         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
6736         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
6737         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
6738         (AARCH64_FL_PROFILE): Move index so flags are ordered.
6739         (AARCH64_ISA_RCPC8_4): New flag.
6740         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
6741         to aarch64_offset_9bit_signed_unscaled_p.
6742         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
6743         and use stlur.
6744         * config/aarch64/constraints.md (Ust): New constraint.
6745         * config/aarch64/predicates.md.
6746         (aarch64_9bit_offset_memory_operand): New predicate.
6747         (aarch64_rcpc_memory_operand): New predicate.
6749 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
6751         PR rtl-optimization/87361
6752         * rtlanal.c (nonzero_bits1): Revert accidental change.
6754 2018-09-19  Richard Biener  <rguenther@suse.de>
6756         PR tree-optimization/87349
6757         PR tree-optimization/87342
6758         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
6760 2018-09-18  Marek Polacek  <polacek@redhat.com>
6762         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
6763         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
6765 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6767         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
6769 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6771         PR rtl-optimization/86882
6772         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
6774 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
6776         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
6777         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
6779 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
6781         PR other/87353
6782         * doc/invoke.texi (Link Options): Fix formatting and grammar.
6784 2018-09-18  Richard Biener  <rguenther@suse.de>
6786         PR middle-end/63155
6787         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
6788         (compute_samebase_partition_bases): Likewise.
6789         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
6790         (gimple_can_coalesce_p): Simplify.
6792 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
6794         Handle a library implementation of ffs calling __builtin_ffs.
6795         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
6796         (mmix_init_libfuncs): New function: make __builtin_ffs expand
6797         to __ffsdi2.
6799 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6801         * diagnostic-show-locus.c (class layout_range): Add field
6802         "m_original_idx".
6803         (layout_range::layout_range): Add "original_idx" param and use it
6804         to initialize new field.
6805         (make_range): Use 0 for original_idx.
6806         (layout::layout): Pass in index to calls to
6807         maybe_add_location_range.
6808         (layout::maybe_add_location_range): Add param "original_idx" and
6809         pass it on to layout_range.
6810         (layout::print_any_labels): Pass on range->m_original_idx to
6811         get_text call.
6812         (gcc_rich_location::add_location_if_nearby): Use 0 for
6813         original_idx.
6814         * gcc-rich-location.h (text_range_label::get_text): Update for new
6815         param.
6816         (range_label_for_type_mismatch::get_text): Likewise.
6818 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6820         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
6822 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6824         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
6825         format_string_diagnostic_t.
6826         (fmtwarn_n): Likewise.
6827         * substring-locations.c
6828         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
6829         (format_warning_n_va): Convert to...
6830         (format_string_diagnostic_t::emit_warning_n_va): ...this.
6831         (format_warning_va): Convert to...
6832         (format_string_diagnostic_t::emit_warning_va): ...this.
6833         (format_warning_at_substring): Convert to...
6834         (format_string_diagnostic_t::emit_warning): ...this.
6835         (format_warning_at_substring_n): Convert to...
6836         (format_string_diagnostic_t::emit_warning_n): ...this.
6837         * substring-locations.h (class format_string_diagnostic_t): New
6838         class.
6839         (format_warning_va): Convert to
6840         format_string_diagnostic_t::emit_warning_va.
6841         (format_warning_n_va): Convert to
6842         format_string_diagnostic_t::emit_warning_n_va.
6843         (format_warning_at_substring): Convert to
6844         format_string_diagnostic_t::emit_warning.
6845         (format_warning_at_substring_n): Convert to
6846         format_string_diagnostic_t::emit_warning_n.
6848 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
6849             Bernd Schmidt <bernds_cb1@t-online.de>
6851         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
6852         SImode args.
6854 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6856         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
6857         operand 0 predicate to nonimmediate operand.
6858         (rint<mode>2_frndint): Remove insn pattern.
6859         (rint<mode>2): Change operand 1 predicate to general_operand.
6860         Extend operand 1 to XFmode and generate rintxf2 insn.
6861         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
6862         Do not use X87MODEF mode macro.
6863         (frndintxf2_<rounding>_i387): Rename from
6864         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
6865         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
6866         to XFmode and generate significandxf3 insn.
6868 2018-09-17  Richard Biener  <rguenther@suse.de>
6870         PR tree-optimization/87328
6871         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
6872         visiting unexecutable backedges when not iterating.
6873         (do_rpo_vn): Mark all edges not executable even when not
6874         iterating.
6876 2018-09-17  Martin Jambor  <mjambor@suse.cz>
6878         PR c/63886
6879         * doc/invoke.texi (Warning Options): Likewise.
6881 2018-09-17  Richard Biener  <rguenther@suse.de>
6883         PR tree-optimization/87301
6884         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
6885         clean EH info from leftover copy assignments.
6887 2018-09-17  Martin Liska  <mliska@suse.cz>
6889         PR gcov-profile/85871
6890         * gcov.c (output_intermediate_file): Fix out of bounds
6891         access.
6893 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
6895         * config/arc/arc.c: Object attributes for core4 not reflected
6896         correctly.
6897         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
6898         core3).
6900 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
6902         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
6904 2018-09-17  Martin Liska  <mliska@suse.cz>
6906         * doc/gcov.texi: Document new option --use-hotness-colors.
6907         * gcov.c (struct source_info): Declare new field.
6908         (source_info::source_info): Set default for maximum_count.
6909         (print_usage): Add new -q option.
6910         (process_args): Process it.
6911         (accumulate_line_info): Save src->maximum_count.
6912         (output_line_beginning): Make color line number if
6913         flag_use_hotness_colors is set.
6914         (output_line_details): Pass default argument value.
6915         (output_lines): Pass src->maximum_count.
6917 2018-09-17  Martin Liska  <mliska@suse.cz>
6919         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
6920         Use processor_names table.
6921         * config/i386/i386.c (ix86_default_align): Use
6922         processor_cost_table for alignment values.
6923         (ix86_option_override_internal): Use processor_names.
6924         (ix86_function_specific_print): Likewise.
6925         * config/i386/i386.h (struct processor_costs):
6926         Add alignment values.
6927         (struct ptt): Remove and replace with const char *.
6928         * config/i386/x86-tune-costs.h (struct processor_costs):
6929         Declare default alignments for all costs.
6931 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
6933         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
6934         symbolics or VR_VARYING ranges for ABS_EXPR.
6935         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
6936         when range will wrap.
6938 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
6940         PR middle-end/86864
6941         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
6942         before and after a JUMP_TABLE_DATA.
6944 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
6946         PR middle-end/87188
6947         * dojump.c (do_compare_and_jump): Canonicalize function pointers
6948         when one operand is a function pointer.  Use POINTER_TYPE_P and
6949         FUNC_OR_METHOD_TYPE_P.
6950         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
6951         * fold-const.c (build_range_check): Likewise.
6952         * match.pd (simple_comparison): Likewise.
6954 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
6956         PR c/82967
6957         * spellcheck.c (get_edit_distance_cutoff): New function.
6958         (selftest::test_edit_distance_unit_test_oneway): Rename to...
6959         (selftest::test_get_edit_distance_one_way): ...this.
6960         (selftest::test_get_edit_distance_unit): Rename to...
6961         (selftest::test_get_edit_distance_both_ways): ...this.
6962         (selftest::test_edit_distances): Move tests to this new function,
6963         and test some more pairs of strings.  Update for above renaming.
6964         (selftest::get_old_cutoff): New function.
6965         (selftest::test_get_edit_distance_cutoff): New function.
6966         (selftest::assert_suggested_for): New function.
6967         (ASSERT_SUGGESTED_FOR): New macro.
6968         (selftest::assert_not_suggested_for): New function.
6969         (ASSERT_NOT_SUGGESTED_FOR): New macro.
6970         (selftest::test_suggestions): New function.
6971         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
6972         tests to selftest::test_edit_distances and call it.  Add calls to
6973         selftest::test_get_edit_distance_cutoff and
6974         selftest::test_suggestions.
6975         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
6976         (best_match::consider): Replace hard-coded cutoff calculation with
6977         a call to...
6978         (best_match::get_cutoff): New declaration.
6979         (best_match::get_best_meaningful_candidate): Likewise.
6981 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6983         * builtins.c (fold_builtin_strlen): Remove TODO comment.
6985 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6987         revert:
6988         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6990         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
6991         terminated string literal.
6993 2018-09-14  Martin Sebor  <msebor@redhat.com>
6995         * builtins.c (unterminated_array): Handle ARRAY_REF.
6996         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
6997         * builtins.h (unterminated_array): Declare extern.
6998         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
6999         arrays.
7000         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
7001         calls.
7003 2018-09-14  Martin Sebor  <msebor@redhat.com>
7004             Jeff Law  <law@redhat.com>
7006         * builtins.c (unterminated_array): New.
7007         (expand_builtin_strcpy): Adjust.
7008         (expand_builtin_strcpy_args): Detect unterminated arrays.
7009         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
7010         unterminated arrays.
7011         * gimple-fold.h (get_maxval_strlen): Add argument.
7012         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
7014         * gimple-fold.c (get_range_strlen): Add argument.
7015         (get_maxval_strlen): Adjust.
7016         * gimple-fold.h (get_range_strlen): Add argument.
7018 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
7020         * config/i386/movdirintrin.h: Fix copyright year.
7022 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
7024         * reg-stack.c: Include regs.h.
7025         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
7026         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
7027         FIRST_STACK_REG, not DFmode.
7028         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
7029         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
7030         (change stack): Default register mode to the reg_raw_mode of
7031         FIRST_STACK_REG, not DFmode.
7032         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
7033         (*swapxf): Rename from swapxf.
7035 2018-09-14  Carl Love  <cel@us.ibm.com>
7037         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
7038         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
7040 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
7042         PR target/87224
7043         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
7044         alternatives.
7046 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
7048         PR target/85628
7049         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
7051 2018-09-14  Jason Merrill  <jason@redhat.com>
7053         Fix --enable-gather-detailed-mem-stats.
7054         * hash-table.c (hash_table_usage): Change from variable to function.
7055         * hash-table.h: Adjust.
7056         * Makefile.in: Add missing dependencies on hash-table.h.
7058 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7060         PR tree-optimization/87259
7061         PR lto/87283
7062         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
7063         execute_cse_reciprocals_1 has tried transforming.
7065 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
7067         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
7068         VR_VARYING for PLUS/MINUS_EXPR.
7070 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
7072         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
7073         formatting.
7075 2018-09-14  Richard Biener  <rguenther@suse.de>
7077         PR middle-end/63155
7078         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
7079         bits for the merged partition.
7081 2018-09-13  Martin Sebor  <msebor@redhat.com>
7082             Bernd Edlinger  <bernd.edlinger@hotmail.de>
7084         * builtins.h (c_srlen): Add argument.
7085         * builtins.c (warn_string_no_nul): New function.
7086         (c_strlen): Add argument and use it.  Update recursive calls.
7087         Pass DECL argument to string_constant to get info on non
7088         terminated strings.  Update *NONSTR as needed.
7089         (fold_builtin_strlen): Add argument to calls to c_strlen.
7090         Warn for unterminated arrays.
7091         (warn_string_no_null): Add prototype.
7092         * expr.c (string_constant): Update arguments.  Update recursive
7093         calls appropriately.  Detect missing NUL terminator and outermost
7094         declaration its missing in.
7095         Improve checks for arrays with nonzero lower bound or elements
7096         that are not a single byte.  Simplify offset computation.
7097         Simplify checks for non-NUL terminated strings.
7098         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
7099         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
7101 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7103         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
7104         correctly.
7105         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
7106         argument.  Fix range checks.
7107         * fold-const.h (c_getstr): Adjust protoype.
7108         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
7109         string is constant but contains no NUL byte.
7111         * expr.c (string_constant): Adjust function comment.
7112         Remove bogus check for zero termination.
7114         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
7116         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
7117         (get_constant_size): Don't make STRING_CSTs larger than they are.
7118         (check_string_literal): New check function for STRING_CSTs.
7119         (output_constant): Use it.
7121 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
7123         PR target/86812
7124         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
7126 2018-09-13  Richard Biener  <rguenther@suse.de>
7128         PR tree-optimization/87263
7129         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
7130         (struct unwind_state): Add max_rpo field.
7131         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
7132         Compute max_rpo, the max RPO number a block can be backwards reached
7133         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
7134         separating it from the iterating mode.
7136 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
7138         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
7139         (rfs_decision): New scheduling decision.
7141 2018-09-13  Richard Biener  <rguenther@suse.de>
7143         PR bootstrap/87134
7144         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
7145         (vn_nary_op_insert_pieces_predicated): Do not write useless
7146         valid_dominated_by_p entry outside of the allocated storage.
7148 2018-09-13  Omar Sandoval  <osandov@osandov.com>
7149             Tom de Vries  <tdevries@suse.de>
7151         PR debug/86985
7152         * dwarf2out.c (is_c): New function.
7153         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
7155 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
7157         PR target/85628
7158         * config/aarch64/aarch64.md (*aarch64_bfxil):
7159         Define.
7160         * config/aarch64/constraints.md (Ulc): Define.
7161         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
7162         Define.
7163         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
7164         New function.
7166 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
7168         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
7169         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
7170         aarch64_layout_frame call.
7171         (aarch64_expand_epilogue): Likewise.
7172         (aarch64_initial_elimination_offset): Likewise.
7173         (aarch64_get_separate_components): Likewise.
7174         (aarch64_use_return_insn_p): Likewise.
7175         (aarch64_layout_frame): Remove unneeded check.
7177 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7179         * configure.ac: Only append
7180         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
7181         gcc_config_arguments if it was never reconfigured or last reconfigure
7182         was with different arguments.
7183         * configure: Regenerated.
7185 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7186             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7188         PR middle-end/87290
7189         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
7190         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
7192 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7194         PR tree-optimization/87287
7195         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
7196         X % C == 0 to X % (unsigned) C == 0 optimization to ...
7197         * match.pd (X % C == 0): ... here.  New optimization.
7199 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7201         PR middle-end/82853
7202         * expr.h (maybe_optimize_mod_cmp): Declare.
7203         * expr.c (mod_inv): New function.
7204         (maybe_optimize_mod_cmp): New function.
7205         (do_store_flag): Use it.
7206         * cfgexpand.c (expand_gimple_cond): Likewise.
7208 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
7209             Julian Brown  <julian@codesourcery.com>
7211         PR middle-end/86336
7212         * gimplify.c (gimplify_scan_omp_clauses): Set
7213         target_firstprivatize_array_bases in OpenACC parallel and kernels
7214         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
7215         OpenACC data regions.
7217 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
7219         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
7220         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
7222 2018-09-12  Richard Biener  <rguenther@suse.de>
7224         PR tree-optimization/87280
7225         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
7226         edge but unreachable target.
7227         (do_rpo_vn): For conservatively handling a PHI only mark
7228         the backedge executable but not the block reachable.
7230 2018-09-12  Richard Biener  <rguenther@suse.de>
7232         PR tree-optimization/87266
7233         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
7234         visited blocks.
7236 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
7238         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
7239         constants.
7240         ("trunc<BFP:mode><DFP_ALL:mode>2")
7241         ("trunc<DFP_ALL:mode><BFP:mode>2")
7242         ("extend<BFP:mode><DFP_ALL:mode>2")
7243         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
7244         according to the target operand type.
7246 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7247             Andreas Krebbel  <krebbel@linux.ibm.com>
7249         PR tree-optimization/86844
7250         * gimple-ssa-store-merging.c
7251         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
7252         there are any overlapping stores in between them, make sure they are
7253         also coalesced or we give up completely.
7255 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7257         PR middle-end/87248
7258         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
7259         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
7261 2018-09-12  Tom de Vries  <tdevries@suse.de>
7263         * common.opt (gdescribe-dies): Add option.
7264         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
7265         attribute for artifical and nameless decls.
7266         (dwarf2out_register_external_die): Add description attribute to
7267         external reference die.
7268         (add_desc_attribute): New functions.
7269         (gen_subprogram_die): Add description attribute to
7270         DW_TAG_call_site_parameter.
7271         * tree-pretty-print.c (print_generic_expr_to_str): New function.
7272         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
7273         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
7274         -gno-describe-dies.
7275         (@item -gdescribe-dies): Add.
7277 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
7279         * tree-vrp.c (vrp_shift_undefined_p): Remove.
7280         (extract_range_from_binary_expr_1: Call
7281         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
7282         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
7283         depend on sign.
7285 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
7287         * gimple-ssa-warn-alloca.c
7288         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
7289         field for ALLOCA_BOUND_*_LARGE.
7291 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7293         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
7295 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7297         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
7298         for clobbers.  Remove obsolete comment.
7300 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7302         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
7303         mpxchk, mpxld and mpxst types.
7304         (define_attr length_immediate): Remove all processing of mpx types.
7305         (define_attr prefix_0f): Ditto.
7306         (define_attr memory): Ditto.
7308 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7310         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
7311         (log<mode>2): Change operand 1 predicate to general_operand.
7312         Extend operand 1 to XFmode and generate logxf3 insn.
7313         (log10<mode>2): Change operand 1 predicate to general_operand.
7314         Extend operand 1 to XFmode and generate log10xf3 insn.
7315         (log2<mode>2): Change operand 1 predicate to general_operand.
7316         Extend operand 1 to XFmode and generate log2xf3 insn.
7317         (fyl2xp1_extend<mode>xf3_i387): Remove.
7318         (log1p<mode>2): Change operand 1 predicate to general_operand.
7319         Extend operand 1 to XFmode and generate log1pxf3 insn.
7320         (fxtract_extend<mode>xf3_i387): Remove.
7321         (logb<mode>2): Change operand 1 predicate to general_operand.
7322         Extend operand 1 to XFmode and generate logbxf3 insn.
7323         (ilogb<mode>2): Change operand 1 predicate to general_operand.
7324         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
7325         (significand<mode>2): Change operand 1 predicate to general_operand.
7326         Extend operand 1 to XFmode and generate significandxf3 insn.
7328 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7330         * gcc.c (perror_with_name, pfatal_with_name): Delete.
7331         (load_specs): Use fatal_error.
7332         (DELETE_IF_ORDINARY, process_command): Use error.
7333         (execute, run_attempt): Use fatal_error.
7335 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
7337         * diagnostic-core.h (sorry_at): New prototype.
7338         * diagnostic.c (sorry_at): New function.
7340 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
7342         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
7343         by zero as VR_UNDEFINED.
7345 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
7347         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
7348         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
7349         (<sincos>mode2): New expander.
7350         (sincos_extend<mode>xf3_i387): Remove insn pattern.
7351         (sincos -> sin, cos splitters): Remove splitter patterns.
7352         (sincos<mode>3): Change operand 2 predicate to general_operand.
7353         Extend operand 2 to XFmode and generate sincosxf3 insn.
7354         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
7355         Change operand 3 predicate to const1_operand.
7356         (fptan_extend<mode>xf4_i387): Remove insn pattern.
7357         (tanxf2): Update operands in the call to fptanxf4_i387.
7358         (tan<mode>2): Change operand 1 predicate to general_operand.
7359         Extend operand 1 to XFmode and generate tanxf3 insn.
7360         (atan2xf3): Rename from *fpatanxf3_i387.
7361         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
7362         (atan2xf3): Remove expander.
7363         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
7364         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
7365         (atan<mode>2): Change operand 1 predicate to general_operand.
7366         Extend operand 1 to XFmode and generate atanxf3 insn.
7368 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
7370         * config/i386/i386.md (x87/SSE constant load splitter): Use
7371         memory_operand instead of nonimmediate_operand for input operand
7372         predicate.
7374 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
7376         * config/i386/i386.md (float partial SSE register stall splitter): Move
7377         splitter near its instruction pattern.
7378         (float_extend partial SSE register stall splitter): Ditto.
7379         (float_truncate partial SSE register stall splitter): Ditto.
7381 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
7383         PR target/86794
7384         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7385         to speculation_safe_value_not_needed.
7387         PR target/85666
7388         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
7389         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
7390         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
7391         leaf_function_p, instead use has_hard_reg_initial_val.
7393 2018-09-09  Nathan Sidwell  <nathan@acm.org>
7395         * gcc.h (pfatal_with_name): Don't declare here.
7396         * gcc.c (pfatal_with_name): Make static.
7398 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
7400         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
7401         earlyclobber.
7403 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
7405         PR rtl-optimization/85458
7406         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
7407         priority hook to reduce the priority of EXPR.
7409 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
7411         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
7412         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
7413         with X87_ENABLE_FLOAT.  Remove preparation code.
7414         (*float<SWI48:mode><MODEF:mode>2): Rename from
7415         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
7416         math using "enabled" attribute.
7417         (*floatdi<MODEF:mode>2_i387): Rename from
7418         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
7419         enable for 32bit targets only.
7420         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
7421         splitter.
7422         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
7423         as operand 1 predicate.  Rewrite as define_insn_and_split.
7424         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
7426 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7428         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
7429         to fallthru to FLOAT case.
7431 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7433         PR target/86731
7434         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
7435         around folding of vec_sl to handle out of range shift values.
7437 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7439         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
7440         Update callers to gen_fix_trunc<mode>_i387_fisttp
7441         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
7442         nonimmediate_operand.
7443         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
7444         and corresponding splitters.
7445         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
7446         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
7447         (fix_truncdi_i387_with_temp): Remove insn pattern
7448         and corresponding splitters.
7449         (fix_trunc<mode>_i387): Change operand 0 predicate to
7450         nonimmediate_operand.
7451         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
7452         and corresponding splitters.
7453         (*fistdi2_1): Remove.
7454         (fistdi2): Ditto.
7455         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
7456         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
7457         (*fist<mode>2_1): Remove.
7458         (fist<mode>2): Ditto.
7459         (fist<mode>2_with_temp): Remove insn pattern and corresponding
7460         splitters.
7461         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
7462         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
7463         (fistdi2_<rounding>): Change operand 0 predicate to
7464         nonimmediate_operand.
7465         (fistdi2_<rounding>_with_temp): Remove insn pattern
7466         and corresponding splitters.
7467         (fist<mode>2_<rounding>): Change operand 0 predicate to
7468         nonimmediate_operand.
7469         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
7470         and corresponding splitters.
7472         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
7474 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7476         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
7477         the init value.
7479 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7481         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7482         early gimple folding of vec_splat().
7483         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
7484         * gimple-fold.h: Add an extern define for tree_vec_extract().
7486 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7488         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
7489         wrappers around TREE_TYPE comparisons.
7491 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7493         PR target/80080
7494         * config/s390/predicates.md: Add nonsym_memory_operand.
7495         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
7496         contains a SYMBOL_REF, load it into an intermediate pseudo.
7497         (s390_emit_compare_and_swap): Legitimize operand.
7498         * config/s390/s390.md: Use the new nonsym_memory_operand
7499         with UNSPECV_CAS patterns.
7501 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7503         PR target/80080
7504         * config/s390/s390-passes.def: New file.
7505         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
7506         declaration.
7507         (make_pass_s390_early_mach): Add declaration.
7508         * config/s390/s390.c (make_pass_s390_early_mach):
7509         (s390_option_override): Remove dynamic registration.
7510         * config/s390/t-s390: Add s390-passes.def.
7512 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7514         * config/s390/s390.c (s390_decompose_constant_pool_ref):
7515         Remove UNSPEC_LTREL_BASE check.
7516         (annotate_constant_pool_refs): Likewise.
7517         (find_constant_pool_ref): Likewise.
7518         (find_ltrel_base): Removed.
7519         (replace_ltrel_base): Removed.
7520         (s390_mainpool_finish): Remove replace_ltrel_base call.
7521         (s390_chunkify_start): Remove pending LTREL_BASE logic.
7522         (s390_chunkify_finish): Remove replace_ltrel_base call.
7523         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
7525 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
7527         PR target/86779
7528         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7529         to speculation_safe_value_not_needed.
7531 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
7532             Bernd Schmidt  <bernds_cb1@t-online.de>
7534         * config/nvptx/nvptx-opts.h: New file.
7535         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
7536         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
7537         (ASM_SPEC): Define.
7538         (TARGET_SM35): New macro.
7539         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
7540         correct predicate.
7541         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
7542         values.
7543         (misa=): New option.
7544         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
7546 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
7548         * config/i386/i386.md (truncdfsf2): Remove expander.
7549         (truncdfsf2_with_temp): Ditto.
7550         (truncxf<mode>2): Ditto.
7551         (*truncdfsf_fast_mixed): Remove insn pattern.
7552         (*truncdfsf_fast_i387): Ditto.
7553         (*truncdfsf_mixed): Ditto.
7554         (*truncdfsf_i387): Ditto.
7555         (*truncdfsf2_i387_1): Ditto.
7556         (*truncxfsf2_mixed): Ditto.
7557         (*truncxfdf2_mixed): Ditto.
7558         (*truncxf<mode>2_i387_noop): Ditto. Update callers
7559         to call gen_truncxf<mode>2 instead.
7560         (*truncxf<mode>2_i387): Remove.
7561         (reg->reg splitters): Remove splitter pattern.
7562         (reg->mem splitters): Ditto.
7564         (truncdfsf2): New insn pattern.
7565         (truncxf<mode>2): Ditto.
7567 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7569         * tree-ssa-math-opts.c (is_mult_by): New function.
7570         (is_square_of): Use the above.
7571         (optimize_recip_sqrt): New function.
7572         (pass_cse_reciprocals::execute): Use the above.
7574 2018-09-05  Richard Biener  <rguenther@suse.de>
7576         PR bootstrap/87134
7577         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
7578         to zero-init the emplaced vec.
7580 2018-09-05  Martin Liska  <mliska@suse.cz>
7582         PR tree-optimization/87205
7583         * tree-switch-conversion.c (pass_lower_switch::execute):
7584         Group cases for switch statements.
7586 2018-09-05  Richard Biener  <rguenther@suse.de>
7588         PR tree-optimization/87217
7589         * tree-ssa-sccvn.c (vuse_valueize): New.
7590         (vn_reference_lookup_pieces): Use it.
7591         (vn_reference_lookup): Likewise.
7593 2018-09-05  Nathan Sidwell  <nathan@acm.org>
7595         PR c++/87137
7596         * stor-layout.c (place_field): Scan forwards to check last
7597         bitfield when ms_bitfield_placement is in effect.
7599 2018-09-05  Richard Biener  <rguenther@suse.de>
7601         PR bootstrap/87225
7602         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
7603         return.
7605 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7606             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
7608         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
7609         * config.gcc (extra_objs): Build it.
7610         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
7611         Likewise.
7612         * config/aarch64/aarch64-passes.def
7613         (pass_tag_collision_avoidance): New pass.
7614         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
7615         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
7616         (aarch64_classify_address): Remove static qualifier.
7617         (aarch64_address_info, aarch64_address_type): Move to...
7618         * config/aarch64/aarch64-protos.h: ... here.
7619         (make_pass_tag_collision_avoidance): New function.
7620         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
7621         New tuning flag.
7623 2018-09-05  Martin Liska  <mliska@suse.cz>
7625         * doc/gcov.texi: Update documentation of humar
7626         readable mode.
7627         * gcov.c (format_count): Print one decimal place, it provides
7628         more fine number of situations like '1G' vs. '1.4G'.
7630 2018-09-05  Martin Liska  <mliska@suse.cz>
7632         PR target/87164
7633         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
7634         * optc-gen.awk: Allow 'Var' for Deprecated options in order
7635         to generate a MASK value.
7637 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
7639         PR debug/86593
7640         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
7641         if frame pointer isn't used.
7642         (compute_frame_pointer_to_fb_displacement): Likewise.
7644 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
7646         PR target/87198
7647         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
7648         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
7649         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
7650         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
7651         and OPTION_MASK_ISA_XSAVEC_UNSET.
7653 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
7655         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
7656         XOR operations in NAND case.
7658 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7660         * wide-int-range.cc (wide_int_range_convert): New.
7661         * wide-int-range.h (wide_int_range_convert): New.
7662         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
7663         code into wide_int_range_convert.
7664         (extract_range_into_wide_ints): Do not munge anti range constants
7665         into the entire domain.  Just return the range back.
7667 2018-09-04  Martin Liska  <mliska@suse.cz>
7669         * genmatch.c (output_line_directive): Add new argument
7670         fnargs.
7671         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
7673 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7675         * doc/invoke.texi (Option Summary): Add whitespace.
7677         * doc/invoke.texi (Option Summary): Add -Waligned-new.
7679 2018-09-04  Richard Biener  <rguenther@suse.de>
7681         PR tree-optimization/87211
7682         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
7683         backedge value we're supposed to treat as VARYING also number
7684         the PHI to VARYING in case it got a different value-number already.
7686 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7688         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
7689         (extract_range_from_binary_expr_1): Do not call
7690         vrp_can_optimize_bit_op.
7691         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
7692         static.
7693         (wide_int_range_get_mask_and_bounds): New.
7694         (wide_int_range_optimize_bit_op): New.
7695         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
7696         (wide_int_range_bit_and): Same.
7697         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
7698         (wide_int_range_optimize_bit_op): New.
7699         (wide_int_range_get_mask_and_bounds): New.
7701 2018-09-04  Richard Biener  <rguenther@suse.de>
7703         PR tree-optimization/87176
7704         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
7705         variable.  When value-numbering a virtual PHI node make sure
7706         to not value-number to the backedge value.
7708 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7710         * doc/extend.texi (Long Long, Hex Floats): Document support for
7711         long long and hex floats in more recent versions of ISO C++.
7713 2018-09-03  Richard Biener  <rguenther@suse.de>
7715         PR tree-optimization/87177
7716         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
7717         cleanup.
7719 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7721         * bb-reorder.c (edge_order): Convert to C-qsort-style
7722         tri-state comparator.
7723         (reorder_basic_blocks_simple): Change std::stable_sort to
7724         gcc_stablesort.
7726 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7728         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
7729         tri-state comparator.
7730         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
7732 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7734         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
7735         (mergesort): ... here as maximum count for using netsort.
7736         (gcc_qsort): Set nlim to 3 if stable sort is requested.
7737         (gcc_stablesort): New.
7738         * system.h (gcc_stablesort): Declare.
7740 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7742         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
7743         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
7744         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
7746 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
7748         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
7749         lxsdx and stxsdx alternatives.
7750         (*mov<mode>_hardfloat64): Ditto.
7751         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
7753 2018-09-03  Richard Biener  <rguenther@suse.de>
7755         PR tree-optimization/87200
7756         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
7757         simplify result.
7759 2018-09-03  Martin Liska  <mliska@suse.cz>
7761         PR tree-optimization/87201
7762         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7763         Fix parenthesis in an expression.
7765 2018-09-03  Richard Biener  <rguenther@suse.de>
7767         PR tree-optimization/87197
7768         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
7769         visited.  CSE the VN_INFO hashtable lookup.
7771         PR tree-optimization/87169
7772         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
7773         iterating make sure there's no extra backedges from irreducible
7774         regions feeding the header.  Mark the destination block
7775         executable.
7777 2018-09-03  Martin Liska  <mliska@suse.cz>
7779         PR driver/83193
7780         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
7781         * common/common-targhooks.c (default_get_valid_option_values):
7782         New function.
7783         * common/common-targhooks.h (default_get_valid_option_values):
7784         Likewise.
7785         * common/config/i386/i386-common.c: Move processor_target_table
7786         from i386.c.
7787         (ix86_get_valid_option_values): New function.
7788         (TARGET_GET_VALID_OPTION_VALUES): New macro.
7789         * config/i386/i386.c (struct ptt): Move to i386-common.c.
7790         (PTA_*): Move all defined masks into i386-common.c.
7791         (ix86_function_specific_restore): Use new processor_cost_table.
7792         * config/i386/i386.h (struct ptt): Moved from i386.c.
7793         (struct pta): Likewise.
7794         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
7795         * doc/tm.texi.in: Likewise.
7796         * opt-suggestions.c (option_proposer::suggest_option):
7797         Pass prefix to build_option_suggestions.
7798         (option_proposer::get_completions): Likewise.
7799         (option_proposer::build_option_suggestions): Use the new target
7800         hook.
7801         * opts.c (struct option_help_tuple): New struct.
7802         (print_filtered_help): Use the new target hook.
7804 2018-09-03  Martin Liska  <mliska@suse.cz>
7806         PR middle-end/59521
7807         * predict.c (set_even_probabilities): Add likely_edges
7808         argument and handle cases where we have precisely one
7809         likely edge.
7810         (combine_predictions_for_bb): Catch also likely_edges.
7811         (tree_predict_by_opcode): Handle gswitch statements.
7812         * tree-cfg.h (find_case_label_for_value): New declaration.
7813         (find_taken_edge_switch_expr): Likewise.
7814         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7815         Find pivot in decision tree based on probabily, not by number of
7816         nodes.
7818 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
7820         * doc/standards.texi (Standards): Update Objective-C reference.
7822 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7824         * doc/install.texi (Prerequisites): Update link for MPC.
7826 2018-09-01  Michael Matz  <matz@suse.de>
7828         PR tree-optimization/87074
7829         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
7830         PHIs for outer-loop uses.
7832 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7834         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
7835         * doc/invoke.texi (C Dialect Options): Ditto.
7837 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7839         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
7841 2018-08-31  Richard Biener  <rguenther@suse.de>
7843         PR tree-optimization/87168
7844         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
7845         (rpo_elim::eliminate_avail): When OP was not visited it must
7846         be available.
7848 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
7850         * tree-vrp.c (copy_value_range): Convert param "from" from
7851         "value_range *" to "const value_range *".
7852         (range_is_null): Likewise for param "vr".
7853         (range_int_cst_p): Likewise.
7854         (range_int_cst_singleton_p): Likewise.
7855         (symbolic_range_p): Likewise.
7856         (value_ranges_intersect_p): Likewise for both params.
7857         (value_range_nonnegative_p): Likewise for param "vr".
7858         (value_range_constant_singleton): Likewise.
7859         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
7860         (extract_range_into_wide_ints): Likewise for param "vr".
7861         (extract_range_from_multiplicative_op): Likewise for params "vr0"
7862         and "vr1".
7863         (vrp_can_optimize_bit_op): Likewise.
7864         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
7865         "vr1_".
7866         (extract_range_from_unary_expr): Likewise.
7867         (debug_value_range): Likewise for param "vr".
7868         (value_range::dump): Add "const" qualifier.
7869         (vrp_prop::check_array_ref): Convert local "vr" from
7870         "value_range *" to "const value_range *".
7871         (vrp_prop::check_mem_ref): Likewise.
7872         (vrp_prop::visit_stmt): Likewise for local "old_vr".
7873         (vrp_intersect_ranges_1): Likewise for param "vr_1".
7874         (vrp_intersect_ranges): Likewise.
7875         (simplify_stmt_for_jump_threading): Likewise for local "vr".
7876         (vrp_prop::vrp_finalize): Likewise.
7877         * tree-vrp.h (value_range::dump): Add "const" qualifier.
7878         (vrp_intersect_ranges): Add "const" qualifier to params as above.
7879         (extract_range_from_unary_expr): Likewise.
7880         (value_range_constant_singleton): Likewise.
7881         (symbolic_range_p): Likewise.
7882         (copy_value_range): Likewise.
7883         (extract_range_from_binary_expr_1): Likewise.
7884         (range_int_cst_p): Likewise.
7885         (vrp_set_zero_nonzero_bits): Likewise.
7886         (range_int_cst_singleton_p): Likewise.
7888 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
7890         * config/aarch64/arm_neon.h (vabsd_s64): New.
7891         (vnegd_s64): Likewise.
7893 2018-08-31  Martin Jambor  <mjambor@suse.cz>
7895         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
7897 2018-08-31  Martin Liska  <mliska@suse.cz>
7899         * ipa-icf.c (sem_item::add_type): Use
7900         sem_item::m_type_hash_cache.
7901         * ipa-icf.h: Move the cache from sem_item_optimizer
7902         to sem_item.
7904 2018-08-31  Nathan Sidwell  <nathan@acm.org>
7906         * doc/extend.texi (Backwards Compatibility): Remove implicit
7907         extern C leeway of () being (...).
7909 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7911         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
7913 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
7915         PR target/86684
7916         PR target/87149
7917         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
7919 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
7921         PR middle-end/87138
7922         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
7923         gen_int_mode.  Formatting fixes.
7925 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
7927         * target.def (custom_function_descriptors): Improve documentation.
7928         * doc/tm.texi.in (Trampolines): Expand discussion of function
7929         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
7930         beginning of the section.
7931         * doc/tm.texi: Regenerated.
7933 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
7935         * cfg.h (class auto_edge_flag): Spell out the template-id of the
7936         base class in the initializer list.  This is a workaround for
7937         building with older GCC.
7938         (class auto_bb_flag): Likewise.
7940 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
7942         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
7943         (altivec_vcmpequ<VI_char>_p): Remove star.
7944         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
7945         vector load modes.
7946         (expand_strncmp_vec_sequence): New function.
7947         (emit_final_str_compare_vec): New function.
7948         (expand_strn_compare): Add support for vector strncmp.
7949         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
7950         length specification to bytes.
7951         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
7952         (vcmpnezb_p): New pattern.
7953         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
7954         for option -mstring-compare-inline-limit.
7956 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
7958         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
7959         (PTA_SKYLAKE): Add PTA_AES.
7960         (PTA_GOLDMONT): Likewise.
7962 2018-08-29  Jan Hubicka  <jh@suse.cz>
7964         PR lto/86517
7965         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
7966         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
7968 2018-08-29  Jan Hubicka  <jh@suse.cz>
7970         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
7971         TYPE_STUB_DECL.
7972         (hash_tree): Do not visit TYPE_STUB_DECL.
7973         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
7974         stream TYPE_STUB_DECL.
7975         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
7976         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
7977         after free_lang_data.
7978         (type_in_anonymous_namespace_p): Likewise.
7980 2018-08-29  Jan Hubicka  <jh@suse.cz>
7982         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7983         comment that it has to be even number.
7984         (class sreal): Change m_sig type to int32_t.
7985         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7986         int64_t for temporary calculations.
7987         (sreal_verify_basics): Drop one bit from minimum and maximum.
7989 2018-08-30  Richard Biener  <rguenther@suse.de>
7991         PR tree-optimization/87147
7992         * tree-ssa-sccvn.c (SSA_VISITED): New function.
7993         (visit_phi): When the degenerate result is from the backedge and
7994         we didn't visit its definition yet drop to VARYING.
7995         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
7997 2018-08-29  Jan Hubicka  <jh@suse.cz>
7999         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
8000         DECL_VINDEX.
8001         (hash_tree): Likewise.
8003 2018-08-29  Jan Hubicka  <jh@suse.cz>
8005         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
8006         and TYPE_NEXT_REF_TO.
8008 2018-08-29  Jan Hubicka  <jh@suse.cz>
8010         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
8011         comment that it has to be even number.
8012         (class sreal): Change m_sig type to int32_t.
8013         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
8014         int64_t for temporary calculations.
8015         (sreal_verify_basics): Drop one bit from minimum and maximum.
8017 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
8019         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
8021 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
8023         PR middle-end/86995
8024         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
8025         if to_add is negative.
8027 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8029         PR middle-end/87053
8030         * builtins.c (c_strlen): Improve range checks.
8032 2018-08-29  Martin Sebor  <msebor@redhat.com>
8033             Jeff Law  <law@redhat.com>
8035         PR tree-optimization/86714
8036         PR tree-optimization/86711
8037         * builtins.c (c_strlen): Add arguments to call to string_constant.
8038         * expr.c (string_constant): Add argument.  Detect missing nul
8039         terminator and outermost declaration it's missing in.
8040         * expr.h (string_constant): Add argument.
8041         * fold-const.c (read_from_constant_string): Add arguments to call to
8042         string_constant.
8043         (c_getstr): Likewise.
8044         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
8045         to string_constant.
8046         * tree-ssa-strlen.c (get_stridx): Likewise.
8048 2018-08-29  Jan Hubicka  <jh@suse.cz>
8050         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
8051         Do not stream DECL_VINDEX.
8052         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
8053         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
8054         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
8056 2018-08-29  Richard Biener  <rguenther@suse.de>
8058         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
8059         virtual operands that are not default defs to honor region
8060         boundaries.
8061         (rpo_vn_valueize): Remove ineffective code here.
8063 2018-08-29  Richard Biener  <rguenther@suse.de>
8065         PR tree-optimization/87132
8066         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
8067         when skipping defs reachable over backedges.
8069 2018-08-29  Richard Biener  <rguenther@suse.de>
8071         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
8072         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
8073         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
8074         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
8075         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
8076         (vn_reference_lookup_3): Remove use of const_parms.
8077         (free_rpo_vn): Do not free const_parms.
8078         (do_rpo_vn): Do not call init_const_parms.
8079         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
8080         SSA_NAME_POINTS_TO_READONLY_MEMORY.
8081         (call_may_clobber_ref_p_1): Likewise.
8083 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
8085         PR other/86726
8086         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
8087         (-O): Ditto.
8088         (-ftree-scev-cprop): Document.
8090 2018-08-29  Jan Hubicka  <jh@suse.cz>
8092         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
8093         parameters.
8094         (sreal constructor): Update.
8095         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
8096         sreal:operator/): Update.
8098 2018-08-29  Martin Liska  <mliska@suse.cz>
8100         * tree-switch-conversion.c (switch_conversion::expand):
8101         Strenghten assumption about gswitch statements.
8103 2018-08-29  Richard Biener  <rguenther@suse.de>
8105         PR tree-optimization/87117
8106         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
8107         re-value-number released SSA VDEFs.
8109 2018-08-29  Richard Biener  <rguenther@suse.de>
8111         PR tree-optimization/87126
8112         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
8114 2018-08-28  Jim Wilson  <jimw@sifive.com>
8116         * config/riscv/pic.md: Rewrite.
8117         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
8118         invalid address.
8119         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
8120         (SOFTF, default_load, softload, softstore): New.
8122 2018-08-28  Jeff Law  <law@redhat.com>
8124         * fold-const.c (fold_binary_loc): Remove recently added assert.
8126 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
8128         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
8129         to OP parmeter of generated function.
8131 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
8133         PR tree-optimization/87009
8134         * match.pd: Add boolean optimizations.
8136 2018-08-28  Martin Sebor  <msebor@redhat.com>
8138         PR middle-end/86631
8139         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
8140         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
8141         (pass_walloca::gate): Use it.
8142         (alloca_call_type): Same.
8143         (pass_walloca::execute): Same.
8144         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
8146 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
8148         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
8149         GCC_VERSION for usage of "__gcc_dump_printf__" format from
8150         >= 3005 to >= 9000.
8152 2018-08-28  Richard Biener  <rguenther@suse.de>
8154         PR tree-optimization/87124
8155         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
8156         constants before looking up avail.
8158 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
8160         PR middle-end/87099
8161         * calls.c (maybe_warn_nonstring_arg): Punt early if
8162         warn_stringop_overflow is zero.  Don't call get_range_strlen
8163         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
8164         Swap comparison operands to have constants on rhs.  Only use
8165         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
8166         increment lenrng[0].
8168 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
8170         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
8171         use of tree_to_shwi.  Remove duplicated test for the size being
8172         a whole number of bytes.
8174 2018-08-28  Richard Biener  <rguenther@suse.de>
8176         PR tree-optimization/87117
8177         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
8178         Handle removed stmt without LHS (GIMPLE_NOP).
8180 2018-08-28  Richard Biener  <rguenther@suse.de>
8182         PR tree-optimization/87117
8183         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
8184         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
8186 2018-08-28  Richard Biener  <rguenther@suse.de>
8188         PR tree-optimization/87117
8189         * tree-ssa-pre.c (compute_avail): Do not make expressions
8190         with predicated values available.
8191         (get_expr_value_id): Assert we do not run into predicated value
8192         expressions.
8194 2018-08-28  Richard Biener  <rguenther@suse.de>
8196         PR tree-optimization/87117
8197         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
8198         get virtual operands.
8199         (get_expr_operands): Handle STRING_CST like other decls.
8201 2018-08-28  Martin Liska  <mliska@suse.cz>
8203         * tree.h: Update documentation of fndecl_built_in_p
8204         functions.
8207 2018-08-27  Jeff Law  <law@redhat.com>
8208         PR tree-optimization/87110
8209         * tree-ssa-dse.c (compute_trims): Handle non-constant
8210         TYPE_SIZE_UNIT.
8212 2018-08-27  Martin Sebor  <msebor@redhat.com>
8214         PR tree-optimization/86914
8215         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
8217 2018-08-27  Martin Sebor  <msebor@redhat.com>
8219         PR tree-optimization/87112
8220         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
8221         the type of the bound argument.
8223 2018-08-27  Jeff Law  <law@redhat.com>
8225         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
8226         type does not have a TYPE_SIZE_UNIT.
8228 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
8230         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
8231         with include of backend.h.
8233 2018-08-27  Richard Biener  <rguenther@suse.de>
8235         PR tree-optimization/86927
8236         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
8237         use const cond reduction code.
8239 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
8241         PR tree-optimization/85758
8242         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
8244 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
8246         PR c++/87091
8247         * diagnostic-show-locus.c (class layout_range): Update for
8248         conversion of show_caret_p to a tri-state.
8249         (layout_range::layout_range): Likewise.
8250         (make_range): Likewise.
8251         (layout::maybe_add_location_range): Likewise.
8252         (layout::should_print_annotation_line_p): Don't show annotation
8253         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
8254         (layout::get_state_at_point): Update for conversion of
8255         show_caret_p to a tri-state.  Bail out early for
8256         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
8257         underlining or source colorization.
8258         (gcc_rich_location::add_location_if_nearby): Update for conversion
8259         of show_caret_p to a tri-state.
8260         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
8261         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
8262         Likewise.
8263         (selftest::test_one_liner_labels): Likewise.
8264         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
8265         conversion of show_caret_p to a tri-state.
8266         * pretty-print.c (text_info::set_location): Likewise.
8267         * pretty-print.h (text_info::set_location): Likewise.
8268         * substring-locations.c (format_warning_n_va): Likewise.
8269         * tree-diagnostic.c (default_tree_printer): Likewise.
8270         * tree-pretty-print.c (newline_and_indent): Likewise.
8272 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
8274         PR c++/87091
8275         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
8276         line above for line-insertion fix-it hints.
8277         (selftest::test_fixit_insert_containing_newline): Update the
8278         expected results, and add a test with line-numbering enabled.
8280 2018-08-27  Martin Liska  <mliska@suse.cz>
8282         PR sanitizer/86962
8283         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
8284         params with DECL_HAS_VALUE_EXPR_P.
8286 2018-08-27  Martin Liska  <mliska@suse.cz>
8288         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
8289         selected expansion strategy.
8291 2018-08-27  Martin Liska  <mliska@suse.cz>
8293         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
8294         * builtins.c (is_builtin_fn): Likewise.
8295         * attribs.c (diag_attr_exclusions): Use new function
8296         fndecl_built_in_p and remove check for FUNCTION_DECL if
8297         possible.
8298         (builtin_mathfn_code): Likewise.
8299         (fold_builtin_expect): Likewise.
8300         (fold_call_expr): Likewise.
8301         (fold_builtin_call_array): Likewise.
8302         (fold_call_stmt): Likewise.
8303         (set_builtin_user_assembler_name): Likewise.
8304         (is_simple_builtin): Likewise.
8305         * calls.c (gimple_alloca_call_p): Likewise.
8306         (maybe_warn_nonstring_arg): Likewise.
8307         * cfgexpand.c (expand_call_stmt): Likewise.
8308         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
8309         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
8310         (cgraph_node::verify_node): Likewise.
8311         * cgraphclones.c (build_function_decl_skip_args): Likewise.
8312         (cgraph_node::create_clone): Likewise.
8313         * config/arm/arm.c (arm_insert_attributes): Likewise.
8314         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
8315         * dse.c (scan_insn): Likewise.
8316         * expr.c (expand_expr_real_1): Likewise.
8317         * fold-const.c (operand_equal_p): Likewise.
8318         (fold_binary_loc): Likewise.
8319         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
8320         * gimple-low.c (lower_stmt): Likewise.
8321         * gimple-pretty-print.c (dump_gimple_call): Likewise.
8322         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8323         Likewise.
8324         * gimple.c (gimple_build_call_from_tree): Likewise.
8325         (gimple_call_builtin_p): Likewise.
8326         (gimple_call_combined_fn): Likewise.
8327         * gimplify.c (gimplify_call_expr): Likewise.
8328         (gimple_boolify): Likewise.
8329         (gimplify_modify_expr): Likewise.
8330         (gimplify_addr_expr): Likewise.
8331         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
8332         * ipa-cp.c (determine_versionability): Likewise.
8333         * ipa-fnsummary.c (compute_fn_summary): Likewise.
8334         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
8335         * ipa-split.c (visit_bb): Likewise.
8336         (split_function): Likewise.
8337         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
8338         * lto-cgraph.c (input_node): Likewise.
8339         * lto-streamer-out.c (write_symbol): Likewise.
8340         * omp-low.c (setjmp_or_longjmp_p): Likewise.
8341         (lower_omp_1): Likewise.
8342         * predict.c (strip_predict_hints): Likewise.
8343         * print-tree.c (print_node): Likewise.
8344         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
8345         * trans-mem.c (is_tm_irrevocable): Likewise.
8346         (is_tm_load): Likewise.
8347         (is_tm_simple_load): Likewise.
8348         (is_tm_store): Likewise.
8349         (is_tm_simple_store): Likewise.
8350         (is_tm_abort): Likewise.
8351         (tm_region_init_1): Likewise.
8352         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
8353         * tree-cfg.c (verify_gimple_call): Likewise.
8354         (move_stmt_r): Likewise.
8355         (stmt_can_terminate_bb_p): Likewise.
8356         * tree-eh.c (lower_eh_constructs_2): Likewise.
8357         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
8358         * tree-inline.c (remap_gimple_stmt): Likewise.
8359         (copy_bb): Likewise.
8360         (estimate_num_insns): Likewise.
8361         (fold_marked_statements): Likewise.
8362         * tree-sra.c (scan_function): Likewise.
8363         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
8364         (optimize_stack_restore): Likewise.
8365         (pass_fold_builtins::execute): Likewise.
8366         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
8367         (mark_all_reaching_defs_necessary_1): Likewise.
8368         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
8369         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
8370         (pass_forwprop::execute): Likewise.
8371         * tree-ssa-loop-im.c (stmt_cost): Likewise.
8372         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
8373         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
8374         * tree-ssa-strlen.c (get_string_length): Likewise.
8375         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
8376         (find_func_aliases_for_call): Likewise.
8377         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
8378         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
8379         * tree-tailcall.c (find_tail_calls): Likewise.
8380         * tree.c (need_assembler_name_p): Likewise.
8381         (free_lang_data_in_decl): Likewise.
8382         (get_call_combined_fn): Likewise.
8383         * ubsan.c (is_ubsan_builtin_p): Likewise.
8384         * varasm.c (incorporeal_function_p): Likewise.
8385         * tree.h (DECL_BUILT_IN): Remove and replace with
8386         fndecl_built_in_p.
8387         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
8388         (fndecl_built_in_p): New.
8390 2018-08-27  Martin Liska  <mliska@suse.cz>
8392         PR tree-optimization/86847
8393         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
8394         Dump also subtree probability.
8395         (switch_decision_tree::do_jump_if_equal): New function.
8396         (switch_decision_tree::emit_case_nodes): Handle special
8397         situations in balanced tree that can be emitted much simpler.
8398         Fix calculation of probabilities that happen in tree expansion.
8399         * tree-switch-conversion.h (struct cluster): Add
8400         is_single_value_p.
8401         (struct simple_cluster): Likewise.
8402         (struct case_tree_node): Add new function has_child.
8403         (do_jump_if_equal): New.
8405 2018-08-27  Martin Liska  <mliska@suse.cz>
8407         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
8408         Add new argument to bit_test_cluster constructor.
8409         (bit_test_cluster::emit): Set bits really number of values
8410         handlel by a test.
8411         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
8412         probability argument.
8413         * tree-switch-conversion.h (struct bit_test_cluster):
8414         Add m_handles_entire_switch.
8416 2018-08-27  Martin Liska  <mliska@suse.cz>
8418         PR tree-optimization/86702
8419         * tree-switch-conversion.c (jump_table_cluster::emit):
8420         Make probabilities even for values in jump table
8421         according to number of cases handled.
8422         (switch_decision_tree::compute_cases_per_edge): Pass
8423         argument to reset_out_edges_aux function.
8424         (switch_decision_tree::analyze_switch_statement): Likewise.
8425         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
8426         Make it static.
8428 2018-08-27  Martin Liska  <mliska@suse.cz>
8430         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
8431         cfun argument explicitly.
8432         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
8433         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
8434         function gimple_switch_default_bb.
8435         (convert_switch_statements):
8436         (expand_builtins):
8437         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
8438         * stmt.c (label_to_block_fn): Use label_to_block and pass
8439         cfun argument explicitly and use gimple_switch_label_bb.
8440         (expand_case): Likewise.
8441         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
8442         cfun argument explicitly. Likewise.
8443         (make_edges_bb): Likewise.
8444         (make_cond_expr_edges): Likewise.
8445         (get_cases_for_edge): Likewise.
8446         (make_gimple_switch_edges): Likewise.
8447         (label_to_block_fn): Likewise.
8448         (label_to_block): Likewise.
8449         (make_goto_expr_edges): Likewise.
8450         (make_gimple_asm_edges): Likewise.
8451         (main_block_label): Likewise.
8452         (group_case_labels_stmt): Likewise.
8453         (find_taken_edge_computed_goto): Likewise.
8454         (find_taken_edge_switch_expr): Likewise.
8455         (gimple_verify_flow_info): Likewise.
8456         (gimple_redirect_edge_and_branch): Likewise.
8457         (gimple_switch_label_bb): New function.
8458         (gimple_switch_default_bb): Likewise.
8459         (gimple_switch_edge): Likewise.
8460         (gimple_switch_default_edge): Likewise.
8461         * tree-cfg.h (label_to_block_fn): Remove and replace ...
8462         (label_to_block): ... with this.
8463         (gimple_switch_label_bb): New.
8464         (gimple_switch_default_bb): Likewise.
8465         (gimple_switch_edge): Likewise.
8466         (gimple_switch_default_edge): Likewise.
8467         * tree-cfgcleanup.c (convert_single_case_switch): Use
8468         new gimple functions and pass new argument to label_to_block.
8469         (cleanup_control_flow_bb):
8470         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
8471         cfun argument explicitly.
8472         (make_eh_edges): Likewise.
8473         (redirect_eh_dispatch_edge): Likewise.
8474         (lower_resx): Likewise.
8475         (lower_eh_dispatch): Likewise.
8476         (maybe_remove_unreachable_handlers): Likewise.
8477         (unsplit_eh): Likewise.
8478         (cleanup_empty_eh): Likewise.
8479         (verify_eh_edges): Likewise.
8480         (verify_eh_dispatch_edge): Likewise.
8481         * tree-ssa-dom.c (record_edge_info): Likewise.
8482         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
8483         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
8484         (thread_through_normal_block): Likewise.
8485         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
8486         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
8487         * tree-switch-conversion.c (switch_conversion::collect): Use new
8488         gimple functions.
8489         (switch_conversion::check_final_bb): Likewise.
8490         (switch_conversion::gather_default_values): Pass new argument
8491         to label_to_block.
8492         (switch_conversion::build_constructors): Likewise.
8493         (switch_decision_tree::compute_cases_per_edge): Use new
8494         gimple_switch_edge function.
8495         (switch_decision_tree::analyze_switch_statement): Pass new argument
8496         to label_to_block.
8497         (switch_decision_tree::try_switch_expansion): Use
8498         gimple_switch_default_edge.
8499         * tree-vrp.c (find_switch_asserts): Pass new argument
8500         to label_to_block.
8501         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
8502         (vr_values::simplify_switch_using_ranges): Likewise.
8504 2018-08-27  Richard Biener  <rguenther@suse.de>
8506         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
8507         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
8509         * tree-ssa-sccvn.h (struct vn_pval): New structure.
8510         (struct vn_nary_op_s): Add unwind_to member.  Add
8511         predicated_values flag and put result into a union together
8512         with a linked list of vn_pval.
8513         (struct vn_ssa_aux): Add name member to make maintaining
8514         a map of SSA name to vn_ssa_aux possible.  Remove no longer
8515         needed info, dfsnum, low, visited, on_sccstack, use_processed
8516         and range_info_anti_range_p members.
8517         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
8518         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
8519         New functions.
8520         (vn_valueize): New global.
8521         (vn_context_bb): Likewise.
8522         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
8523         VN_INFO_PTR_INFO): Remove.
8524         * tree-ssa-sccvn.c: ... (rewrite)
8525         (pass_fre::execute): For -O2+ initialize loops and run
8526         RPO VN in optimistic mode (iterating).  For -O1 and -Og
8527         run RPO VN in non-optimistic mode.
8528         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
8529         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
8530         * doc/invoke.texi (sccvn-max-scc-size): Remove.
8531         (rpo-vn-max-loop-depth): Document.
8532         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
8533         when valuezing the VUSE signals we walked out of the region.
8534         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
8535         (phi_translate): Set VN context block to use for availability
8536         lookup.
8537         (compute_avail): Likewise.
8538         (pre_valueize): New function.
8539         (pass_pre::execute): Adjust to the RPO VN API.
8541         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
8542         (propagate_constants_for_unrolling): Remove.
8543         (tree_unroll_loops_completely): Perform value-numbering
8544         on the unrolled bodies loop parent.
8546 2018-08-27  Richard Biener  <rguenther@suse.de>
8548         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
8549         for partial antic compute.
8551 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
8553         PR rtl-optimization/87065
8554         * combine.c (simplify_if_then_else): Formatting fix.
8555         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
8556         check.
8557         (known_cond): Don't return const_true_rtx for vector modes.  Use
8558         CONST0_RTX instead of const0_rtx.  Formatting fixes.
8560 2018-08-27  Martin Liska  <mliska@suse.cz>
8562         PR gcov-profile/87069
8563         * gcov.c (process_file): Record files already processed
8564         and warn about a file being processed multiple times.
8566 2018-08-27  Martin Liska  <mliska@suse.cz>
8568         PR driver/83193
8569         * config/aarch64/aarch64.c (aarch64_override_options_internal):
8570         Set default values for x_aarch64_*_string strings.
8571         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
8572         prefix.  For -mabi do not print '=ABI' in help and use
8573         <option_value> format for -msve-vector-bits and -moverride
8574         options.
8576 2018-08-26  Jeff Law  <law@redhat.com>
8578         * config/mips/frame-header-opt.c: Include "backend.h" rather than
8579         "cfg.h"
8581 2018-08-26  Marek Polacek  <polacek@redhat.com>
8583         PR c++/87029, Implement -Wredundant-move.
8584         * doc/invoke.texi: Document -Wredundant-move.
8586 2018-08-25  Martin Sebor  <msebor@redhat.com>
8588         PR tree-optimization/87059
8589         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
8590         to the same type as the other.
8591         * fold-const.c (fold_binary_loc): Assert expectation.
8593 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
8595         * config/darwin.c (machopic_legitimize_pic_address): Clean up
8596         extraneous parentheses, dead code section and formatting.
8598 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
8600         PR c++/87091
8601         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
8602         wide enough for jumps in the line-numbering to be visible.
8603         (layout::print_gap_in_line_numbering): New member function.
8604         (layout::calculate_line_spans): When using line numbering, merge
8605         line spans that are only 1 line apart.
8606         (diagnostic_show_locus): When printing line numbers, show gaps in
8607         line numbering directly, rather than printing headers.
8608         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
8609         line-numbering with multiple line spans.
8610         (selftest::test_fixit_insert_containing_newline_2): Add test of
8611         line-numbering, in which the spans are close enough to be merged.
8613 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8615         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
8616         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
8617         * tree-vrp.c (range_is_nonnull): Remove.
8618         (range_includes_zero_p): Accept value_range instead of min/max.
8619         (extract_range_from_binary_expr_1): Do not early bail on
8620         POINTER_PLUS_EXPR.
8621         Use range_includes_zero_p instead of range_is_nonnull.
8622         (extract_range_from_unary_expr): Use range_includes_zero_p instead
8623         of range_is_nonnull.
8624         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
8625         special case VR_ANTI_RANGE.
8626         (vrp_finalize): Same.
8627         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
8628         instead of min/max.
8629         (range_is_nonnull): Remove.
8630         * vr-values.c (vrp_stmt_computes_nonzero): Use
8631         range_includes_zero_p instead of range_is_nonnull.
8632         (extract_range_basic): Pass value_range to range_includes_zero_p
8633         instead of range_is_nonnull.
8635 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
8637         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
8638         * emit-rtl.h (rtl_data): Remove return_bnd.
8639         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
8640         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
8641         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
8642         (POINTER_BOUNDS_MODE): Remove definition.
8643         (make_pointer_bounds_mode): Remove.
8644         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
8645         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
8646         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
8647         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
8648         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
8649         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
8650         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
8652         * config/i386/i386-modes.def (BND32, BND64): Remove.
8653         * config/i386/i386.c (dbx_register_map): Remove bound registers.
8654         (dbx64_register_map): Ditto.
8655         (svr4_dbx_register_map): Ditto.
8656         (indirect_thunk_bnd_needed): Remove.
8657         (indirect_thunks_bnd_used): Ditto.
8658         (indirect_return_bnd_needed): Ditto.
8659         (indirect_return_via_cx_bnd): Ditto.
8660         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
8661         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
8662         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
8663         (output_indirect_thunk_function): Remove handling of
8664         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
8665         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
8666         (ix86_save_reg): Remove handling of crtl->return_bnd.
8667         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
8668         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
8669         and UNSPEC_BNDLX_ADDR.
8670         (ix86_output_indirect_branch_via_reg): Remove handling of
8671         indirect_thunk_prefix_bnd.
8672         (ix86_output_indirect_branch_via_push): Ditto.
8673         (ix86_output_function_return): Ditto.
8674         (ix86_output_indirect_function_return): Ditto.
8675         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
8676         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
8677         (CALL_USED_REGISTERS): Ditto.
8678         (REG_ALLOC_ORDER): Update for removal of bound registers.
8679         (HI_REGISTER_NAMES): Ditto.
8680         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
8681         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
8682         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
8683         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
8684         (FIRST_PSEUDO_REG): Update.
8685         (BND): Remove mode iterator.
8686         * config/i386/predicates.md (bnd_mem_operator): Remove.
8688 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8690         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
8691         vectors.
8693 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8695         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
8696         the case in which the permute needs only a single element and
8697         repeats for every vector of the result.  Extend that case to
8698         handle variable-length vectors.
8699         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8701 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
8703         PR debug/79342
8704         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
8705         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
8707 2018-08-24  Richard Biener  <rguenther@suse.de>
8709         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
8710         bb_flags_allocated members.
8711         (auto_flag): New RAII class for allocating flags.
8712         (auto_edge_flag): New RAII class for allocating edge flags.
8713         (auto_bb_flag): New RAII class for allocating bb flags.
8714         * cfgloop.c (verify_loop_structure): Allocate temporary edge
8715         flag dynamically.
8716         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
8717         in favor of temporarily allocated BB flag.
8718         * hsa-brig.c: Re-order includes.
8719         * hsa-dump.c: Likewise.
8720         * hsa-regalloc.c: Likewise.
8721         * print-rtl.c: Likewise.
8722         * profile-count.c: Likewise.
8724 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
8726         PR target/86989
8727         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
8728         the TOC register.
8730 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8732         PR 87073/bootstrap
8733         * wide-int-range.cc (wide_int_range_div): Do not ignore result
8734         from wide_int_range_multiplicative_op.
8736 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
8738         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
8739         "permutaion".
8741 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
8743         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
8744         to 'expanded'.
8746 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
8748         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
8749         full GENERIC expression used for replacement.
8751 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
8753         * tree-vrp.c (abs_extent_range): Remove.
8754         (extract_range_into_wide_ints): Pass wide ints by reference.
8755         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
8756         Pass wide ints by reference in all calls to
8757         extract_range_into_wide_ints.
8758         * wide-int-range.cc (wide_int_range_div): New.
8759         * wide-int-range.h (wide_int_range_div): New.
8760         (wide_int_range_includes_zero_p): New.
8761         (wide_int_range_zero_p): New.
8763 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
8765         * config/aarch64/aarch64.md (arches): New enum.
8766         (arch): New enum attr.
8767         (arch_enabled): New attr.
8768         (enabled): Now uses arch_enabled only.
8769         (simd, sve, fp16): Removed attribute.
8770         (fp): Attr now defined in terms of 'arch'.
8771         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
8772         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
8773         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
8774         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
8775         attributes into 'arch'.
8776         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
8777         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
8778         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
8779         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
8780         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
8781         'simd' attribute into 'arch'.
8782         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
8783         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
8784         Convert use of 'fp' attribute to 'arch'.
8785         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
8786         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
8787         into 'arch'.
8788         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
8789         (different modes) Merge 'fp' and 'simd' into 'arch'.
8790         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
8791         'simd' into 'arch'.
8793 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
8795         PR rtl-optimization/87026
8796         * expmed.c (canonicalize_comparison): If we can no longer create
8797         pseudoregisters, don't.
8799 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
8801         PR target/86951
8802         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
8803         prototype.
8804         * config/arm/arm.c (speculation_barrier_libfunc): New static
8805         variable.
8806         (arm_init_libfuncs): Initialize it.
8807         (arm_emit_speculation_barrier): New function.
8808         * config/arm/arm.md (speculation_barrier): Call
8809         arm_emit_speculation_barrier for architectures that do not have
8810         DSB or ISB.
8811         (speculation_barrier_insn): Only match on Armv7 or later.
8813 2018-08-23  Richard Biener  <rguenther@suse.de>
8815         PR middle-end/87024
8816         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
8817         calls.
8819 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8821         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
8822         of single-vector TBLs.
8823         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
8824         one input is given.
8826 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8828         PR target/85910
8829         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
8830         aarch64_evpc_tbl guard.
8832 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8834         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
8835         behaviour.
8837 2018-08-22  Martin Sebor  <msebor@redhat.com>
8839         PR middle-end/87052
8840         * tree-pretty-print.c (pretty_print_string): Add argument.
8841         (dump_generic_node): Call to pretty_print_string with string size.
8843 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
8845         PR rtl-optimization/86771
8846         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
8847         of two SETs into those two SETs, one to be placed at i2, if that SETs
8848         destination is modified between i2 and i3.
8850 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8852         PR tree-optimization/86725
8853         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
8854         function.
8855         (vect_analyze_scalar_cycles_1): Check it.
8857 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8859         PR tree-optimization/86725
8860         * tree-vect-loop.c (vect_is_simple_reduction): When treating
8861         an outer loop phi as a double reduction, make sure that the
8862         single user of the phi result is an inner loop phi.
8864 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8866         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
8867         grouped stores with gaps to a strided group.
8869 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8871         * tree-vect-stmts.c (get_group_load_store_type)
8872         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
8873         first statement in a group.
8875 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8877         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
8878         the sequence used in gcc/gcc.c.
8880 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8882         PR other/704
8883         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
8884         building it.
8886 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8888         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
8889         Darwin10-specific unwinder-shim.
8890         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
8891         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
8892         New to cater for Darwin10 Rosetta.
8894 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8896         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
8897         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
8899 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8901         PR bootstrap/81033
8902         PR target/81733
8903         PR target/52795
8904         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
8905         (dwarf2out_switch_text_section): Generate a local label for the
8906         second function sub-section and apply it as the second FDE start
8907         label.
8908         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
8909         second sub-section start.
8911 2018-08-22  Richard Biener  <rguenther@suse.de>
8913         PR tree-optimization/86988
8914         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
8916 2018-08-22  Richard Biener  <rguenther@suse.de>
8918         PR tree-optimization/86945
8919         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
8921 2018-08-22  Alexandre Oliva <oliva@adacore.com>
8923         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
8924         a comment about how uses of r2 for .sdata2 come about.
8926 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
8928         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
8930 2018-08-21  Marek Polacek  <polacek@redhat.com>
8932         PR c++/86981, Implement -Wpessimizing-move.
8933         * doc/invoke.texi: Document -Wpessimizing-move.
8935 2018-08-21  Jan Hubicka  <jh@suse.cz>
8937         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
8938         * tree.h (is_redundant_typedef): Remove.
8939         * dwarf2out.c (is_redundant_typedef): Turn into static function.
8941 2018-08-21  Jan Hubicka  <jh@suse.cz>
8943         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
8944         when possible.
8946 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
8948         * expmed.c (extract_low_bits): Reject invalid subregs early.
8950 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8952         PR middle-end/86121
8953         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
8954         behaviour.
8956 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8958         * config/vxworks.h: Guard vxworks_asm_out_constructor and
8959         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
8960         * config/vxworks.c: Likewise.
8962 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8964         * config/vxworks.c: Set targetm.have_ctors_dtors
8965         if HAVE_INITFINI_ARRAY_SUPPORT.
8966         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
8967         if HAVE_INITFINI_ARRAY_SUPPORT.
8969 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8971         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
8972         default search path for VxWorks < 7.
8974 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8976         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
8977         (get_format_string): Refer to c_getstr.
8979 2018-08-21  Tom de Vries  <tdevries@suse.de>
8981         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8982         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8983         (debuginfo_early_stop): Declare.
8984         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8985         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8986         (debuginfo_early_stop): New function.
8987         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
8988         and debuginfo_early_stop.
8989         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
8990         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
8991         (general_init): Call debuginfo_early_init.
8992         (finalize): Call debuginfo_fini.
8993         (do_compile): Call debuginfo_init.
8994         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
8995         -fdump-early-debug.
8996         (@item -fdump-debug, @item -fdump-earlydebug): Add.
8998 2018-08-21  Tom de Vries  <tdevries@suse.de>
9000         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
9001         flag_dump_noaddr and flag_dump_unnumbered.
9003 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
9005         * wide-int-range.cc (wide_int_range_abs): New.
9006         (wide_int_range_order_set): Rename from wide_int_range_min_max.
9007         * wide-int-range.h (wide_int_range_abs): New.
9008         (wide_int_range_min_max): New.
9009         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
9010         case to call wide_int_range_abs.
9011         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
9012         (extract_range_from_abs_expr): Delete.
9014 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
9016         PR target/87033
9017         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
9018         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
9019         for indexed loads.
9021 2018-08-20  Nathan Sidwell  <nathan@acm.org>
9022             Jeff Law <law@redhat.com>
9024         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
9025         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
9027 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
9029         PR other/84889
9030         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
9031         (decl_attributes): Likewise.
9032         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
9033         instance.
9034         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9035         * diagnostic-core.h (class auto_diagnostic_group): New class.
9036         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
9037         (diagnostic_report_diagnostic): Handle the first diagnostics within
9038         a group.
9039         (emit_diagnostic): Add auto_diagnostic_group instance.
9040         (inform): Likewise.
9041         (inform_n): Likewise.
9042         (warning): Likewise.
9043         (warning_at): Likewise.
9044         (warning_n): Likewise.
9045         (pedwarn): Likewise.
9046         (permerror): Likewise.
9047         (error): Likewise.
9048         (error_n): Likewise.
9049         (error_at): Likewise.
9050         (sorry): Likewise.
9051         (fatal_error): Likewise.
9052         (internal_error): Likewise.
9053         (internal_error_no_backtrace): Likewise.
9054         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
9055         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
9056         * diagnostic.h (struct diagnostic_context): Add fields
9057         "diagnostic_group_nesting_depth",
9058         "diagnostic_group_emission_count", "begin_group_cb",
9059         "end_group_cb".
9060         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
9061         Add auto_diagnostic_group instance(s).
9062         (find_explicit_erroneous_behavior): Likewise.
9063         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
9064         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
9065         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
9066         (gimplify_va_arg_expr): Likewise.
9067         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
9068         (HSA_SORRY_AT): Likewise.
9069         * ipa-devirt.c (compare_virtual_tables): Likewise.
9070         (warn_odr): Likewise.
9071         * multiple_target.c (expand_target_clones): Likewise.
9072         * opts-common.c (cmdline_handle_error): Likewise.
9073         * reginfo.c (globalize_reg): Likewise.
9074         * substring-locations.c (format_warning_n_va): Likewise.
9075         * tree-inline.c (expand_call_inline): Likewise.
9076         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
9077         * tree-ssa-loop-niter.c
9078         (do_warn_aggressive_loop_optimizations): Likewise.
9079         * tree-ssa-uninit.c (warn_uninit): Likewise.
9080         * tree.c (warn_deprecated_use): Likewise.
9082 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
9084         PR target/87014
9085         * config/i386/i386.md (eh_return): Always update EH return
9086         address in word_mode.
9088 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
9090         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
9091         TARGET_SPLIT_COMPLEX_ARG is defined.
9093 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9095         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
9097 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9099         PR target/86984
9100         * expr.c (expand_assignment): Assert that bitpos is positive.
9101         (store_field): Likewise
9102         (expand_expr_real_1): Make sure that bitpos is positive.
9103         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
9104         integer overflow.
9106 2018-08-20  Nathan Sidwell  <nathan@acm.org>
9108         * Makefile.in (CPP_ID_DATA_H): Delete.
9109         (CPP_INTERNAL_H): Don't add it.
9110         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
9111         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
9113 2018-08-20  Richard Biener  <rguenther@suse.de>
9115         PR tree-optimization/78655
9116         * tree-vrp.c (extract_range_from_binary_expr_1): Make
9117         pointer + offset nonnull if either operand is nonnull work.
9119 2018-08-20  Tom de Vries  <tdevries@suse.de>
9121         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
9122         unless the referenced die describes the added property using
9123         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
9124         Otherwise, add a DW_AT_location to the referenced die.
9126 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
9128         PR target/86994
9129         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
9130         register_operand when calling ix86_set_reg_reg_cost.
9131         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
9132         Set *total to 0 for operands that satisfy x86_64_immediate_operand
9133         predicate and to 1 otherwise.
9135 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
9137         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
9138         emit a diagnostic that it is not supported and reset the option.
9139         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
9140         supported and consume the option.  (ASM_FINAL_SPEC): New.
9142 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
9144         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
9145         a sentence.
9147 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
9149         C-SKY port: Documentation
9151         * doc/extend.texi (C-SKY Function Attributes): New section.
9152         * doc/invoke.texi (Option Summary): Add C-SKY options.
9153         (C-SKY Options): New section.
9154         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
9156 2018-08-17  Jojo  <jijie_rong@c-sky.com>
9157             Huibin Wang  <huibin_wang@c-sky.com>
9158             Sandra Loosemore  <sandra@codesourcery.com>
9159             Chung-Lin Tang  <cltang@codesourcery.com>
9161         C-SKY port: Backend implementation
9163         * config/csky/*: New.
9164         * common/config/csky/*: New.
9166 2018-08-17  Jojo  <jijie_rong@c-sky.com>
9167             Huibin Wang  <huibin_wang@c-sky.com>
9168             Sandra Loosemore  <sandra@codesourcery.com>
9169             Chung-Lin Tang  <cltang@codesourcery.com>
9170             Andrew Jenner  <andrew@codesourcery.com>
9172         C-SKY port: Configury
9174         * config.gcc (csky-*-*): New.
9175         * configure.ac: Add csky to targets for dwarf2 debug_line support.
9176         * configure: Regenerated.
9178 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
9180         * dump-context.h: Include "dumpfile.h".
9181         (dump_context::dump_printf_va): Convert final param from va_list
9182         to va_list *.  Convert from ATTRIBUTE_PRINTF to
9183         ATTRIBUTE_GCC_DUMP_PRINTF.
9184         (dump_context::dump_printf_loc_va): Likewise.
9185         * dumpfile.c: Include "stringpool.h".
9186         (make_item_for_dump_printf_va): Delete.
9187         (make_item_for_dump_printf): Delete.
9188         (class dump_pretty_printer): New class.
9189         (dump_pretty_printer::dump_pretty_printer): New ctor.
9190         (dump_pretty_printer::emit_items): New member function.
9191         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
9192         function.
9193         (dump_pretty_printer::emit_item): New member function.
9194         (dump_pretty_printer::stash_item): New member function.
9195         (dump_pretty_printer::format_decoder_cb): New member function.
9196         (dump_pretty_printer::decode_format): New member function.
9197         (dump_context::dump_printf_va): Reimplement in terms of
9198         dump_pretty_printer.
9199         (dump_context::dump_printf_loc_va): Convert final param from va_list
9200         to va_list *.
9201         (dump_context::begin_scope): Reimplement call to
9202         make_item_for_dump_printf.
9203         (dump_printf): Update for change to dump_printf_va.
9204         (dump_printf_loc): Likewise.
9205         (selftest::test_capture_of_dump_calls): Convert "stmt" from
9206         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
9207         with %T, %E, and %G.
9208         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
9209         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
9210         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
9211         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
9212         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
9213         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
9214         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
9215         within a dump_printf_loc call to "%wu".
9216         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
9217         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
9218         missing space after "=".
9219         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
9220         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
9221         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
9222         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
9223         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
9224         duplicate "vectorized" from message.
9226 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
9228         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
9229         polyNxK_t element's TYPE_STRING_FLAG.
9231 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
9233         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
9234         (they were unnamed before).  Fix comments.
9236 2018-08-17  Nathan Sidwell  <nathan@acm.org>
9238         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
9240 2018-08-17  Richard Biener  <rguenther@suse.de>
9242         PR tree-optimization/86841
9243         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
9245 2018-08-17  Martin Liska  <mliska@suse.cz>
9247         * common.opt: Remove Warn, Init and Report for options with
9248         Ignore/Deprecated flag. Warning is done automatically for
9249         Deprecated flags.
9250         * config/i386/i386.opt: Likewise.
9251         * config/ia64/ia64.opt: Likewise.
9252         * config/rs6000/rs6000.opt: Likewise.
9253         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
9254         Remove usage of flag_check_pointer_bounds.
9255         * lto-wrapper.c (merge_and_complain): Do not handle
9256         OPT_fcheck_pointer_bounds.
9257         (append_compiler_options): Likewise.
9258         * opt-functions.awk: Do not handle Deprecated.
9259         * optc-gen.awk: Check that Var, Report and Init are not
9260         used for an option with Ignore/Deprecated flag.
9261         * opts-common.c (decode_cmdline_option): Do not report
9262         CL_ERR_DEPRECATED.
9263         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
9264         options.
9265         * opts.h (struct cl_option): Remove cl_deprecated flag.
9266         (CL_ERR_DEPRECATED): Remove error enum value.
9268 2018-08-17  Richard Biener  <rguenther@suse.de>
9270         PR middle-end/86505
9271         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
9272         across a va-arg-pack using call adjust its return value accordingly.
9274 2018-08-16  Martin Sebor  <msebor@redhat.com>
9276         PR tree-optimization/86853
9277         * gimple-ssa-sprintf.c (struct format_result): Rename member.
9278         (struct fmtresult): Add member and initialize it in ctors.
9279         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
9280         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
9281         (format_directive): Set POSUNDER4K when MAYFAIL is set.
9282         (parse_directive): Handle %C same as %c.
9283         (sprintf_dom_walker::compute_format_length): Adjust.
9284         (is_call_safe): Adjust.
9286 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9288         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
9289         for determining how to count the elements.
9290         * builtins.h (c_strlen): Adjust prototype.
9291         * expr.c (string_constant): Add new parameter mem_size.
9292         Set *mem_size appropriately.
9293         * expr.h (string_constant): Adjust protoype.
9294         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
9295         * gimple-fold.h (get_range_strlen): Adjust prototype.
9296         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
9297         (format_string): Call get_string_length with eltsize.
9299 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9301         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
9302         to emit the span, rather than setting it as the prefix.
9304 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9306         * diagnostic-show-locus.c (layout::start_annotation_line): Add
9307         "margin_char" parameter, defaulting to space.  Use it in place
9308         of pp_space for the initial part of the margin.
9309         (layout::print_leading_fixits): Use '+' when filling the margin
9310         of line-insertion fix-it hints.
9312 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9314         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
9315         Delete.
9317 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9319         * config/rs6000/altivec.md: Don't set length attribute to the default
9320         value.
9321         * config/rs6000/darwin.md: Ditto.
9322         * config/rs6000/dfp.md: Ditto.
9323         * config/rs6000/htm.md: Ditto.
9324         * config/rs6000/rs6000.md: Ditto.
9325         * config/rs6000/sync.md: Ditto.
9326         * config/rs6000/vsx.md: Ditto.
9328 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9330         * config/rs6000/altivec.md: Don't set length attribute to the default
9331         value, for branch instructions.
9332         * config/rs6000/darwin.md: Ditto.
9333         * config/rs6000/rs6000.md: Ditto.
9335 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9337         * config/rs6000/rs6000.md (length): Always define as const_int 4.
9338         (unnamed conditional branch define_insn): Set length to 4 or 8
9339         depending on offset.
9340         (<bd>_<mode>): Similar, for alternative 0.
9341         (<bd>tf_<mode>): Ditto.
9343 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9345         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
9347 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
9349         * doc/rtl.texi: Replace old RTX class names with new names.
9352 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
9354         * expmed.h (canonicalize_comparison): New declaration.
9355         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
9356         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
9357         * optabs.c (prepare_cmp_insn): Likewise.
9358         * rtl.h (unsigned_condition_p): New function which checks if a
9359         comparison operator is unsigned.
9361 2018-08-16  Nathan Sidwell  <nathan@acm.org>
9363         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
9364         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
9366 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9368         PR target/84711
9369         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
9370         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
9371         (mov<mov>): ..this and enable unconditionally.
9373 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9375         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
9377 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
9379         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
9380         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
9381         "Common" with "Target".
9383 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
9385         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
9386         * doc/invoke.texi (mmitigate-rop): Remove.
9387         * config/i386/i386.c: Do not include "regrename.h".
9388         (ix86_rop_should_change_byte_p, reg_encoded_number)
9389         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
9390         Remove.
9391         (ix86_reorg): Remove call to ix86_mitigate_rop.
9392         * config/i386/i386.md (attr "modrm_class"): Remove.
9393         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
9394         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
9395         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
9397 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
9399         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
9400         allow folding of mergeh() and mergel() for the float and double types.
9401         (fold_mergehl_helper): Rework to handle building a permute tree
9402         for float vectors.
9404 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
9406         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
9407         for TARGET_SSE.
9409 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
9411         * common.opt (fdiagnostics-show-labels): New option.
9412         * diagnostic-show-locus.c (class layout_range): Add field
9413         "m_label".
9414         (class layout): Add field "m_show_labels_p".
9415         (layout_range::layout_range): Add param "label" and use it to
9416         initialize m_label.
9417         (make_range): Pass in NULL for new "label" param of layout_range's
9418         ctor.
9419         (layout::layout): Initialize m_show_labels_p.
9420         (layout::maybe_add_location_range): Pass in loc_range->m_label
9421         when constructing layout_range instances.
9422         (struct line_label): New struct.
9423         (layout::print_any_labels): New member function.
9424         (layout::print_line): Call it if label-printing is enabled.
9425         (selftest::test_one_liner_labels): New test.
9426         (selftest::test_diagnostic_show_locus_one_liner): Call it.
9427         * diagnostic.c (diagnostic_initialize): Initialize
9428         context->show_labels_p.
9429         * diagnostic.h (struct diagnostic_context): Add field
9430         "show_labels_p".
9431         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9432         -fno-diagnostics-show-labels.
9433         * dwarf2out.c (gen_producer_string): Add
9434         OPT_fdiagnostics_show_labels to the ignored options.
9435         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
9436         param.
9437         (gcc_rich_location::maybe_add_expr): Likewise.
9438         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
9439         label" param, defaulting to NULL.
9440         (gcc_rich_location::add_expr): Add "label" param.
9441         (gcc_rich_location::maybe_add_expr): Likewise.
9442         (class text_range_label): New class.
9443         (class range_label_for_type_mismatch): New class.
9444         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
9445         of format_warning_va.
9446         (fmtwarn_n): Likewise for new params of format_warning_n_va.
9447         * lto-wrapper.c (merge_and_complain): Add
9448         OPT_fdiagnostics_show_labels to the "pick one setting" options.
9449         (append_compiler_options): Likewise to the dropped options.
9450         (append_diag_options): Likewise to the passed-on options.
9451         * opts.c (common_handle_option): Handle the new option.
9452         * selftest-diagnostic.c
9453         (test_diagnostic_context::test_diagnostic_context): Enable
9454         show_labels_p.
9455         * substring-locations.c: Include "gcc-rich-location.h".
9456         (format_warning_n_va): Add "fmt_label" and "param_label" params
9457         and use them as appropriate.
9458         (format_warning_va): Add "fmt_label" and "param_label" params,
9459         passing them on to format_warning_n_va.
9460         (format_warning_at_substring): Likewise.
9461         (format_warning_at_substring_n): Likewise.
9462         * substring-locations.h (format_warning_va): Add "fmt_label" and
9463         "param_label" params.
9464         (format_warning_n_va): Likewise.
9465         (format_warning_at_substring): Likewise.
9466         (format_warning_at_substring_n): Likewise.
9467         * toplev.c (general_init): Initialize global_dc->show_labels_p.
9469 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
9471         PR testsuite/86519
9472         * builtins.c (expand_builtin_memcmp): Do not expand the call
9473         when overflow is detected.
9475 2018-08-15  Martin Sebor  <msebor@redhat.com>
9477         PR tree-optimization/71625
9478         * config/aarch64/aarch64-builtins.c
9479         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
9481 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
9483         * config/s390/s390.c (s390_reorg): Remove loop.
9485 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9487         * config/darwin.c
9488          (darwin_function_switched_text_sections): Delete.
9489         * gcc/config/darwin.h
9490          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
9492 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9494         PR target/81685
9495         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
9496         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
9497         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
9499 2018-08-15  Martin Liska  <mliska@suse.cz>
9501         PR tree-optimization/86925
9502         * predict.c (expr_expected_value_1): When taking
9503         later predictor, assign also probability.
9504         Use fold_build2_initializer_loc in order to fold
9505         the expression in -frounding-math.
9507 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
9509         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
9510         patterns.
9511         (expand_vec_perm_1): Try the new method.
9513 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
9515         PR target/86547
9516         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
9517         Check whether lra_live_max_point is 0 before dividing.
9519 2018-08-14  Martin Sebor  <msebor@redhat.com>
9521         PR tree-optimization/86650
9522         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
9523         (vrp_prop::check_mem_ref): Same.
9525 2018-08-13  Liu Hao <lh_mouse@126.com>
9527         * pretty-print.c (eat_esc_sequence): Swap the foreground and
9528         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
9529         and clear it thereafter, as it only works for DBCS.
9531 2018-08-13  Liu Hao <lh_mouse@126.com>
9533         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
9534         handle returned by _get_osf_handle().
9536 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9538         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
9539         for folding vec_perm.
9541 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9543         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
9544         Add support for gimple-folding of vec_pack() and vec_unpack()
9545         intrinsics.
9547 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
9549         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
9550         vec_xst variants to the list.
9551         (rs6000_gimple_fold_builtin): Add support for folding unaligned
9552         vector loads and stores.
9554 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
9556         * config.gcc (rs6000-ibm-aix4.x): Delete.
9557         (rs6000-ibm-aix5.1): Delete.
9558         (rs6000-ibm-aix5.2): Delete.
9559         (rs6000-ibm-aix5.3): Delete.
9560         * config/rs6000/aix43.h: Delete.
9561         * config/rs6000/aix51.h: Delete.
9562         * config/rs6000/aix52.h: Delete.
9563         * config/rs6000/t-aix43: Delete.
9565 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
9567         * config/s390/s390.c (s390_decompose_constant_pool_ref):
9568         New function.
9569         (s390_decompose_address): Factor out constant pool ref
9570         decomposition.
9572 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9574         * config/nds32/nds32-predicates.c
9575         (nds32_can_use_bclr_p): Change return type as bool.
9576         (nds32_can_use_bset_p): Ditto.
9577         (nds32_can_use_btgl_p): Ditto.
9578         (nds32_can_use_bitci_p): Ditto.
9579         * config/nds32/nds32-protos.h
9580         (nds32_can_use_bclr_p): Change declaration.
9581         (nds32_can_use_bset_p): Ditto.
9582         (nds32_can_use_btgl_p): Ditto.
9583         (nds32_can_use_bitci_p): Ditto.
9585 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9587         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
9588         Support -msched-prolog-epilog option.
9589         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
9591 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9593         * common/config/nds32/nds32-common.c
9594         (nds32_option_optimization_table): Enalbe -malways-align.
9596 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9598         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
9599         extra_headers.
9600         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
9601         OPT_misr_secure_ case.
9602         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
9603         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
9604         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
9605         secure attribute.
9606         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
9607         (nds32_isr_info): New field security_level.
9608         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
9609         * config/nds32/nds32.md (return_internal): Consider critical attribute.
9610         * config/nds32/nds32.opt (misr-secure): New option.
9611         * config/nds32/nds32_init.inc: New file.
9612         * config/nds32/nds32_isr.h: New file.
9614 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
9616         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
9617         Update comment for atomic instructions.
9618         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
9619         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
9620         Remove.
9621         (atomic_loaddi): Revise fence expansion to only emit fence prior to
9622         load for __ATOMIC_SEQ_CST model.
9623         (atomic_loaddi_1): Remove float register target.
9624         (atomic_storedi): Handle CONST_INT values.
9625         (atomic_storedi_1): Remove float register source.  Add special case
9626         for zero value.
9627         (memory_barrier): New expander and insn.
9629 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
9631         PR tree-optimization/86835
9632         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
9633         new_stmt after def_gsi, make sure to insert new_square_stmt after
9634         that stmt, not 2 stmts before it.
9636 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
9638         PR target/82418
9639         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
9640         instead of SWI48.
9642 2018-08-10  Martin Liska  <mliska@suse.cz>
9644         PR target/83610
9645         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
9646         function type.
9647         * builtins.c (expand_builtin_expect_with_probability):
9648         New function.
9649         (expand_builtin_expect_with_probability): New function.
9650         (build_builtin_expect_predicate): Add new argumnet probability
9651         for BUILT_IN_EXPECT_WITH_PROBABILITY.
9652         (fold_builtin_expect):
9653         (fold_builtin_2):
9654         (fold_builtin_3):
9655         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
9656         * builtins.h (fold_builtin_expect): Set new argument.
9657         * doc/extend.texi: Document __builtin_expect_with_probability.
9658         * doc/invoke.texi: Likewise.
9659         * gimple-fold.c (gimple_fold_call): Pass new argument.
9660         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
9661         also BUILT_IN_EXPECT_WITH_PROBABILITY.
9662         * predict.c (get_predictor_value): New function.
9663         (expr_expected_value): Add new argument probability. Assume
9664         that predictor and probability are always non-null.
9665         (expr_expected_value_1): Likewise.  For __builtin_expect and
9666         __builtin_expect_with_probability set probability.  Handle
9667         combination in binary expressions.
9668         (tree_predict_by_opcode): Simplify code by simply calling
9669         get_predictor_value.
9670         (pass_strip_predict_hints::execute): Add handling of
9671         BUILT_IN_EXPECT_WITH_PROBABILITY.
9672         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
9673         new predictor.
9674         * tree.h (DECL_BUILT_IN_P): New function.
9676 2018-08-10  Martin Liska  <mliska@suse.cz>
9678         PR tree-optimization/85799
9679         * passes.def: Add argument for pass_strip_predict_hints.
9680         * predict.c (class pass_strip_predict_hints): Add new argument
9681         early_p.
9682         (strip_predictor_early): New function.
9683         (pass_strip_predict_hints::execute): Call the function to
9684         strip predictors.
9685         (strip_predict_hints): New function.
9686         * predict.def: Fix comment.
9688 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
9690         * Makefile.in: Clarify which tm.texi to copy over to assert the
9691         right to grant a GFDL license for all.
9693 2018-08-09  Jeff Law  <law@redhat.com>
9695         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
9696         unused variable.
9698 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
9700         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
9701         prototype.
9703 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9705         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
9706         reductions for variable-length vectors.
9708 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
9710         PR other/84889
9711         * common.opt (fdiagnostics-show-line-numbers): New option.
9712         * diagnostic-show-locus.c (class layout): Add fields
9713         "m_show_line_numbers_p" and "m_linenum_width";
9714         (num_digits): New function.
9715         (test_num_digits): New function.
9716         (layout::layout): Initialize new fields.  Update m_x_offset
9717         logic to handle any left margin.
9718         (layout::print_source_line): Print line number when requested.
9719         (layout::start_annotation_line): New member function.
9720         (layout::print_annotation_line): Call it.
9721         (layout::print_leading_fixits): Likewise.
9722         (layout::print_trailing_fixits): Likewise.  Update calls to
9723         move_to_column for new parameter.
9724         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
9725         it to potentially call start_annotation_line.
9726         (layout::show_ruler): Call start_annotation_line.
9727         (selftest::test_line_numbers_multiline_range): New selftest.
9728         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
9729         and selftest::test_line_numbers_multiline_range.
9730         * diagnostic.c (diagnostic_initialize): Initialize
9731         show_line_numbers_p.
9732         * diagnostic.h (struct diagnostic_context): Add field
9733         "show_line_numbers_p".
9734         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9735         -fno-diagnostics-show-line-numbers.
9736         * dwarf2out.c (gen_producer_string): Add
9737         OPT_fdiagnostics_show_line_numbers to the ignored options.
9738         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
9739         one setting" options.
9740         (append_compiler_options): Likewise to the dropped options.
9741         (append_diag_options): Likewise to the passed-on options.
9742         * opts.c (common_handle_option): Handle the new option.
9743         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
9745 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9747         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
9748         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
9749         third argument of type "const signed char" to descriptions of
9750         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
9751         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
9752         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
9753         __builtin_bcdsub_ov functions.
9755 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9757         PR tree-optimization/86858
9758         * tree-vect-loop.c (vect_is_simple_reduction): Restore
9759         flow_bb_inside_loop_p calls.
9761 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9763         PR tree-optimization/86871
9764         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
9765         instead of gimple_assign_lhs.
9767 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
9769         PR target/86887
9770         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
9771         register constraint to operand 0.
9772         (add<mode>3_carryinC): Likewise.
9773         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
9775 2018-08-09  Martin Liska  <mliska@suse.cz>
9777         PR c/86895
9778         * common.opt: Remove extra line.
9780 2018-08-09  Martin Liska  <mliska@suse.cz>
9782         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
9783         at the end of a line, make first letter capital and end up
9784         a sentence with a dot.
9785         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
9786         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
9787         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
9788         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
9789         (PARAM_MAX_ISL_OPERATIONS): Likewise.
9790         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
9791         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
9792         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
9793         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
9794         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
9795         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
9796         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
9797         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
9798         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
9799         (PARAM_TREE_REASSOC_WIDTH): Likewise.
9800         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
9801         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
9802         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
9804 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
9806         PR target/84332
9807         * config/s390/s390.c (s390_option_override_internal): Reduce the
9808         stack-clash-protection-probe-interval param if it would be too big
9809         for z900.
9811 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
9813         PR target/46179
9814         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
9815         * config/m68k/m68k.c (handle_move_double): Don't call
9816         m68k_final_prescan_insn.
9817         (m68k_adjust_decorated_operand): Renamed from
9818         m68k_final_prescan_insn, remove first and third operand and
9819         simplify.
9820         (print_operand): Call it.
9821         (print_operand_address): Call it.
9823 2018-08-08  Nathan Sidwell  <nathan@acm.org>
9825         * diagnostic.c (diagnostic_report_current_module): Use
9826         linemap_included_from & linemap_included_from_linemap.
9828 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
9830         * config/aarch64/aarch64-cores.def: Add phecda core.
9831         * config/aarch64/aarch64-tune.md: Regenerate.
9832         * doc/invoke.texi: Add phecda core.
9834 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9836         PR target/85295
9837         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
9838         definitions.
9839         * config/s390/s390.md ("movti"): Add more alternatives for
9840         constant to GPR copies.
9842 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9844         * config/s390/s390.c: Fix whitespace damage throughout the file.
9845         * config/s390/s390.h: Likewise.
9846         * config/s390/tpf.h: Likewise.
9848 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9850         * config/s390/s390.c (s390_loadrelative_operand_p):
9851         Remove TARGET_CPU_ZARCH usages.
9852         (s390_rtx_costs): Likewise.
9853         (s390_legitimate_constant_p): Likewise.
9854         (s390_cannot_force_const_mem): Likewise.
9855         (legitimate_reload_constant_p): Likewise.
9856         (s390_preferred_reload_class): Likewise.
9857         (legitimize_pic_address): Likewise.
9858         (legitimize_tls_address): Likewise.
9859         (s390_split_branches): Removed.
9860         (s390_add_execute): Removed.
9861         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
9862         (s390_mainpool_start): Likewise.
9863         (s390_mainpool_finish): Likewise.
9864         (s390_mainpool_cancel): Removed.
9865         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
9866         (s390_chunkify_cancel): Likewise.
9867         (s390_return_addr_rtx): Likewise.
9868         (s390_register_info): Remove split_branches_pending_p uages.
9869         (s390_optimize_register_info): Likewise.
9870         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
9871         split_branches_pending_p usages.
9872         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
9873         (s390_load_got): Likewise.
9874         (s390_expand_split_stack_prologue): Likewise.
9875         (output_asm_nops): Likewise.
9876         (s390_function_profiler): Likewise.
9877         (s390_emit_call): Likewise.
9878         (s390_conditional_register_usage): Likewise.
9879         (s390_optimize_prologue): Likewise.
9880         (s390_reorg): Remove TARGET_CPU_ZARCH and
9881         split_branches_pending_p usages.
9882         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
9883         usages.
9884         (s390_output_indirect_thunk_function): Likewise.
9885         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
9886         (TARGET_CPU_ZARCH_P): Removed.
9887         (struct machine_function): Remove split_branches_pending_p.
9888         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
9890 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9892         * common/config/s390/s390-common.c (processor_flags_table):
9893         Remove flags.
9894         * config.gcc: Remove with_arch/with_tune support.
9895         * config/s390/2064.md: Remove cpu attribute comparisons.
9896         * config/s390/driver-native.c (s390_host_detect_local_cpu):
9897         Remove MTN.
9898         * config/s390/linux.h (ASM_SPEC):
9899         Remove -march support.
9900         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
9901         Use a table to get an arch level.
9902         * config/s390/s390-opts.h (enum processor_type):
9903         Remove enum values.
9904         * config/s390/s390.c
9905         (processor_table): Remove entries, add arch_level values.
9906         (s390_issue_rate): Remove cases.
9907         (s390_option_override): Adjust
9908         s390_option_override_internal() call.
9909         (s390_option_override_internal): Remove deprecation warning.
9910         (s390_valid_target_attribute_tree): Adjust
9911         s390_option_override_internal() call.
9912         * config/s390/s390.h (struct s390_processor):
9913         Share with s390-c.c, add arch_level field.
9914         * config/s390/s390.md:
9915         Remove occurrences in cpu attribute.
9916         * config/s390/s390.opt: Remove -march/-mtune support.
9917         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
9918         * doc/invoke.texi: Remove deprecation warning.
9920 2018-08-08  Luis Machado  <luis.machado@linaro.org>
9922         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
9923         global.
9924         (qdf24xx_tunings): Set vector cost structure to
9925         qdf24xx_vector_cost.
9927         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
9928         <register_sextend>: Set to 3.
9930 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
9932         PR target/86838
9933         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
9934         * config/aarch64/aarch64-simd.md
9935         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
9936         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
9937         (aarch64_frecpx<mode>): ...this new pattern.
9938         * config/aarch64/aarch64-simd-builtins.def: Remove comment
9939         about aarch64_frecp<FRECP:frecp_suffix><mode>.
9941 2018-08-07  Martin Liska  <mliska@suse.cz>
9943         PR middle-end/83023
9944         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
9945         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
9946         * predict.def (PRED_MALLOC_NONNULL): New predictor.
9947         * doc/extend.texi: Document that malloc attribute adds
9948         hit to compiler.
9950 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
9952         PR target/86785
9953         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9954         Define to speculation_safe_value_not_needed.
9956 2018-08-06  Jeff Law  <law@redhat.com>
9958         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
9959         the vr_values instance to cprop_into_stmt.
9960         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
9961         (cprop_operand): Also query EVRP to determine if OP is a constant.
9963 2018-08-06  Nathan Sidwell  <nathan@acm.org>
9965         * diagnostic.c (diagnostic_report_current_module): Reroll
9966         included-at loop.  Translate text.
9968 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
9970         * function-tests.c (selftest::test_expansion_to_rtl): Call
9971         free_after_compilation.
9973 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9975         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
9977 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
9979         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
9980         loops with memory block operations from getting unrolled.
9982 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
9984         PR target/86807
9985         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9986         Define to speculation_safe_value_not_needed.
9988 2018-08-06  Jeff Law  <law@redhat.com>
9990         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
9991         assert.
9993 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9995         PR target/86662
9996         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
9997         with all enabled __intN types.
9999         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
10001 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10003         * alias.c (record_set): Check for clobber high.
10004         * cfgexpand.c (expand_gimple_stmt): Likewise.
10005         * combine-stack-adj.c (single_set_for_csa): Likewise.
10006         * combine.c (find_single_use_1): Likewise.
10007         (set_nonzero_bits_and_sign_copies): Likewise.
10008         (get_combine_src_dest): Likewise.
10009         (is_parallel_of_n_reg_sets): Likewise.
10010         (try_combine): Likewise.
10011         (record_dead_and_set_regs_1): Likewise.
10012         (reg_dead_at_p_1): Likewise.
10013         (reg_dead_at_p): Likewise.
10014         * dce.c (deletable_insn_p): Likewise.
10015         (mark_nonreg_stores_1): Likewise.
10016         (mark_nonreg_stores_2): Likewise.
10017         * df-scan.c (df_find_hard_reg_defs): Likewise.
10018         (df_uses_record): Likewise.
10019         (df_get_call_refs): Likewise.
10020         * dwarf2out.c (mem_loc_descriptor): Likewise.
10021         * haifa-sched.c (haifa_classify_rtx): Likewise.
10022         * ira-build.c (create_insn_allocnos): Likewise.
10023         * ira-costs.c (scan_one_insn): Likewise.
10024         * ira.c (equiv_init_movable_p): Likewise.
10025         (rtx_moveable_p): Likewise.
10026         (interesting_dest_for_shprep): Likewise.
10027         * jump.c (mark_jump_label_1): Likewise.
10028         * postreload-gcse.c (record_opr_changes): Likewise.
10029         * postreload.c (reload_cse_simplify): Likewise.
10030         (struct reg_use): Add source expr.
10031         (reload_combine): Check for clobber high.
10032         (reload_combine_note_use): Likewise.
10033         (reload_cse_move2add): Likewise.
10034         (move2add_note_store): Likewise.
10035         * print-rtl.c (print_pattern): Likewise.
10036         * recog.c (decode_asm_operands): Likewise.
10037         (store_data_bypass_p): Likewise.
10038         (if_test_bypass_p): Likewise.
10039         * regcprop.c (kill_clobbered_value): Likewise.
10040         (kill_set_value): Likewise.
10041         * reginfo.c (reg_scan_mark_refs): Likewise.
10042         * reload1.c (maybe_fix_stack_asms): Likewise.
10043         (eliminate_regs_1): Likewise.
10044         (elimination_effects): Likewise.
10045         (mark_not_eliminable): Likewise.
10046         (scan_paradoxical_subregs): Likewise.
10047         (forget_old_reloads_1): Likewise.
10048         * reorg.c (find_end_label): Likewise.
10049         (try_merge_delay_insns): Likewise.
10050         (redundant_insn): Likewise.
10051         (own_thread_p): Likewise.
10052         (fill_simple_delay_slots): Likewise.
10053         (fill_slots_from_thread): Likewise.
10054         (dbr_schedule): Likewise.
10055         * resource.c (update_live_status): Likewise.
10056         (mark_referenced_resources): Likewise.
10057         (mark_set_resources): Likewise.
10058         * rtl.c (copy_rtx): Likewise.
10059         * rtlanal.c (reg_referenced_p): Likewise.
10060         (single_set_2): Likewise.
10061         (noop_move_p): Likewise.
10062         (note_stores): Likewise.
10063         * sched-deps.c (sched_analyze_reg): Likewise.
10064         (sched_analyze_insn): Likewise.
10066 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10068         * cse.c (invalidate_reg): New function extracted from...
10069         (invalidate): ...here.
10070         (canonicalize_insn): Check for clobber high.
10071         (invalidate_from_clobbers): invalidate clobber highs.
10072         (invalidate_from_sets_and_clobbers): Likewise.
10073         (count_reg_usage): Check for clobber high.
10074         (insn_live_p): Likewise.
10075         * cselib.c (cselib_expand_value_rtx_1):Likewise.
10076         (cselib_invalidate_regno): Check for clobber in setter.
10077         (cselib_invalidate_rtx): Pass through setter.
10078         (cselib_invalidate_rtx_note_stores):
10079         (cselib_process_insn): Check for clobber high.
10080         * cselib.h (cselib_invalidate_rtx): Add operand.
10082 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10084         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
10085         (mark_not_eliminable): Likewise.
10086         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
10087         * lra-lives.c (process_bb_lives): Check for clobber high.
10088         * lra.c (new_insn_reg): Remember clobber highs.
10089         (collect_non_operand_hard_regs): Check for clobber high.
10090         (lra_set_insn_recog_data): Likewise.
10091         (add_regs_to_insn_regno_info): Likewise.
10092         (lra_update_insn_regno_info): Likewise.
10094 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10096         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
10097         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
10099 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10101         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
10102         (copy_insn_1): Likewise.
10103         (gen_hard_reg_clobber_high): New gen function.
10104         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
10105         * genemit.c (gen_exp): Likewise.
10106         (gen_emit_seq): Pass through info.
10107         (gen_insn): Check for CLOBBER_HIGH.
10108         (gen_expand): Pass through info.
10109         (gen_split): Likewise.
10110         (output_add_clobbers): Likewise.
10111         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
10112         (remove_clobbers): Likewise.
10113         * rtl.h (gen_hard_reg_clobber_high): New declaration.
10115 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10117         * doc/rtl.texi (clobber_high): Add.
10118         (parallel): Add in clobber high
10119         * rtl.c (rtl_check_failed_code3): Add function.
10120         * rtl.def (CLOBBER_HIGH): Add expression.
10121         * rtl.h (RTL_CHECKC3): Add macro.
10122         (rtl_check_failed_code3): Add declaration.
10123         (XC3EXP): Add macro.
10125 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
10127         PR target/86386
10128         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
10129         cfun->machine->max_used_stack_alignment if needed.
10131 2018-08-04  Martin Sebor  <msebor@redhat.com>
10133         PR tree-optimization/86571
10134         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
10135         NaN output to 4.
10137 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
10139         PR target/86799
10140         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10141         Define.
10143 2018-08-03  Jeff Law  <law@redhat.com>
10145         PR target/86795
10146         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10147         Define to speculation_safe_value_not_needed.
10149 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
10151         * doc/gcov.texi (-x): Remove duplicate "to".
10152         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
10153         (-Wif-not-aligned): Remove duplicate "is".
10154         (-flto): Remove duplicate "the".
10155         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
10156         duplicate "v5.00.b".
10157         (MSP430 Options): Remove duplicate "and" from the description
10158         of "-mgprel-sec=regexp".
10159         (x86 Options): Remove duplicate copies of "vmldLog102" and
10160         vmlsLog104 from description of "-mveclibabi=type".
10162 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
10164         * internal-fn.h (first_commutative_argument): Declare.
10165         * internal-fn.c (first_commutative_argument): New function.
10166         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
10167         restrictions for pattern statements.  Use first_commutative_argument
10168         to look for commutative operands in calls to internal functions.
10170 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
10172         * Makefile.in (wide-int-range.o): New.
10173         * tree-vrp.c: Move all the wide_int_* functions to...
10174         * wide-int-range.cc: ...here.
10175         * tree-vrp.h: Move all the wide_int_* prototypes to...
10176         * wide-int-range.h: ...here.
10178 2018-08-03  Tom de Vries  <tdevries@suse.de>
10180         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
10181         UI_NONE.
10182         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
10183         * except.c (output_function_exception_table): Do early exit if
10184         targetm_common.except_unwind_info (&global_options) == UI_NONE.
10186 2018-08-03  Martin Liska  <mliska@suse.cz>
10188         * predict.c (dump_prediction): Change to 2 digits
10189         in fraction part.
10191 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
10193         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
10194         neon_dup_q to...
10195         (falkor_am_1_gtov_gtov): ... a new insn reservation.
10197 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
10199         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
10200         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
10201         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
10202         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
10203         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
10205 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
10207         * diagnostic-show-locus.c (diagnostic_show_locus): Use
10208         pp_take_prefix when saving the existing prefix.
10209         * diagnostic.c (diagnostic_append_note): Likewise.
10210         * langhooks.c (lhd_print_error_function): Likewise.
10211         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
10212         param's type.  Free the existing prefix.
10213         (pp_take_prefix): New function.
10214         (pretty_printer::pretty_printer): Drop the prefix parameter.
10215         Rename the length parameter to match the comment.
10216         (pretty_printer::~pretty_printer): Free the prefix.
10217         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
10218         parameter.
10219         (struct pretty_printer): Drop the "const" from "prefix" field's
10220         type and clarify memory management.
10221         (pp_set_prefix): Drop the "const" from the 2nd param.
10222         (pp_take_prefix): New decl.
10224 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
10226         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
10227         for word_mode_ok here instead of passing as argument.
10228         (expand_block_compare): Change select_block_compare_mode() call.
10229         (expand_strncmp_gpr_sequence): New function.
10230         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
10232 2018-08-02  Jeff Law  <law@redhat.com>
10234         PR target/86790
10235         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10236         Define to speculation_safe_value_not_needed.
10238         PR target/86784
10239         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10240         Define to speculation_safe_value_not_needed.
10242 2018-08-02  Tom de Vries  <tdevries@suse.de>
10244         PR target/86660
10245         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
10246         function.  Return UI_TARGET unconditionally.
10247         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
10248         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
10250 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10252         * genemit.c (print_overload_test): Fix typo.
10254 2018-08-02  Richard Biener  <rguenther@suse.de>
10256         PR tree-optimization/86816
10257         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
10258         which checks for value availability before querying it.
10259         (gvn_uses_equal): Use it.
10260         (same_succ_hash): Likewise.
10261         (gimple_equal_p): Likewise.
10263 2018-08-02  Nick Clifton  <nickc@redhat.com>
10265         PR target/86813
10266         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10267         Define to speculation_safe_value_not_needed.
10269         PR target/86810
10270         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10271         Define to speculation_safe_value_not_needed.
10273         PR target/86810
10274         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10275         Define to speculation_safe_value_not_needed.
10277         PR target/86803
10278         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10279         Define to speculation_safe_value_not_needed.
10281         PR target/86797
10282         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10283         Define to speculation_safe_value_not_needed.
10285         PR target/86791
10286         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10287         Define to speculation_safe_value_not_needed.
10289         PR target/86789
10290         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10291         Define to speculation_safe_value_not_needed.
10293         PR target/86787
10294         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10295         Define to speculation_safe_value_not_needed.
10297         PR target/86782
10298         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10299         speculation_safe_value_not_needed.
10301         PR target/86781
10302         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
10303         to speculation_safe_value_not_needed.
10305 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10307         * doc/md.texi: Expand the documentation of instruction names
10308         to mention port-local uses.  Document '@' in pattern names.
10309         * read-md.h (overloaded_instance, overloaded_name): New structs.
10310         (mapping): Declare.
10311         (md_reader::handle_overloaded_name): New member function.
10312         (md_reader::get_overloads): Likewise.
10313         (md_reader::m_first_overload): New member variable.
10314         (md_reader::m_next_overload_ptr): Likewise.
10315         (md_reader::m_overloads_htab): Likewise.
10316         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
10317         m_next_overload_ptr and m_overloads_htab.
10318         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
10319         (get_mode_token, get_code_token, get_int_token): New functions.
10320         (map_attr_string): Add an optional argument that passes back
10321         the associated iterator.
10322         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
10323         (md_reader::handle_overloaded_name, add_overload_instance): New
10324         functions.
10325         (apply_iterators): Handle '@' names.  Report an error if '@'
10326         is used without iterators.
10327         (initialize_iterators): Initialize the new iterator_group fields.
10328         * genopinit.c (handle_overloaded_code_for)
10329         (handle_overloaded_gen): New functions.
10330         (main): Use them to print declarations of maybe_code_for_* and
10331         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
10332         * genemit.c (print_overload_arguments, print_overload_test)
10333         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
10334         (main): Use it to print definitions of maybe_code_for_* and
10335         maybe_gen_* functions.
10336         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
10337         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
10338         instead of explicit mode checks.
10339         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
10340         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
10341         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
10342         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
10343         (aarch64_expand_compare_and_swap): Likewise
10344         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
10345         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
10346         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
10347         (aarch64_constant_pool_reload_icode): Delete.
10348         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
10349         instead of aarch64_constant_pool_reload_icode.  Use
10350         code_for_aarch64_reload_mov instead of explicit mode checks.
10351         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
10352         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
10353         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
10354         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
10355         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
10356         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
10357         (aarch64_atomic_load_op_code): Delete.
10358         (aarch64_emit_atomic_load_op): Likewise.
10359         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
10360         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
10361         instead of aarch64_emit_atomic_load_op.
10362         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
10363         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
10364         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
10365         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
10366         character before the pattern name.
10367         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
10368         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
10369         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
10370         (aarch64_frecps<mode>): Likewise.
10371         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
10372         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
10373         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
10374         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
10375         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
10377 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10379         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
10380         Allow HFmode constants if TARGET_FP_F16INST.
10382 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
10384         PR target/86014
10385         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
10386         No longer check last store for clobber of address register.
10388 2018-08-02  Martin Liska  <mliska@suse.cz>
10390         PR gcov-profile/86817
10391         * gcov.c (process_all_functions): New function.
10392         (main): Call it.
10393         (process_file): Move functions processing to
10394         process_all_functions.
10396 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
10398         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
10399         "const" to the "gimple *" and "rtx_insn *" parameters.
10400         * dumpfile.h (dump_user_location_t::dump_user_location_t):
10401         Likewise.
10402         (dump_location_t::dump_location_t): Likewise.
10404 2018-08-01  Martin Sebor  <msebor@redhat.com>
10406         PR tree-optimization/86650
10407         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
10408         rather than a "gcall *".  Directly pass the data of interest
10409         to percent_K_format, rather than building a temporary CALL_EXPR
10410         to hold it.
10411         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
10412         (gimple_fold_builtin_strncat): Adjust.
10413         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
10414         gcall* argument with gimple*.
10415         * gimple-ssa-warn-restrict.c (check_call): Same.
10416         (wrestrict_dom_walker::before_dom_children): Same.
10417         (builtin_access::builtin_access): Same.
10418         (check_bounds_or_overlap): Same
10419         (maybe_diag_overlap): Same.
10420         (maybe_diag_offset_bounds): Same.
10421         * tree-diagnostic.c (default_tree_printer): Move usage of
10422         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
10423         to this callsite.
10424         * tree-pretty-print.c (percent_K_format): Add argument.
10425         * tree-pretty-print.h: Add argument.
10426         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
10427         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
10428         (maybe_diag_stxncpy_trunc): Same.
10429         (handle_builtin_stxncpy): Same.
10430         (handle_builtin_strcat): Same.
10432 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10434         * match.pd: Optimise pointer range checks.
10436 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10438         PR tree-optimization/86758
10439         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
10440         to remove pattern statements.
10442 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10444         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
10445         result of dfs_enumerate_from when constructing stmt_vec_infos,
10446         instead of additionally calling get_loop_body.
10448 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10450         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
10451         parameter.
10452         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
10453         When creating an iv, assert that the step is not known to be zero.
10454         (vect_setup_realignment): Update call accordingly.
10455         * tree-vect-stmts.c (vectorizable_store): Likewise.
10456         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
10458 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10460         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
10461         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
10462         (vectorizable_reduction): Likewise.
10463         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10464         (vect_detect_hybrid_slp_stmts): Likewise.
10465         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10467 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
10469         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
10470         (wide_int_set_zero_nonzero_bits): ...this.
10471         (zero_nonzero_bits_from_vr): Rename to...
10472         (vrp_set_zero_nonzero_bits): ...this.
10473         (extract_range_from_multiplicative_op_1): Abstract wide int
10474         code...
10475         (wide_int_range_multiplicative_op): ...here.
10476         (extract_range_from_binary_expr_1): Extract wide int binary
10477         operations into their own functions.
10478         (wide_int_range_lshift): New.
10479         (wide_int_range_can_optimize_bit_op): New.
10480         (wide_int_range_shift_undefined_p): New.
10481         (wide_int_range_bit_xor): New.
10482         (wide_int_range_bit_ior): New.
10483         (wide_int_range_bit_and): New.
10484         (wide_int_range_trunc_mod): New.
10485         (extract_range_into_wide_ints): New.
10486         (vrp_shift_undefined_p): New.
10487         (extract_range_from_multiplicative_op): New.
10488         (vrp_can_optimize_bit_op): New.
10489         * tree-vrp.h (value_range::dump): New.
10490         (wide_int_range_multiplicative_op): New.
10491         (wide_int_range_lshift):New.
10492         (wide_int_range_shift_undefined_p): New.
10493         (wide_int_range_bit_xor): New.
10494         (wide_int_range_bit_ior): New.
10495         (wide_int_range_bit_and): New.
10496         (wide_int_range_trunc_mod): New.
10497         (zero_nonzero_bits_from_bounds): Rename to...
10498         (wide_int_set_zero_nonzero_bits): ...this.
10499         (zero_nonzero_bits_from_vr): Rename to...
10500         (vrp_set_zero_nonzero_bits): ...this.
10501         (range_easy_mask_min_max): Rename to...
10502         (wide_int_range_can_optimize_bit_op): this.
10503         * vr-values.c (simplify_bit_ops_using_ranges): Rename
10504         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
10506 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10508         * tree-vectorizer.h (vect_orig_stmt): New function.
10509         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
10510         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
10511         (vect_create_epilog_for_reduction): Likewise.
10512         (vectorizable_live_operation): Likewise.
10513         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10514         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
10515         * tree-vect-stmts.c (vectorizable_call): Likewise.
10516         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
10518 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10520         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
10521         argument.
10522         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
10523         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
10524         (vect_transform_loop): Likewise.
10525         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10527 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10529         * tree-vectorizer.h (vect_schedule_slp): Return void.
10530         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10531         (vect_schedule_slp): Likewise.
10533 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10535         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
10536         argument.
10537         (vect_transform_loop): Update calls accordingly.  Schedule SLP
10538         instances before the main loop, if any exist.
10540 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10542         PR tree-optimization/86749
10543         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
10544         If the lhs is used in a COND_EXPR, check that it is being used
10545         as the "then" or "else" value.
10547 2018-08-01  Tom de Vries  <tdevries@suse.de>
10549         PR target/86800
10550         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10551         speculation_safe_value_not_needed.
10553 2018-08-01  Richard Biener  <rguenther@suse.de>
10555         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
10556         as base and offset.
10558 2018-08-01  Martin Liska  <mliska@suse.cz>
10560         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
10561         format how successful transformation is dumped.
10562         (gimple_mod_pow2_value_transform): Likewise.
10563         (gimple_mod_subtract_transform): Likewise.
10564         (gimple_stringops_transform): Likewise.
10566 2018-08-01  Martin Liska  <mliska@suse.cz>
10568         PR value-prof/35543
10569         * value-prof.c (interesting_stringop_to_profile_p):
10570         Simplify the code and add BUILT_IN_MEMMOVE.
10571         (gimple_stringops_transform): Likewise.
10573 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
10575         * config/aarch64/aarch64-simd.md
10576         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
10577         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
10578         use GPI iterator instead of SI mode.
10580 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10582         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
10583         rs6000_speculation_barrier.
10584         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
10585         new barrier pattern name.
10587 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10589         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
10590         (speculation_barrier): New insn.
10592 2018-08-01  Richard Biener  <rguenther@suse.de>
10594         PR bootstrap/86724
10595         * graphite.h: Include isl/id.h and isl/space.h to allow build
10596         with ISL 0.20.
10598 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
10600         PR target/86651
10601         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10602         mode for COFF targets.
10603         * defaults.h (TARGET_COFF): Define.
10604         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
10605         TARGET_COFF): Define.
10606         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
10607         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
10608         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
10610 2018-07-31  Alexandre Oliva <oliva@adacore.com>
10612         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
10613         * gimple-streamer-out.c (output_bb): Save it.
10614         * lto-streamer-in.c (input_struct_function_base): Restore
10615         instance discriminator if available.  Create map on demand.
10616         * lto-streamer-out.c (output_struct_function_base): Save it if
10617         available.
10618         * final.c (decl_to_instance_map): Document LTO strategy.
10620 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
10621             Olivier Hainque  <hainque@adacore.com>
10623         * debug.h (decl_to_instance_map_t): New type.
10624         (decl_to_instance_map): Declare.
10625         (maybe_create_decl_to_instance_map): New inline function.
10626         * final.c (bb_discriminator, last_bb_discriminator): New statics,
10627         to track basic block discriminators.
10628         (final_start_function_1): Initialize them.
10629         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
10630         bb_discriminator.
10631         (decl_to_instance_map): New variable.
10632         (map_decl_to_instance, maybe_set_discriminator): New functions.
10633         (notice_source_line): Set discriminator.
10635 2018-07-31  Ian Lance Taylor  <iant@golang.org>
10637         * targhooks.c (default_have_speculation_safe_value): Add
10638         ATTRIBUTE_UNUSED.
10640 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10642         * dump-context.h: Include "pretty-print.h".
10643         (dump_context::refresh_dumps_are_enabled): New decl.
10644         (dump_context::emit_item): New decl.
10645         (class dump_context): Add fields "m_test_pp" and
10646         "m_test_pp_flags".
10647         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10648         (temp_dump_context::get_dumped_text): New decl.
10649         (class temp_dump_context): Add field "m_pp".
10650         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
10651         (dump_context::refresh_dumps_are_enabled): ...and add a test for
10652         m_test_pp.
10653         (set_dump_file): Update for above change.
10654         (set_alt_dump_file): Likewise.
10655         (dump_loc): New overload, taking a pretty_printer *.
10656         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
10657         to any test pretty-printer.
10658         (make_item_for_dump_gimple_stmt): New function, adapted from
10659         optinfo::add_gimple_stmt.
10660         (dump_context::dump_gimple_stmt): Call it, and use the result,
10661         eliminating the direct usage of dump_file and alt_dump_file in
10662         favor of indirectly using them via emit_item.
10663         (make_item_for_dump_gimple_expr): New function, adapted from
10664         optinfo::add_gimple_expr.
10665         (dump_context::dump_gimple_expr): Call it, and use the result,
10666         eliminating the direct usage of dump_file and alt_dump_file in
10667         favor of indirectly using them via emit_item.
10668         (make_item_for_dump_generic_expr): New function, adapted from
10669         optinfo::add_tree.
10670         (dump_context::dump_generic_expr): Call it, and use the result,
10671         eliminating the direct usage of dump_file and alt_dump_file in
10672         favor of indirectly using them via emit_item.
10673         (make_item_for_dump_printf_va): New function, adapted from
10674         optinfo::add_printf_va.
10675         (make_item_for_dump_printf): New function.
10676         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
10677         and use the result, eliminating the direct usage of dump_file and
10678         alt_dump_file in favor of indirectly using them via emit_item.
10679         (make_item_for_dump_dec): New function.
10680         (dump_context::dump_dec): Call it, and use the result,
10681         eliminating the direct usage of dump_file and alt_dump_file in
10682         favor of indirectly using them via emit_item.
10683         (make_item_for_dump_symtab_node): New function, adapted from
10684         optinfo::add_symtab_node.
10685         (dump_context::dump_symtab_node): Call it, and use the result,
10686         eliminating the direct usage of dump_file and alt_dump_file in
10687         favor of indirectly using them via emit_item.
10688         (dump_context::begin_scope): Reimplement, avoiding direct usage
10689         of dump_file and alt_dump_file in favor of indirectly using them
10690         via emit_item.
10691         (dump_context::emit_item): New member function.
10692         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10693         Set up test pretty-printer on the underlying context.  Call
10694         refresh_dumps_are_enabled.
10695         (temp_dump_context::~temp_dump_context): Call
10696         refresh_dumps_are_enabled.
10697         (temp_dump_context::get_dumped_text): New member function.
10698         (selftest::verify_dumped_text): New function.
10699         (ASSERT_DUMPED_TEXT_EQ): New macro.
10700         (selftest::test_capture_of_dump_calls): Run all tests twice, with
10701         and then without optinfo enabled.  Add uses of
10702         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
10703         * dumpfile.h: Update comment for the dump_* API.
10704         * optinfo-emit-json.cc
10705         (selftest::test_building_json_from_dump_calls): Update for new
10706         param for temp_dump_context ctor.
10707         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
10708         and "m_owned" field.
10709         (optinfo_item::~optinfo_item): Likewise.
10710         (optinfo::add_item): New member function.
10711         (optinfo::emit): Update comment.
10712         (optinfo::add_string): Delete.
10713         (optinfo::add_printf): Delete.
10714         (optinfo::add_printf_va): Delete.
10715         (optinfo::add_gimple_stmt): Delete.
10716         (optinfo::add_gimple_expr): Delete.
10717         (optinfo::add_tree): Delete.
10718         (optinfo::add_symtab_node): Delete.
10719         (optinfo::add_dec): Delete.
10720         * optinfo.h (class dump_context): New forward decl.
10721         (optinfo::add_item): New decl.
10722         (optinfo::add_string): Delete.
10723         (optinfo::add_printf): Delete.
10724         (optinfo::add_printf_va): Delete.
10725         (optinfo::add_gimple_stmt): Delete.
10726         (optinfo::add_gimple_expr): Delete.
10727         (optinfo::add_tree): Delete.
10728         (optinfo::add_symtab_node): Delete.
10729         (optinfo::add_dec): Delete.
10730         (optinfo::add_poly_int): Delete.
10731         (optinfo_item::optinfo_item): Remove "owned" param.
10732         (class optinfo_item): Remove field "m_owned".
10734 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10736         PR middle-end/86705
10737         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
10738         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
10739         requested variable alignment.
10740         (expand_one_ssa_partition): Likewise.
10741         (expand_one_var): Likewise.
10743 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10745         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10746         to speculation_safe_value_not_needed.
10748 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10750         * targhooks.h (speculation_safe_value_not_needed): New prototype.
10751         * targhooks.c (speculation_safe_value_not_needed): New function.
10752         * target.def (have_speculation_safe_value): Update documentation.
10753         * doc/tm.texi: Regenerated.
10755 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10757         * config/aarch64/iterators.md (ALLI_TI): New iterator.
10758         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
10759         expand.
10760         (despeculate_copy<ALLI:mode>_insn): New insn.
10761         (despeculate_copyti_insn): New insn.
10762         (despeculate_simple<ALLI:mode>): New insn
10763         (despeculate_simpleti): New insn.
10764         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
10765         function.
10766         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
10767         aarch64_speculation_safe_value.
10768         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
10770 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10772         * config/aarch64/aarch64-speculation.cc: New file.
10773         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
10774         before pass_reorder_blocks.
10775         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
10776         prototype.
10777         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
10778         X14 and X15 when tracking speculation.
10779         * config/aarch64/aarch64.md (register name constants): Add
10780         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
10781         (unspec): Add UNSPEC_SPECULATION_TRACKER.
10782         (speculation_barrier): New insn attribute.
10783         (cmp<mode>): Allow SP in comparisons.
10784         (speculation_tracker): New insn.
10785         (speculation_barrier): Add speculation_barrier attribute.
10786         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
10787         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
10788         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
10790 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10792         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
10793         aarch64_track_speculation is true.
10794         (tb<optab><mode>1): Likewise.
10795         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
10796         generate CB[N]Z when tracking speculation.
10797         (aarch64_split_compare_and_swap): Likewise.
10798         (aarch64_split_atomic_op): Likewise.
10800 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10802         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
10804 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10806         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
10807         (speculation_barrier): New insn.
10809 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10811         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
10812         * config/arm/arm.md (speculation_barrier): New expand.
10813         (speculation_barrier_insn): New pattern.
10815 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10817         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
10818         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
10819         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
10820         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
10821         list.
10822         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
10823         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
10824         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
10825         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
10826         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
10827         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
10828         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
10829         * builtins.c (expand_speculation_safe_value): New function.
10830         (expand_builtin): Call it.
10831         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
10832         * doc/extend.texi: Document __builtin_speculation_safe_value.
10833         * doc/md.texi: Document "speculation_barrier" pattern.
10834         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
10835         TARGET_HAVE_SPECULATION_SAFE_VALUE.
10836         * doc/tm.texi: Regenerated.
10837         * target.def (have_speculation_safe_value, speculation_safe_value): New
10838         hooks.
10839         * targhooks.c (default_have_speculation_safe_value): New function.
10840         (default_speculation_safe_value): New function.
10841         * targhooks.h (default_have_speculation_safe_value): Add prototype.
10842         (default_speculation_safe_value): Add prototype.
10844 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10846         * dump-context.h (dump_context::dump_loc): New decl.
10847         * dumpfile.c (dump_context::dump_loc): New member function.
10848         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
10849         and dump_gimple_stmt.
10850         (dump_context::dump_gimple_expr_loc): Likewise, using
10851         dump_gimple_expr.
10852         (dump_context::dump_generic_expr_loc): Likewise, using
10853         dump_generic_expr.
10854         (dump_context::dump_printf_loc_va): Likewise, using
10855         dump_printf_va.
10856         (dump_context::begin_scope): Explicitly using the global function
10857         "dump_loc", rather than the member function.
10859 2018-07-31  Martin Sebor  <msebor@redhat.com>
10861         PR tree-optimization/86741
10862         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
10864 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
10866         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
10867         depend on whether prefetch instructions will be emitted or not.
10868         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
10869         will be emitted or not.
10870         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
10871         (TARGET_SETMEM_PFD): New macros.
10873 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10875         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
10876         (NULL_STMT_VEC_INFO): Delete.
10877         (stmt_vec_info::operator*): Likewise.
10878         (stmt_vec_info::operator gimple *): Likewise.
10879         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
10880         of NULL_STMT_VEC_INFO.
10881         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10882         (vect_reassociating_reduction_p): Likewise.
10883         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10884         (vectorizable_store): Likewise.
10885         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
10886         (vec_info::free_stmt_vec_infos): Likewise.
10888 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10890         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
10891         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
10893 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10895         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
10896         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10897         (vec_info::free_stmt_vec_info): New private member functions.
10898         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
10899         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
10900         * tree-parloops.c (gather_scalar_reductions): Remove calls to
10901         set_stmt_vec_info_vec and free_stmt_vec_infos.
10902         * tree-vect-loop.c (_loop_vec_info): Remove call to
10903         set_stmt_vec_info_vec.
10904         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
10905         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
10906         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
10907         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10908         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
10909         assignments in {vec_info::,}new_stmt_vec_info that are redundant
10910         with the clearing in the xcalloc.
10911         (stmt_vec_info_vec): Delete.
10912         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
10913         (vectorize_loops): Likewise.
10914         (vec_info::~vec_info): Remove argument from call to
10915         free_stmt_vec_infos.
10916         (vec_info::add_stmt): Remove vinfo argument from call to
10917         new_stmt_vec_info.
10919 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10921         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
10922         rather than a gimple stmt.
10923         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
10924         information for pattern statements when passed the original
10925         statement; instead wait to be passed the pattern statement itself.
10926         Don't call set_vinfo_for_stmt here.
10927         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
10928         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
10929         stmt_vec_infos here.
10930         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
10931         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
10932         stmt_vec_infos entry.
10934 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10936         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
10937         * tree-vectorizer.c (vec_info::replace_stmt): New function.
10938         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
10939         * tree-vect-stmts.c (vectorizable_call): Likewise.
10940         (vectorizable_simd_clone_call): Likewise.
10942 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10944         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
10945         * tree-vectorizer.c (vec_info::remove_stmt): New function.
10946         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
10947         * tree-vect-loop.c (vect_transform_loop): Likewise.
10948         * tree-vect-slp.c (vect_schedule_slp): Likewise.
10949         * tree-vect-stmts.c (vect_remove_stores): Likewise.
10951 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10953         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
10954         (vect_dr_stmt): Delete.
10955         * tree-vectorizer.c (vec_info::lookup_dr): New function.
10956         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
10957         of DR_VECT_AUX.
10958         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10959         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
10960         (vect_verify_datarefs_alignment, vect_peeling_supportable)
10961         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10962         (vect_analyze_data_refs): Likewise.
10963         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
10964         argument.
10965         (vect_find_same_alignment_drs): Likewise.
10966         (vect_slp_analyze_node_dependences): Update calls accordingly.
10967         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10968         instead of DR_VECT_AUX.
10969         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
10970         of a vector data references.  Use vec_info::lookup_dr instead of
10971         DR_VECT_AUX.
10972         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
10973         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10974         instead of DR_VECT_AUX.
10976 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10978         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
10979         dr_vec_info.
10980         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10981         accordingly.
10982         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10983         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10984         (vect_gen_prolog_loop_niters): Likewise.
10986 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10988         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
10989         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
10990         (vect_known_alignment_in_bytes, vect_dr_behavior)
10991         (vect_get_scalar_dr_size): Take references as dr_vec_infos
10992         instead of data_references.  Update calls to other routines for
10993         which the same change has been made.
10994         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
10995         dr_vec_infos instead of stmt_vec_infos.
10996         (vect_analyze_data_ref_dependence): Update call accordingly.
10997         (vect_slp_analyze_data_ref_dependence)
10998         (vect_record_base_alignments): Use DR_VECT_AUX.
10999         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
11000         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
11001         (vector_alignment_reachable_p, vect_get_data_access_cost)
11002         (vect_peeling_supportable, vect_analyze_group_access_1)
11003         (vect_analyze_group_access, vect_analyze_data_ref_access)
11004         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
11005         (vect_compile_time_alias, vect_small_gap_p)
11006         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
11007         (vect_supportable_dr_alignment): Take references as dr_vec_infos
11008         instead of data_references.  Update calls to other routines for
11009         which the same change has been made.
11010         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
11011         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
11012         (vect_slp_analyze_and_verify_node_alignment)
11013         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
11014         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
11015         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
11016         above changes.
11017         (_vect_peel_info::dr): Replace with...
11018         (_vect_peel_info::dr_info): ...this new field.
11019         (vect_peeling_hash_get_most_frequent)
11020         (vect_peeling_hash_choose_best_peeling): Update accordingly.
11021         (vect_peeling_hash_get_lowest_cost):
11022         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
11023         routines for which the same change has been made.
11024         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
11025         data_reference.
11026         * tree-vect-loop-manip.c (get_misalign_in_elems)
11027         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
11028         above changes.
11029         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11030         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
11031         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
11032         (get_group_load_store_type, get_negative_load_store_type)
11033         (vect_get_data_ptr_increment, vectorizable_store)
11034         (vectorizable_load): Likewise.
11035         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
11036         Update calls to other routines for which the same change has been made.
11038 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11040         * tree-vectorizer.h (vec_info::move_dr): New member function.
11041         (dataref_aux): Rename to...
11042         (dr_vec_info): ...this and add "dr" and "stmt" fields.
11043         (_stmt_vec_info::dr_aux): Update accordingly.
11044         (_stmt_vec_info::data_ref_info): Delete.
11045         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
11046         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
11047         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
11048         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
11049         of data_ref.
11050         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
11051         (STMT_VINFO_DR_INFO): New macro.
11052         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
11053         (set_dr_misalignment): Update after rename of dataref_aux.
11054         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
11055         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
11056         initialization of STMT_VINFO_DATA_REF.
11057         * tree-vectorizer.c (vec_info::move_dr): New function.
11058         * tree-vect-patterns.c (vect_recog_bool_pattern)
11059         (vect_recog_mask_conversion_pattern)
11060         (vect_recog_gather_scatter_pattern): Use it.
11061         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
11062         the "dr" and "stmt" fields of dr_vec_info instead of
11063         STMT_VINFO_DATA_REF.
11065 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11067         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
11068         (is_pattern_stmt_p): Use it.
11069         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
11070         on pattern statements.
11072 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11074         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
11075         original stmt as a stmt_vec_info rather than a gimple stmt.
11076         (vect_pattern_recog_1): Take the statement directly as a
11077         stmt_vec_info, rather than via a gimple_stmt_iterator.
11078         Update call to vect_mark_pattern_stmts.
11079         (vect_pattern_recog): Update calls accordingly.
11081 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11083         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
11084         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
11085         a vect_def_type for the first argument.
11086         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
11087         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
11088         operand if it isn't defined by a vectorized statement.
11089         (vect_build_gather_load_calls): Remove the mask_dt argument and
11090         update calls to vect_get_vec_def_for_stmt_copy.
11091         (vectorizable_bswap): Likewise the dt argument.
11092         (vectorizable_call): Update calls to vectorizable_bswap and
11093         vect_get_vec_def_for_stmt_copy.
11094         (vectorizable_simd_clone_call, vectorizable_assignment)
11095         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
11096         (vectorizable_comparison): Update calls to
11097         vect_get_vec_def_for_stmt_copy.
11098         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
11099         vect_is_simple_use.
11100         (vect_get_loop_based_defs): Remove dt argument and update call
11101         to vect_get_vec_def_for_stmt_copy.
11102         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
11103         and vect_get_vec_def_for_stmt_copy.
11104         (vectorizable_load): Update calls to vect_build_gather_load_calls
11105         and vect_get_vec_def_for_stmt_copy.
11106         * tree-vect-loop.c (vect_create_epilog_for_reduction)
11107         (vectorizable_reduction, vectorizable_live_operation): Update calls
11108         to vect_get_vec_def_for_stmt_copy.
11110 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11112         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
11113         and gimple stmt arguments with a stmt_vec_info.
11114         (vect_record_base_alignments): Update calls accordingly.
11115         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
11116         and gimple stmt arguments with a stmt_vec_info.
11117         (vect_build_slp_tree_1): Remove vinfo argument and update call
11118         to vect_record_max_nunits.
11119         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
11120         and vect_record_max_nunits.
11122 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11124         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
11125         file and take a stmt_vec_info instead of a gimple stmt.
11126         (supportable_widening_operation, vect_finish_replace_stmt)
11127         (vect_finish_stmt_generation, vect_get_store_rhs)
11128         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
11129         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
11130         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
11131         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
11132         (vect_create_data_ref_ptr, bump_vector_ptr)
11133         (vect_permute_store_chain, vect_setup_realignment)
11134         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
11135         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
11136         (vectorizable_reduction, vectorizable_induction)
11137         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
11138         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
11139         than gimple stmts as arguments.
11140         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
11141         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
11142         (can_group_stmts_p, vect_check_gather_scatter)
11143         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
11144         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
11145         (vect_permute_load_chain, vect_shift_permute_load_chain)
11146         (vect_transform_grouped_load)
11147         (vect_record_grouped_load_vectors): Likewise.
11148         * tree-vect-loop.c (vect_fixup_reduc_chain)
11149         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
11150         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
11151         (vectorizable_reduction, vectorizable_induction)
11152         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
11153         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
11154         (vect_get_load_store_mask): Likewise.
11155         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
11156         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
11157         * tree-vect-stmts.c (vect_mark_relevant)
11158         (is_simple_and_all_uses_invariant)
11159         (exist_non_indexing_operands_for_use_p, process_use)
11160         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
11161         (vect_get_vec_def_for_operand, vect_get_vec_defs)
11162         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
11163         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
11164         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
11165         (get_negative_load_store_type, get_load_store_type)
11166         (vect_check_load_store_mask, vect_check_store_rhs)
11167         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
11168         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11169         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11170         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11171         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
11172         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
11173         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
11174         (supportable_widening_operation): Likewise.
11176 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11178         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
11179         a stmt_vec_info instead of a gcall.
11180         (vect_check_gather_scatter): Update call accordingly.
11181         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
11182         of a gphi.
11183         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
11184         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
11185         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
11186         instead of a gimple stmt.
11187         (vect_transform_loop): Update calls accordingly.
11188         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
11189         stmt_vec_infos instead of gimple stmts.
11190         (vect_analyze_slp_instance): Update use accordingly.
11191         * tree-vect-stmts.c (read_vector_array, write_vector_array)
11192         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
11193         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
11194         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
11195         (vect_gen_widened_results_half, vect_get_loop_based_defs)
11196         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
11197         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
11198         down to subroutines.
11200 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11202         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
11203         of the worklist from a vector of gimple stmts to a vector of
11204         stmt_vec_infos.
11205         * tree-vect-stmts.c (vect_mark_relevant, process_use)
11206         (vect_mark_stmts_to_be_vectorized): Likewise
11208 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11210         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
11211         statement before passing it to vect_analyze_stmt.
11212         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
11213         the chain of phi vector definitions.  Track the exit phi via its
11214         stmt_vec_info.
11215         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
11216         STMT_VINFO_REDUC_DEF.
11217         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
11218         stmt_vec_infos to handle the statement chains.
11219         (vect_get_slp_defs): Record the first statement in the node
11220         using a stmt_vec_info.
11221         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
11222         statements here and pass their stmt_vec_info down to subroutines.
11223         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
11224         down to vect_finish_stmt_generation.
11225         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
11226         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
11227         stmt_vec_infos to subroutines.
11228         (vect_remove_stores): Use stmt_vec_infos to handle the statement
11229         chains.
11231 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11233         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
11234         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
11235         (vect_permute_store_chain, vect_setup_realignment)
11236         (vect_permute_load_chain, vect_shift_permute_load_chain)
11237         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
11238         stmts internally, and when passing values to other vectorizer routines.
11239         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11240         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
11241         (vect_analyze_loop_operations, get_initial_def_for_reduction)
11242         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
11243         (vectorizable_reduction, vectorizable_induction)
11244         (vectorizable_live_operation, vect_transform_loop_stmt)
11245         (vect_transform_loop): Likewise.
11246         * tree-vect-patterns.c (vect_reassociating_reduction_p)
11247         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
11248         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
11249         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11250         (vect_slp_analyze_node_operations_1): Likewise.
11251         * tree-vect-stmts.c (vect_mark_relevant, process_use)
11252         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
11253         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
11254         (vect_finish_stmt_generation_1, get_group_load_store_type)
11255         (get_load_store_type, vect_build_gather_load_calls)
11256         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11257         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11258         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11259         (vectorizable_store, vectorizable_load, vectorizable_condition)
11260         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
11261         (supportable_widening_operation): Likewise.
11262         (vect_get_vector_types_for_stmt): Likewise.
11263         * tree-vectorizer.h (vect_dr_behavior): Likewise.
11265 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11267         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11268         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
11269         (vect_permute_store_chain, vect_permute_load_chain)
11270         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
11271         repeated stmt_vec_info lookups.
11272         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11273         (vect_update_ivs_after_vectorizer): Likewise.
11274         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
11275         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11276         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
11277         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11278         (vect_bb_slp_scalar_cost): Likewise.
11279         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
11281 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11283         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
11284         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
11285         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
11286         to get gassigns and gcalls, rather than operating on generc gimple
11287         stmts.
11288         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
11289         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
11290         (vectorizable_load, vect_analyze_stmt): Likewise.
11291         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
11293 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11295         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
11296         return stmt_vec_infos rather than gimple stmts.  Do not accept
11297         null arguments.
11298         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
11299         of a gimple stmt.
11300         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
11301         Update use of get_later_stmt.
11302         (vect_get_constant_vectors): Update call accordingly.
11303         (vect_schedule_slp_instance): Likewise
11304         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
11305         (vect_slp_analyze_instance_dependence): Likewise.
11306         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
11308 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11310         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
11311         (stmt_info_for_cost::stmt_info): ...this new field.
11312         (add_stmt_costs): Update accordingly.
11313         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11314         (vect_get_known_peeling_cost): Likewise.
11315         (vect_estimate_min_profitable_iters): Likewise.
11316         * tree-vect-stmts.c (record_stmt_cost): Likewise.
11318 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11320         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
11321         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11322         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
11323         accordingly.
11324         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
11326 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11328         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
11329         a gimple stmt to a stmt_vec_info.
11330         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
11332 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11334         * tree-vectorizer.h (vec_info::grouped_stores): Change from
11335         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11336         (_loop_vec_info::reduction_chains): Likewise.
11337         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
11338         accordingly.
11339         * tree-vect-slp.c (vect_analyze_slp): Likewise.
11341 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11343         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
11344         a gimple stmt to a stmt_vec_info.
11345         (_stmt_vec_info::next_element): Likewise.
11346         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
11347         (vect_slp_analyze_and_verify_node_alignment)
11348         (vect_analyze_group_access_1, vect_analyze_group_access)
11349         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
11350         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
11351         (vect_supportable_dr_alignment): Update accordingly.
11352         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
11353         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
11354         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
11355         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11356         * tree-vect-slp.c (vect_build_slp_tree_1)
11357         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
11358         (vect_split_slp_store_group, vect_analyze_slp_instance)
11359         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
11360         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
11361         (get_group_load_store_type, get_load_store_type)
11362         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
11363         (vect_transform_stmt, vect_remove_stores): Likewise.
11365 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11367         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
11368         than a gimple stmt.
11369         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11370         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
11371         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
11372         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
11373         (vector_alignment_reachable_p, vect_get_data_access_cost)
11374         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
11375         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
11376         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
11377         (vect_analyze_group_access_1, vect_analyze_group_access)
11378         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
11379         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
11380         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
11381         result of vect_dr_stmt and use the stmt_vec_info instead of
11382         the associated gimple stmt.
11383         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11384         (vect_gen_prolog_loop_niters): Likewise.
11385         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11387 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11389         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
11390         to a vec<stmt_vec_info>.
11391         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
11392         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
11393         vec<stmt_vec_info>.
11394         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
11395         to a vec<stmt_vec_info>.
11396         (bst_traits::value_type, bst_traits::value_type): Likewise.
11397         (bst_traits::hash): Update accordingly.
11398         (vect_get_and_check_slp_defs): Change the stmts parameter from
11399         a vec<gimple *> to a vec<stmt_vec_info>.
11400         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
11401         (vect_build_slp_tree): Likewise.
11402         (vect_build_slp_tree_2): Likewise.  Update uses of
11403         SLP_TREE_SCALAR_STMTS.
11404         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
11405         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
11406         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
11407         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
11408         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
11409         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
11410         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
11411         (vect_get_constant_vectors, vect_get_slp_defs)
11412         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
11413         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
11414         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
11415         instead of gimple stmts.
11416         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
11417         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
11418         (vect_slp_analyze_instance_dependence): Update uses of
11419         SLP_TREE_SCALAR_STMTS.
11420         (vect_slp_analyze_and_verify_node_alignment): Likewise.
11421         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
11422         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
11423         (get_initial_defs_for_reduction): Likewise.
11424         (vect_create_epilog_for_reduction): Likewise.
11425         (vectorize_fold_left_reduction): Likewise.
11426         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
11427         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
11428         (can_vectorize_live_stmts): Likewise.
11430 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11432         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
11433         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11434         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
11435         than gimple stmts.
11436         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
11437         of a gimple stmt.
11438         (gather_scalar_reductions): Update after above interface changes.
11439         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
11440         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
11441         than gimple stmts.
11442         (vect_force_simple_reduction): Likewise.
11443         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
11444         LOOP_VINFO_REDUCTIONS.
11445         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11447 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11449         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
11450         a gimple stmt to a stmt_vec_info.
11451         * tree-vect-loop.c (vect_active_double_reduction_p)
11452         (vect_force_simple_reduction, vectorizable_reduction): Update
11453         accordingly.
11455 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11457         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
11458         vec<gimple *> to a vec<stmt_vec_info>.
11459         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
11460         the reduction_phis argument from a vec<gimple *> to a
11461         vec<stmt_vec_info>.
11462         (vectorizable_reduction): Likewise the phis local variable that
11463         is passed to vect_create_epilog_for_reduction.  Update for new type
11464         of SLP_TREE_VEC_STMTS.
11465         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
11466         (vectorizable_live_operation): Likewise.
11467         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
11468         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
11470 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11472         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
11473         a gimple stmt to a stmt_vec_info.
11474         (vectorizable_condition, vectorizable_live_operation)
11475         (vectorizable_reduction, vectorizable_induction): Pass back the
11476         vectorized statement as a stmt_vec_info.
11477         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
11478         use of STMT_VINFO_VEC_STMT.
11479         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
11480         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
11481         as stmt_vec_infos rather than gimple stmts.
11482         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
11483         to a stmt_vec_info.
11484         (vectorizable_live_operation): Likewise.
11485         (vectorizable_reduction, vectorizable_induction): Likewise,
11486         updating use of STMT_VINFO_VEC_STMT.
11487         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
11488         of STMT_VINFO_VEC_STMT.
11489         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
11490         (vectorizable_simd_clone_call, vectorizable_conversion)
11491         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11492         (vectorizable_store, vectorizable_load, vectorizable_condition)
11493         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
11494         from a gimple stmt to a stmt_vec_info.
11495         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
11496         pointer to a stmt_vec_info to the vectorizable_* routines.
11498 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11500         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
11501         a gimple stmt to a stmt_vec_info.
11502         (is_pattern_stmt_p): Update accordingly.
11503         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
11504         (vect_record_grouped_load_vectors): Likewise.
11505         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
11506         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
11507         (vect_model_reduction_cost): Likewise.
11508         (vect_create_epilog_for_reduction): Likewise.
11509         (vectorizable_reduction, vectorizable_induction): Likewise.
11510         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11511         Return the stmt_vec_info for the pattern statement.
11512         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
11513         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
11514         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11515         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
11516         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
11517         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
11518         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
11519         (free_stmt_vec_info, vect_is_simple_use): Likewise.
11521 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11523         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
11524         (vect_finish_stmt_generation): Likewise.
11525         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11526         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
11527         (vect_build_gather_load_calls): Use the return value of the above
11528         functions instead of a separate call to vinfo_for_stmt.  Use narrow
11529         scopes for the input gimple stmt and wider scopes for the associated
11530         stmt_vec_info.  Use vec_info::lookup_def when setting these
11531         stmt_vec_infos from an SSA_NAME definition.
11532         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11533         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11534         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11535         (vectorizable_store, vectorizable_load, vectorizable_condition)
11536         (vectorizable_comparison): Likewise.
11537         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
11538         (vectorizable_reduction): Likewise.
11540 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11542         * tree-vectorizer.h (vect_is_simple_use): Add an optional
11543         stmt_vec_info * parameter before the optional gimple **.
11544         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11545         (process_use, vect_get_vec_def_for_operand_1): Update callers.
11546         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
11547         * tree-vect-loop.c (vectorizable_reduction): Likewise.
11548         (vectorizable_live_operation): Likewise.
11549         * tree-vect-patterns.c (type_conversion_p): Likewise.
11550         (vect_look_through_possible_promotion): Likewise.
11551         (vect_recog_rotate_pattern): Likewise.
11552         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11554 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11556         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
11557         a typedef to a wrapper class.
11558         (NULL_STMT_VEC_INFO): New macro.
11559         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
11560         (stmt_vec_info::operator*): New function.
11561         (stmt_vec_info::operator gimple *): Likewise.
11562         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
11563         (add_stmt_costs): Likewise.
11564         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
11565         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11566         (vect_get_known_peeling_cost): Likewise.
11567         (vect_estimate_min_profitable_iters): Likewise.
11568         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11569         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
11570         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
11571         (vectorizable_store, free_stmt_vec_infos): Likewise.
11572         (new_stmt_vec_info): Change return type of xcalloc to
11573         _stmt_vec_info *.
11575 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11577         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
11578         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
11579         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
11580         a single_imm_use-based sequence.
11581         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
11583 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11585         * tree-vectorizer.h (vec_info::lookup_def): Declare.
11586         * tree-vectorizer.c (vec_info::lookup_def): New function.
11587         * tree-vect-patterns.c (vect_get_internal_def): Use it.
11588         (vect_widened_op_tree): Likewise.
11589         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11590         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
11591         (vectorizable_reduction): Likewise.
11592         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
11593         of a gimple *.
11594         (vect_is_slp_reduction): Update calls accordingly.  Use
11595         vec_info::lookup_def.
11596         (vect_is_simple_reduction): Likewise
11597         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
11599 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11601         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
11602         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
11603         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
11604         of vinfo_for_stmt.
11605         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
11606         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
11607         (vect_update_vf_for_slp, vect_analyze_loop_operations)
11608         (vect_is_slp_reduction, vectorizable_induction)
11609         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
11610         * tree-vect-patterns.c (vect_init_pattern_stmt):
11611         (vect_determine_min_output_precision_1, vect_determine_precisions)
11612         (vect_pattern_recog): Likewise.
11613         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
11614         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
11615         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
11616         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11617         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
11618         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
11619         info field from a loop to a loop_vec_info.
11621 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11623         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
11624         (vec_info::add_stmt): Declare.
11625         * tree-vectorizer.c (vec_info::add_stmt): New function.
11626         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
11627         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
11628         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11629         (vectorizable_induction): Likewise.
11630         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
11631         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11632         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
11633         (vectorizable_load): Likewise.
11634         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11635         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
11636         (vect_recog_gather_scatter_pattern): Likewise.
11637         (append_pattern_def_seq): Likewise.  Remove a check that is
11638         performed by add_stmt itself.
11640 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11642         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
11643         which make_ssa_name was called with new_stmt before new_stmt
11644         had been created.
11646 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11648         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
11649         split out from...
11650         (vect_is_slp_reduction): ...here...
11651         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
11652         that are already known to be false.
11654 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11656         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
11657         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
11658         STMT_VINFO_NUM_SLP_USES when it's true.
11659         (vect_free_slp_instance): Add a final_p parameter and pass it to
11660         vect_free_slp_tree.
11661         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
11662         (vect_analyze_slp_instance): Likewise.
11663         (vect_slp_analyze_operations): Likewise.
11664         (vect_slp_analyze_bb_1): Likewise.
11665         * tree-vectorizer.c (vec_info): Likewise.
11666         * tree-vect-loop.c (vect_transform_loop): Likewise.
11668 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11670         * tree-vect-loop.c (vectorizable_reduction): Assert that the
11671         function is not called for second and subsequent members of
11672         a reduction group.
11674 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11676         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
11677         cases for nested loops from here to ...
11678         (vect_create_epilog_for_reduction): ...here.  Only call
11679         vect_is_simple_use for inner-loop reductions.
11681 2018-07-31  Martin Liska  <mliska@suse.cz>
11683         PR gcov-profile/85338
11684         PR gcov-profile/85350
11685         PR gcov-profile/85372
11686         * profile.c (struct location_triplet): New.
11687         (struct location_triplet_hash): Likewise.
11688         (output_location): Do not output a BB that
11689         is already recorded for a line.
11690         (branch_prob): Use streamed_locations.
11692 2018-07-31  Martin Liska  <mliska@suse.cz>
11694         PR gcov-profile/85370
11695         * coverage.c (coverage_begin_function): Do not mark target
11696         clones as artificial functions.
11698 2018-07-31  Martin Liska  <mliska@suse.cz>
11700         PR gcov-profile/83813
11701         PR gcov-profile/84758
11702         PR gcov-profile/85217
11703         PR gcov-profile/85332
11704         * profile.c (branch_prob): Do not record GOTO expressions
11705         for GIMPLE statements which locations are already streamed.
11707 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11709         * gcc.c (handle_spec_function): Accept a soft_matched_part
11710         argument, as do_spec_1.  Pass it down to ...
11711         (eval_spec_function): Accept a soft_matched_part argument,
11712         and pass it down to ...
11713         (do_spec_2): Accept a soft_matched_part argument, and pass
11714         it down to do_spec_1.
11715         (do_spec_1): Pass soft_matched_part to handle_spec_function.
11716         (handle_braces): Update call to handle_spec_function.
11717         (driver::set_up_specs): Update calls to do_spec_2.
11718         (compare_debug_dump_opt_spec_function): Likewise.
11719         (compare_debug_self_opt_spec_function): Likewise.
11721 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11723         * common.opt (nolibc): New option.
11724         * doc/invoke.texi (Link Options): Document it.
11725         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
11726         * config/alpha/linux.h: Likewise.
11727         * config/arc/elf.h: Likewise.
11728         * config/arm/uclinux-elf.h: Likewise.
11729         * config/arm/unknown-elf.h: Likewise.
11730         * config/avr/avrlibc.h: Likewise.
11731         * config/bfin/bfin.h: Likewise.
11732         * config/bfin/linux.h: Likewise.
11733         * config/bfin/uclinux.h: Likewise.
11734         * config/darwin.h: Likewise.
11735         * config/darwin10.h: Likewise.
11736         * config/darwin12.h: Likewise.
11737         * config/gnu-user.h: Likewise.
11738         * config/lm32/uclinux-elf.h: Likewise.
11739         * config/pa/pa-hpux11.h: Likewise.
11740         * config/pa/pa64-hpux.h: Likewise.
11741         * config/sparc/sparc.h: Likewise.
11743 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11745         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
11746         undefined variables.
11748 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11750         PR target/86640
11751         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
11752         instead of GEN_INT.
11754 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11756         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
11757         terminated string literal.
11759 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11761         PR rtl-optimization/85160
11762         * combine.c (is_just_move): New function.
11763         (try_combine): Allow combining two instructions into two if neither of
11764         the original instructions was a move.
11766 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
11768         PR target/86673
11769         * doc/extend.texi (Global Register Variables): Discourage use of type
11770         qualifiers.
11771         (Local Register Variables): Likewise.
11773 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
11775         PR tree-optimization/86506
11776         * hwint.h (ceil_log2): Resync with hwint.c implementation.
11778 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
11780         PR target/86547
11781         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
11782         hard_regno, make sure no insn between `from` and `to` clobbers it.
11784 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
11785             Tom de Vries  <tdevries@suse.de>
11787         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
11788         (PTX_DEFAULT_RUNTIME_DIM): ... this.
11789         (nvptx_goacc_validate_dims): Set default worker and gang dims to
11790         PTX_DEFAULT_RUNTIME_DIM.
11791         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
11793 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
11795         * config/pa/pa.c (pa_output_addr_vec): Align address table.
11796         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
11797         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
11799 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
11801         * config/rs6000/constraints.md (wG constraint): Delete, no longer
11802         used.
11803         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
11804         predicate to reflect toc fusion has been deleted.
11805         (toc_fusion_mem_raw): Delete, no longer used.
11806         (toc_fusion_mem_wrapped): Likewise.
11807         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
11808         fusion mask bit.
11809         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
11810         Delete, no longer used.
11811         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
11812         meant to be used for toc fusion.
11813         (rs6000_debug_print_mode): Delete toc fusion debugging.
11814         (rs6000_debug_reg_global): Likewise.
11815         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
11816         fusion and secondary reload support that were never used.
11817         (rs6000_option_override_internal): Delete TOC fusion, that was only
11818         partially defined, and it did not work unless you also used the
11819         -mcmodel= switch.
11820         (rs6000_legitimate_address_p): Delete TOC fusion support.
11821         (rs6000_opt_masks): Likewise.
11822         (fusion_wrap_memory_address): Delete function, no longer used.
11823         (fusion_split_address); Delete TOC fusion support.
11824         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
11825         longer used with toc fusion being deleted.
11826         (TARGET_TOC_FUSION_FP): Likewise.
11827         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
11828         UNSPEC.
11829         (toc fusion spliter): Delete TOC fusion support.
11830         (toc_fusionload_<mode>): Likewise.
11831         (toc_fusionload_di): Likewise.
11832         (fusion_gpr_load_<mode>): Delete generator function, this insn no
11833         longer needs to be named.  Rename predicate to delete TOC fusion.
11834         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11835         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11836         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11837         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11838         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
11840 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11842         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
11843         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
11844         __int128 in built-in function prototypes.
11845         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
11846         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
11848 2018-07-27  Martin Sebor  <msebor@redhat.com>
11850         PR tree-optimization/86696
11851         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
11852         types, including enums.
11853         (handle_char_store): Be prepared for the above function to fail.
11855 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
11857         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
11858         when optimization level is lower than 2 or optimize for size.
11860 2018-07-26  Martin Sebor  <msebor@redhat.com>
11862         PR tree-optimization/86043
11863         PR tree-optimization/86042
11864         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
11865         (get_string_cst_length): Rename...
11866         (get_min_string_length): ...to this.  Add argument.
11867         (handle_char_store): Extend to handle multi-character stores by
11868         MEM_REF.
11869         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
11870         * tree.h (initializer_zerop): Add argument.
11872 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
11874         PR middle-end/86660
11875         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
11876         declare target to variables if they have always,{to,from,tofrom} map
11877         kinds.
11879 2018-07-26  Martin Liska  <mliska@suse.cz>
11881         PR lto/86548
11882         * lto-wrapper.c: Add linker_output as prefix
11883         for ltrans_output_file.
11885 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
11887         PR rtl-optimization/85805
11888         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
11889         value for hard registers if that was written in the same mode.
11891 2018-07-26  Martin Liska  <mliska@suse.cz>
11893         PR gcov-profile/86536
11894         * gcov.c (format_gcov): Use printf format %.*f directly
11895         and do not handle special values.
11897 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11899         * common/config/arc/arc-common.c (arc_option_optimization_table):
11900         Update default optimizations for size.
11902 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11904         * config/arc/arc.md (movsf_insn): Add short instruction selection.
11905         * config/arc/constraints.md (CfZ): New constraint.
11906         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
11907         (subsf3_fpu): Likewise.
11908         (cmpsf_fpu): Likewise.
11909         (cmpsf_fpu_uneq): Likewise.
11911 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11913         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
11914         accesses as well.
11915         (arc_is_uncached_mem_p): uncached applies to both the variable and
11916         the pointer.
11918 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11920         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
11921         register names.
11923 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
11925         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
11926         field "m_scopes" from vec to auto_vec.
11928 2018-07-25  Martin Liska  <mliska@suse.cz>
11930         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
11931         return type.
11933 2018-07-25  Richard Biener  <rguenther@suse.de>
11935         PR debug/86654
11936         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
11937         special wrt context_die late.
11938         (gen_subprogram_die): Re-use DIEs in local scope.
11940 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
11942         PR tree-optimization/86644
11943         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
11945 2018-07-25  Martin Liska  <mliska@suse.cz>
11947         PR middle-end/86645
11948         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
11949         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
11951 2018-07-25  Martin Liska  <mliska@suse.cz>
11953         PR sanitizer/79635
11954         * params.def: Explain ASan abbreviation and provide
11955         a documentation link.
11957 2018-07-24  Martin Sebor  <msebor@redhat.com>
11959         PR tree-optimization/86622
11960         PR tree-optimization/86532
11961         * builtins.h (string_length): Declare.
11962         * builtins.c (c_strlen): Correct handling of non-constant offsets.
11963         (check_access): Be prepared for non-constant length ranges.
11964         (string_length): Make extern.
11965         * expr.c (string_constant): Only handle the minor non-constant
11966         array index.  Use string_constant to compute the length of
11967         a generic string constant.
11969 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
11971         PR tree-optimization/86618
11972         * tree-vect-stmts.c (vectorizable_call): Don't take the address
11973         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
11975 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
11977         PR tree-optimization/86636
11978         * json.cc (json::object::set): Fix comment.  Add assertions.
11979         (json::array::append): Move here from json.h.  Add comment and an
11980         assertion.
11981         (json::string::string): Likewise.
11982         * json.h (json::array::append): Move to json.cc.
11983         (json::string::string): Likewise.
11984         * optinfo-emit-json.cc
11985         (optrecord_json_writer::impl_location_to_json): Assert that we
11986         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
11987         wrapper around it.  Expand the location once, rather than three
11988         times.
11989         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
11990         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
11991         wrappers.
11992         (optrecord_json_writer::optinfo_to_json): Likewise, in four
11993         places.  Fix some overlong lines.
11995 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
11997         * config/aarch64/aarch64-simd.md
11998         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
11999         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
12000         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
12001         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
12002         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
12003         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
12004         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
12005         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
12006         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
12008 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
12010         PR middle-end/86627
12011         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
12012         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
12013         and abs_d == d, do the power of two handling if profitable.
12015 2018-07-24  Richard Biener  <rguenther@suse.de>
12017         * match.pd: Add BIT_FIELD_REF canonicalizations.
12019 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12021         PR c/86617
12022         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
12024 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12026         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
12027         terminated STRING_CST object.
12029 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12031         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
12033 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
12035         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
12036         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
12037         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
12038         the elements into a register.
12039         (rs6000_split_v4si_init_di_reg): Delete.
12040         (rs6000_split_v4si_init): Delete.
12041         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
12042         (vsx_init_v4si): Rewrite as a define_expand.
12044 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
12046         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
12047         zero_extend argument from memory): New.
12049 2018-07-22  Martin Sebor  <msebor@redhat.com>
12051         PR bootstrap/86621
12052         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
12053         diagnosing calls with unknown arguments unless -Walloca-larger-than
12054         is restricted to less than PTRDIFF_MAX bytes.
12056 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
12058         * doc/gcov.texi (Invoking Gcov): Editorial changes.
12060 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
12062         * pretty-print.c (text_info::set_location): Remove redundant
12063         "line_table" parameter from call to rich_location::set_range.
12065 2018-07-20  Martin Sebor  <msebor@redhat.com>
12067         PR middle-end/82063
12068         * builtins.c (expand_builtin_alloca): Adjust.
12069         * calls.c (alloc_max_size): Simplify.
12070         * cgraphunit.c (cgraph_node::expand): Adjust.
12071         * common.opt (larger_than_size, warn_frame_larger_than): Remove
12072         variables.
12073         (frame_larger_than_size): Same.
12074         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
12075         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
12076         Initialize.
12077         * doc/invoke.texi (GCC Command Options): Document option arguments.
12078         Explain byte-size arguments and suffixes.
12079         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
12080         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
12081         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
12082         * doc/options.texi (UInteger): Expand.
12083         (Host_Wide_Int, ByteSize): Document new properties.
12084         * final.c (final_start_function_1): Include sizes in an error message.
12085         * function.c (frame_offset_overflow): Same.
12086         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
12087         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
12088         Diagnose unbounded alloca calls only for limits of less than
12089         PTRDIFF_MAX.
12090         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
12091         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
12092         for alloca(0).
12093         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
12094         only for limits of less than PTRDIFF_MAX.
12095         * langhooks-def.h (lhd_handle_option): Change function argument
12096         to HOST_WIDE_INT.
12097         * langhooks.c (lhd_handle_option): Same.
12098         * langhooks.h (handle_option): Same.
12099         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
12100         ByteSize flags.
12101         (var_type, var_type_struct): Same.
12102         (var_set): Handle ByteSize flag.
12103         * optc-gen.awk: Add comments to output to ease debugging.  Make
12104         use of HOST_WIDE_INT where appropriate.
12105         * opts-gen-save.awk:  Use %lx to format unsigned long.
12106         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
12107         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
12108         arguments.  Parse bytes-size suffixes.
12109         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
12110         (enum_value_to_arg): Same.
12111         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
12112         (handle_option): Adjust.
12113         (generate_option): Change function argument to HOST_WIDE_INT.
12114         (cmdline_handle_error): Adjust.
12115         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
12116         (set_option): Change function argument to HOST_WIDE_INT.
12117         (option_enabled): Handle cl_host_wide_int.
12118         (get_option_state): Handle CLVC_SIZE.
12119         (control_warning_option): Same.
12120         * opts.c (common_handle_option): Change function argument to
12121         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
12122         OPT_Wvla_larger_than_.
12123         * opts.h (enum cl_var_type): Add an enumerator.
12124         * stor-layout.c (layout_decl): Print a more meaningful warning.
12125         * toplev.c (output_stack_usage): Adjust.
12127 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
12129         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
12130         call to inline_expand_builtin_string_cmp.
12131         (expand_builtin_strcmp): Likewise.
12132         (expand_builtin_strncmp): Likewise.
12133         (inline_string_cmp): Delete the last parameter, change char_type_node
12134         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
12135         two operands.
12136         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
12137         the inlining expansion on target where the type of the call has same or
12138         narrower precision than unsigned char.
12140 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
12142         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
12143         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
12144         * common.opt (fsave-optimization-record): New option.
12145         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
12146         * doc/invoke.texi (-fsave-optimization-record): New option.
12147         * dumpfile.c: Include "optinfo-emit-json.h".
12148         (struct kv_pair): Move to coretypes.h.
12149         (optgroup_options): Make non-static.
12150         (dump_context::end_scope): Call
12151         optimization_records_maybe_pop_dump_scope.
12152         * dumpfile.h (optgroup_options): New decl.
12153         * json.cc: New file.
12154         * json.h: New file.
12155         * optinfo-emit-json.cc: New file.
12156         * optinfo-emit-json.h: New file.
12157         * optinfo.cc: Include "optinfo-emit-json.h".
12158         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
12159         (optinfo_enabled_p): Check optimization_records_enabled_p.
12160         (optinfo_wants_inlining_info_p): Likewise.
12161         * optinfo.h: Update comment.
12162         * profile-count.c (profile_quality_as_string): New function.
12163         * profile-count.h (profile_quality_as_string): New decl.
12164         (profile_count::quality): New accessor.
12165         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
12166         and optinfo_emit_json_cc_tests.
12167         * selftest.h (selftest::json_cc_tests): New decl.
12168         (selftest::optinfo_emit_json_cc_tests): New decl.
12169         * toplev.c: Include "optinfo-emit-json.h".
12170         (compile_file): Call optimization_records_finish.
12171         (do_compile): Call optimization_records_start.
12172         * tree-ssa-live.c: Include optinfo.h.
12173         (remove_unused_scope_block_p): Retain inlining information if
12174         optinfo_wants_inlining_info_p returns true.
12176 2018-07-20  Richard Biener  <rguenther@suse.de>
12178         PR debug/86585
12179         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
12180         to cover -flto-partition=none.
12182 2018-07-20  Martin Liska  <mliska@suse.cz>
12184         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
12185         (get_decl_source_range): Remove unused function.
12187 2018-07-20  Richard Biener  <rguenther@suse.de>
12189         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
12190         (struct vn_phi_s): Likewise.
12191         (struct vn_reference_s): Likewise.
12192         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
12193         for searching the slot of an entry known to be in the hash itself.
12194         (vn_phi_hasher::equal): Likewise.
12195         (vn_reference_hasher::equal): Likewise.
12196         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
12197         globals.
12198         (optimistic_info, current_info): Remove, keeping only valid_info.
12199         (vn_reference_lookup_1): Remove fallback lookup.
12200         (vn_reference_lookup_2): Likewise.
12201         (vn_nary_op_lookup_1): Likewise.
12202         (vn_phi_lookup): Likewise.
12203         (vn_nary_build_or_lookup_1): Make sure to not chain the built
12204         hash element.
12205         (vn_reference_insert): Adjust, chain the inserted hash element
12206         at last_inserted_ref.
12207         (vn_reference_insert_pieces): Likewise.
12208         (visit_reference_op_call): Likewise.
12209         (vn_nary_op_insert_into): Chain the inserted hash element at
12210         last_inserted_nary.
12211         (vn_nary_op_insert_pieces): Adjust.
12212         (vn_nary_op_insert): Likewise.
12213         (vn_nary_op_insert_stmt): Likewise.
12214         (vn_phi_insert): Adjust, chain the inserted hash element at
12215         last_inserted_phi.
12216         (process_scc): Remove clearing and copying the optimistic
12217         table.  Instead remove elements inserted during an optimistic
12218         iteration from the single table we maintain.
12219         (init_scc_vn): Adjust.
12220         (free_scc_vn): Likewise.
12221         (sccvn_dom_walker::record_cond): Likewise.
12222         (sccvn_dom_walker::after_dom_children): Likewise.
12224 2018-07-19  Martin Sebor  <msebor@redhat.com>
12226         PR tree-optimization/84047
12227         PR tree-optimization/83776
12228         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
12229         (check_array_bounds): Call it.
12231 2018-07-19  Martin Sebor  <msebor@redhat.com>
12233         * align.h (align_flags): Use member initialization.
12235 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
12237         * Makefile.in (OBJS): Add optinfo.o.
12238         * coretypes.h (class symtab_node): New forward decl.
12239         (struct cgraph_node): New forward decl.
12240         (class varpool_node): New forward decl.
12241         * dump-context.h: New file.
12242         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
12243         "tree-pass.h".
12244         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
12245         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
12246         (set_alt_dump_file): Likewise.
12247         (dump_context::~dump_context): New dtor.
12248         (dump_gimple_stmt): Move implementation to...
12249         (dump_context::dump_gimple_stmt): ...this new member function.
12250         Add the stmt to any pending optinfo, creating one if need be.
12251         (dump_gimple_stmt_loc): Move implementation to...
12252         (dump_context::dump_gimple_stmt_loc): ...this new member function.
12253         Start a new optinfo and add the stmt to it.
12254         (dump_gimple_expr): Move implementation to...
12255         (dump_context::dump_gimple_expr): ...this new member function.
12256         Add the stmt to any pending optinfo, creating one if need be.
12257         (dump_gimple_expr_loc): Move implementation to...
12258         (dump_context::dump_gimple_expr_loc): ...this new member function.
12259         Start a new optinfo and add the stmt to it.
12260         (dump_generic_expr): Move implementation to...
12261         (dump_context::dump_generic_expr): ...this new member function.
12262         Add the tree to any pending optinfo, creating one if need be.
12263         (dump_generic_expr_loc): Move implementation to...
12264         (dump_context::dump_generic_expr_loc): ...this new member
12265         function.  Add the tree to any pending optinfo, creating one if
12266         need be.
12267         (dump_printf): Move implementation to...
12268         (dump_context::dump_printf_va): ...this new member function.  Add
12269         the text to any pending optinfo, creating one if need be.
12270         (dump_printf_loc): Move implementation to...
12271         (dump_context::dump_printf_loc_va): ...this new member function.
12272         Start a new optinfo and add the stmt to it.
12273         (dump_dec): Move implementation to...
12274         (dump_context::dump_dec): ...this new member function.  Add the
12275         value to any pending optinfo, creating one if need be.
12276         (dump_context::dump_symtab_node): New member function.
12277         (dump_context::get_scope_depth): New member function.
12278         (dump_context::begin_scope): New member function.
12279         (dump_context::end_scope): New member function.
12280         (dump_context::ensure_pending_optinfo): New member function.
12281         (dump_context::begin_next_optinfo): New member function.
12282         (dump_context::end_any_optinfo): New member function.
12283         (dump_context::s_current): New global.
12284         (dump_context::s_default): New global.
12285         (dump_scope_depth): Delete global.
12286         (dumpfile_ensure_any_optinfo_are_flushed): New function.
12287         (dump_symtab_node): New function.
12288         (get_dump_scope_depth): Reimplement in terms of dump_context.
12289         (dump_begin_scope): Likewise.
12290         (dump_end_scope): Likewise.
12291         (selftest::temp_dump_context::temp_dump_context): New ctor.
12292         (selftest::temp_dump_context::~temp_dump_context): New dtor.
12293         (selftest::verify_item): New function.
12294         (ASSERT_IS_TEXT): New macro.
12295         (ASSERT_IS_TREE): New macro.
12296         (ASSERT_IS_GIMPLE): New macro.
12297         (selftest::test_capture_of_dump_calls): New test.
12298         (selftest::dumpfile_c_tests): Call it.
12299         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
12300         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
12301         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
12302         descriptive comment.
12303         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
12304         (dump_node, dump_bb): Move these unrelated decls.
12305         (class dump_manager): Add leading comment.
12306         * optinfo.cc: New file.
12307         * optinfo.h: New file.
12309 2018-07-19  Michael Collison  <michael.collison@arm.com>
12310             Richard Henderson <rth@redhat.com>
12312         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
12313         (subti): Handle op1 zero.
12314         (subvti4, usub4ti4): New.
12315         (*sub<GPI>3_compare1_imm): New.
12316         (sub<GPI>3_carryinCV): New.
12317         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
12318         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
12320 2018-07-19  Michael Collison  <michael.collison@arm.com>
12321             Richard Henderson <rth@redhat.com>
12323         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
12324         (addti3): Create simpler code if low part is already known to be 0.
12325         (addvti4, uaddvti4): New.
12326         (*add<GPI>3_compareC_cconly_imm): New.
12327         (*add<GPI>3_compareC_cconly): New.
12328         (*add<GPI>3_compareC_imm): New.
12329         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
12330         handle constants within this pattern..
12331         (*add<GPI>3_compareV_cconly_imm): New.
12332         (*add<GPI>3_compareV_cconly): New.
12333         (*add<GPI>3_compareV_imm): New.
12334         (add<GPI>3_compareV): New.
12335         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
12336         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
12337         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
12338         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
12339         with 'comparison' operator.
12340         (*add<GPI>3_compareV_cconly_imm): Ditto.
12341         (*add<GPI>3_compareV_cconly): Ditto.
12342         (*add<GPI>3_compareV_imm): Ditto.
12343         (add<GPI>3_compareV): Ditto.
12344         (add<mode>3_carryinC): Ditto.
12345         (*add<mode>3_carryinC_zero): Ditto.
12346         (*add<mode>3_carryinC): Ditto.
12347         (add<mode>3_carryinV): Ditto.
12348         (*add<mode>3_carryinV_zero): Ditto.
12349         (*add<mode>3_carryinV): Ditto.
12351 2018-07-19  Michael Collison  <michael.collison@arm.com>
12352             Richard Henderson <rth@redhat.com>
12354         * config/aarch64/aarch64-modes.def (CC_V): New.
12355         * config/aarch64/aarch64-protos.h
12356         (aarch64_addti_scratch_regs): Declare
12357         (aarch64_subvti_scratch_regs): Declare.
12358         (aarch64_expand_subvti): Declare.
12359         (aarch64_gen_unlikely_cbranch): Declare
12360         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
12361         for signed overflow using CC_Vmode.
12362         (aarch64_get_condition_code_1): Handle CC_Vmode.
12363         (aarch64_gen_unlikely_cbranch): New function.
12364         (aarch64_addti_scratch_regs): New function.
12365         (aarch64_subvti_scratch_regs): New function.
12366         (aarch64_expand_subvti): New function.
12368 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12370         * config/aarch64/aarch64-option-extensions.def: New entry for profile
12371         extension.
12372         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
12373         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
12374         extension.
12376 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12378         PR target/83009
12379         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12380         address check not strict.
12382 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12384         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
12385         Umq with Umn.
12386         (store_pair_lanes<mode>): Likewise.
12387         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
12388         enum value 'ADDR_QUERY_LDP_STP_N'.
12389         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
12390         (aarch64_print_address_internal): Add declaration.
12391         (aarch64_print_ldpstp_address): Remove.
12392         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
12393         (aarch64_print_operand): Change printing of 'y'.
12394         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
12395         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
12396         'true' rather than '1'.
12397         * config/aarch64/constraints.md (Uml): Likewise.
12398         (Uml): Rename to Umn.
12399         (Umq): Remove.
12401 2018-07-19  Richard Biener  <rguenther@suse.de>
12403         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
12404         a trailing array.
12405         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
12406         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
12407         (vn_reference_hasher): Likewise.
12408         (struct vn_tables_s): Remove obstack and alloc-pool members.
12409         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
12410         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
12411         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
12412         (vn_reference_insert_pieces): Likewise.
12413         (alloc_vn_nary_op_noinit): Adjust.
12414         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
12415         (vn_phi_eq): Adjust.
12416         (shared_lookup_phiargs): Remove.
12417         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
12418         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
12419         (visit_reference_op_call): Likewise.
12420         (copy_nary, copy_phi, copy_reference): Remove.
12421         (process_scc): Rewind the obstack when iterating.  Do not
12422         copy the elements to valid_info but just move them from one
12423         hashtable to the other.
12424         (allocate_vn_table): Adjust.
12425         (free_vn_table): Likewise.
12426         (init_scc_vn): Likewise.
12427         (free_scc_vn): Likewise.
12429 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
12431         PR target/86560
12432         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
12433         indirect_return as function type attribute.
12434         (ix86_attribute_table): Change indirect_return to function
12435         type attribute.
12436         * doc/extend.texi: Update indirect_return attribute.
12438 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
12440         * wide-int.h (widest2_int): New.
12441         * gimple-fold.c (arith_overflowed_p): Use it.
12442         * tree.h (widest2_int_cst): New.
12443         * tree-vrp.c (wide_int_binop_overflow): Rename from
12444         vrp_int_const_binop.
12445         Rewrite to work on trees.
12446         (extract_range_from_multiplicative_op_1): Abstract code to...
12447         (wide_int_range_min_max): ...here.
12448         (wide_int_range_cross_product): ...and here.
12449         (extract_range_from_binary_expr_1): Abstract overflow code to...
12450         (wide_int_range_mult_wrapping): ...here.
12451         * tree-vrp.h (wide_int_range_cross_product): New.
12452         (wide_int_range_mult_wrapping): New.
12454 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
12455             Julia Koval  <julia.koval@intel.com>
12457         * config/i386/x86-tune-costs.h (skylake_memcpy,
12458         skylake_memset): Replace rep_prefix with unrolling for size 512.
12460 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
12462         PR middle-end/86544
12463         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
12464         comparision with EQ_EXPR in last stmt.
12466 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12468         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
12469         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
12470         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
12471         previously known as "PowerPC AltiVec Built-in Functions".  Move
12472         some material to new subsubsections "PowerPC AltiVec Built-in
12473         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
12474         ISA 2.07".
12475         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
12476         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
12477         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
12478         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
12480 2018-07-18  Richard Biener  <rguenther@suse.de>
12482         PR tree-optimization/86557
12483         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
12484         EXACT_DIV_EXPR.
12486 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
12488         * config/s390/s390.c (s390_function_profiler): Generate CFI.
12490 2018-07-17  Jeff Law  <law@redhat.com>
12492         * config/arm/arm.c (get_label_padding): Update for recent
12493         changes to label_to_alignment.
12495         PR tree-optimization/86010
12496         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
12498         * config/mips/mips.c (vr4130_align_insns): Update for recent
12499         changes to label_to_alignment.
12501         * config/frv/frv.c (frv_label_align): Update for recent changes
12502         to label_to_alignment.
12504         * config/nios2/nios2.c (nios2_label_align): Update for recent
12505         changes which dropped ALIGN_LABELS_LOG.
12507 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
12509         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
12510         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
12512 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
12514         * config/arc/arc.c (arc_label_align): Use align_labels instead of
12515         deprecated align_labels_log.
12517 2018-07-17  Richard Biener  <rguenther@suse.de>
12519         PR lto/86456
12520         * dwarf2out.c (init_sections_and_labels): Always generate
12521         a debug_line_str_section for early LTO debug.
12522         (dwarf2out_finish): Reset debug_line_str_hash output early.
12523         Bump counter for extra dwarf5 .debug_loc labels to not conflict
12524         with fat LTO part.
12525         (dwarf2out_early_finish): Output debug_line_str.
12527 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12529         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
12530         index register on z196 or later.
12532 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12534         * config/s390/s390.c (s390_default_align): Set default function
12535         alignment to 16.
12536         (s390_override_options_after_change): Call s390_default align.
12537         (s390_option_override_internal): Call s390_default_align.
12538         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
12540 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12542         PR middle-end/86542
12543         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
12544         to _looptemp_ clauses, other than the first two.
12546 2018-07-17  Martin Liska  <mliska@suse.cz>
12548         * opts.c: Do not enable OPT_falign_* for -Os.
12550 2018-07-17  Martin Liska  <mliska@suse.cz>
12552         * align.h (MAX_CODE_ALIGN): New.
12553         (MAX_CODE_ALIGN_VALUE): New.
12554         * common/config/i386/i386-common.c (ix86_handle_option):
12555         (MAX_CODE_ALIGN): Moved to align.h.
12556         * final.c (MAX_CODE_ALIGN): Likewise.
12557         * opts.c (parse_and_check_align_values):
12558         (MAX_CODE_ALIGN): Likewise.
12559         (MAX_CODE_ALIGN_VALUE): Likewise.
12561 2018-07-17  Martin Liska  <mliska@suse.cz>
12563         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
12564         in order to fulfil coding style.
12565         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
12566         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12567         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12568         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
12569         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
12570         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
12571         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
12572         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
12573         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12575 2018-07-17  Martin Liska  <mliska@suse.cz>
12577         * align.h: New file.
12578         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
12579         directly.
12580         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
12581         align_flags of label_to_alignment.
12582         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
12583         align_flags class.
12584         * config/m68k/m68k.c: Do not use removed align_labels_value and
12585         align_loops_value.
12586         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
12587         (LOOP_ALIGN): Likewise.
12588         (LABEL_ALIGN): Likewise.
12589         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12590         Remove not used macro.
12591         (rs6000_loop_align): Change return type to align_flags.
12592         (rs6000_loop_align_max_skip): Remove.
12593         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
12594         Change return type to align_flags.
12595         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12596         Remove not used macro.
12597         (rs6000_loop_align):  Change return type to align_flags.
12598         (rs6000_loop_align_max_skip): Remove.
12599         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
12600         * config/rx/rx-protos.h (rx_align_for_label): Make it
12601         static function.
12602         * config/rx/rx.c (rx_align_for_label): Change return type
12603         to align_flags.
12604         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
12605         macro definitions.
12606         into align_flags class.
12607         (LABEL_ALIGN): Likewise.
12608         (LOOP_ALIGN): Likewise.
12609         * config/s390/s390.c (s390_label_align): Use align_flags
12610         class member.
12611         (s390_asm_output_function_label): Likewise.
12612         * config/sh/sh.c (sh_override_options_after_change):
12613         Use align_flags class directly without macros.
12614         (find_barrier): Likewise.
12615         (barrier_align): Likewise.
12616         (sh_loop_align): Likewise.
12617         * config/spu/spu.c (spu_option_override):
12618         Use align_flags_tuple::get_value instead of removed macros.
12619         (spu_sched_init): Likewise.
12620         * config/spu/spu.h (GTY): Likewise.
12621         * config/visium/visium.c (visium_option_override):
12622         Set "8" as default secondary alignment.
12623         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
12624         in order to guarantee secondary alignment of 8.
12625         * coretypes.h: Include align.h header file.
12626         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
12627         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
12628         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
12629         * doc/tm.texi.in: Likewise.
12630         * final.c (struct label_alignment): Remove not used structure.
12631         (LABEL_ALIGN): Change type to align_flags.
12632         (LOOP_ALIGN): Likewise.
12633         (JUMP_ALIGN): Likewise.
12634         (default_loop_align_max_skip): Remove.
12635         (default_label_align_max_skip): Likewise.
12636         (default_jump_align_max_skip): Likewise.
12637         (default_label_align_after_barrier_max_skip):
12638         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
12639         (LABEL_TO_MAX_SKIP): Remove.
12640         (label_to_alignment): Return align_flags type instead of integer.
12641         (label_to_max_skip): Remove.
12642         (align_fuzz): Use align_flags type.
12643         (compute_alignments): Use align_flags type and use align_flags::max
12644         to combine multiple alignments.
12645         (grow_label_align): Grow vec instead of C array.
12646         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
12647         (shorten_branches):  Use align_flags type and use align_flags::max
12648         to combine multiple alignments.
12649         (final_scan_insn_1): Remove usage of secondary alignment that comes
12650         from label alignment, but instead use proper secondary alignment
12651         which is computed in grow_label_align.
12652         * flags.h (struct align_flags_tuple): Move to align.h.
12653         (struct align_flags): Likewise.
12654         (state_align_loops): Rename to align_loops.
12655         (state_align_jumps): Rename to align_jumps.
12656         (state_align_labels): Rename to align_labels.
12657         (state_align_functions): Rename to align_functions.
12658         (align_loops_log): Remove.
12659         (align_jumps_log): Remove.
12660         (align_labels_log): Remove.
12661         (align_functions_log): Remove.
12662         (align_loops_max_skip): Remove.
12663         (align_jumps_max_skip): Remove.
12664         (align_labels_max_skip): Remove.
12665         (align_functions_max_skip): Remove.
12666         (align_loops_value): Remove.
12667         (align_jumps_value): Remove.
12668         (align_labels_value): Remove.
12669         (align_functions_value): Remove.
12670         * output.h (label_to_alignment): Change return type to align_flags.
12671         (label_to_max_skip): Remove.
12672         * target.def: Remove loop_align_max_skip, label_align_max_skip,
12673         jump_align_max_skip macros.
12674         * targhooks.h (default_loop_align_max_skip): Remove.
12675         (default_label_align_max_skip): Likewise.
12676         (default_jump_align_max_skip): Likewise.
12677         (default_label_align_after_barrier_max_skip): Remove.
12678         * toplev.c (read_log_maxskip): Use ::normalize function.
12679         (parse_N_M): Remove not used argument and also call ::normalize.
12680         (parse_alignment_opts): Do not pass unused arguments.
12681         * varasm.c (assemble_start_function): Use directly align_functions
12682         instead of removed macros.
12683         * system.h: Do not poison removed macros.
12685 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12687         PR middle-end/86539
12688         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
12689         and cond temporaries don't have reference type if iterator has
12690         pointer type.  For init use &for_pre_body instead of pre_p if
12691         for_pre_body is non-empty.
12693 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12695         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
12696         double-double modes to SFmode directly directly.
12697         (trunc<mode>sf2_fprs): Delete.
12699 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12701         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
12702         for conversions between IFmode and the decimal floating point modes.
12703         (init_float128_ieee): Use the correct names for conversions between
12704         KFmode and the decimal floating point modes.
12706 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12708         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
12709         for the conversions between TDmode and IFmode.
12710         (init_float128_ieee): Use more correct names for the conversions
12711         between TDmode and KFmode.
12713 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
12715         PR tree-optimization/86526
12716         * builtins.c (expand_builtin_memcmp): Formatting fixes.
12717         (inline_expand_builtin_string_cmp): Likewise.
12718         (inline_string_cmp): Likewise.  Use c_readstr instead of
12719         builtin_memcpy_read_str.  Add unit_mode temporary.
12721 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12723         PR middle-end/86528
12724         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
12725         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
12727 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12729         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
12730         Alphabetize prototypes of built-in functions, separating out
12731         built-in functions that are listed in this section but should be
12732         described elsewhere.
12734 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
12736         PR target/86511
12737         * expmed.c (emit_store_flag): Do not emit setcc followed by a
12738         conditional move when trapping comparison was split to a
12739         non-trapping one (and vice versa).
12741 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12743         * config/s390/s390.c (s390_function_profiler): Generate nops
12744         instead of profiler call sequences.
12745         * config/s390/s390.opt: Add the new option.
12747 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12749         * config/s390/s390.c (s390_function_profiler): Generate
12750         __mcount_loc section.
12751         * config/s390/s390.opt: Add the new option.
12753 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12755         * common.opt: Add the new warning.
12756         * config/s390/s390.c (s390_function_profiler): Emit "brasl
12757         %r0,__fentry__" when -mfentry is specified.
12758         (s390_option_override_internal): Disallow -mfentry for 31-bit
12759         CPUs.
12760         * config/s390/s390.opt: Add the new option.
12762 2018-07-16  Richard Biener  <rguenther@suse.de>
12764         PR lto/86523
12765         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
12766         for function-local FUNCTION_DECL and RESULT_DECL immediately.
12768 2018-07-16  Martin Liska  <mliska@suse.cz>
12770         PR ipa/86529
12771         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
12772         to ::get_create.
12774 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12776         * config/arc/arcHS.md: Update ARCHS scheduling rules.
12778 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12780         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
12781         for ARCHS4x.
12782         * config/arc/arc-cpus.def (hs4x): New cpu.
12783         (hs4xd): Likewise.
12784         * config/arc/arc-tables.opt: Regenerate.
12785         * config/arc/arc.c (arc_sched_issue_rate): New function.
12786         (TARGET_SCHED_ISSUE_RATE): Define.
12787         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
12788         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
12789         fpu_cvt.
12790         (attr tune): Add ARCHS4x tune values.
12791         (attr tune_dspmpy): Define.
12792         (*tst): Correct instruction type.
12793         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
12794         * config/arc/arcHS4x.md: New file.
12795         * config/arc/fpu.md: Update instruction type attributes.
12796         * config/arc/t-multilib: Regenerate.
12798 2018-07-16  Tom de Vries  <tdevries@suse.de>
12800         PR debug/86455
12801         * var-tracking.c (vt_initialize): Fix pre_dec handling.
12803 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
12805         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
12806         early clobber.
12808 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
12810         PR tree-optimization/86514
12811         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
12812         conversion to a boolean type from a type with greater precision.
12814 2018-07-16  Tom de Vries  <tdevries@suse.de>
12816         * var-tracking.c (vt_initialize): Print adjusted insn slim if
12817         dump_flags request TDF_SLIM.
12819 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
12821         * fold-const.c (int_const_binop_1): Abstract...
12822         (wide_int_binop): ...wide int code here.
12823         (poly_int_binop): ...poly int code here.
12824         Abstract the rest of int_const_binop_1 into int_const_binop.
12825         * fold-const.h (wide_int_binop): New.
12826         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
12827         Remove useless PLUS/MINUS_EXPR case.
12828         (zero_nonzero_bits_from_vr): Move wide int code...
12829         (zero_nonzero_bits_from_bounds): ...here.
12830         (extract_range_from_binary_expr_1): Move mask optimization code...
12831         (range_easy_mask_min_max): ...here.
12832         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
12833         (range_easy_mask_min_max): New.
12835 2018-07-15  Jeff Law  <law@redhat.com>
12837         PR target/85993
12838         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
12839         block.
12841 2018-07-14  Jim Wilson  <jimw@sifive.com>
12843         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
12845 2018-07-14  Paul Koning  <ni1d@arrl.net>
12847         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
12849 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
12851         * lto-streamer-out.c (copy_function_or_variable): Dump info about
12852         copying section.
12854 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
12855             Steve Munroe  <munroesj52@gmail.com>
12857         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
12858         (_mm_andnot_si128): Likewise.
12859         (_mm_or_si128): Likewise.
12860         (_mm_xor_si128): Likewise.
12862 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
12864         PR middle-end/78809
12865         * builtins.c (expand_builtin_memcmp): Inline the calls first
12866         when result_eq is false.
12867         (expand_builtin_strcmp): Inline the calls first.
12868         (expand_builtin_strncmp): Likewise.
12869         (inline_string_cmp): New routine. Expand a string compare
12870         call by using a sequence of char comparison.
12871         (inline_expand_builtin_string_cmp): New routine. Inline expansion
12872         a call to str(n)cmp/memcmp.
12873         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
12874         New option.
12875         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
12877 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
12879         * config/arm/driver-arm.c: Include arm-native.h.
12880         (host_detect_local_cpu): Use auto-generated data tables.
12881         (vendors, arm_cpu_table): Delete.  Move part information to ...
12882         * config/arm/arm-cpus.in: ... here.
12883         * config/arm/parsecpu.awk (gen_native): New function.
12884         (vendor, part): New CPU fields.
12885         (END): Add support for building the native CPU detection tables.
12886         * config/arm/t-arm (arm-native.h): Add build rule.
12887         (driver-arm.o): Add dependency on arm-native.h.
12889 2018-07-13  Richard Biener  <rguenther@suse.de>
12891         PR middle-end/85974
12892         * match.pd (addr1 - addr2): Allow either of the operand to
12893         have a conversion.
12895 2018-07-13  Tom de Vries  <tdevries@suse.de>
12897         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
12898         in remap_ssa_name.
12900 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
12902         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
12903         arrays instead of numbered variables.
12905 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
12907         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
12908         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
12910 2018-07-13  Richard Biener  <rguenther@suse.de>
12912         PR debug/86452
12913         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
12914         instead of get_context_die.
12916 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
12917             Richard Biener  <rguenther@suse.de>
12919         PR middle-end/86489
12920         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
12921         that the loop latch destination where phi is defined.
12923 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
12925         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
12926         (riscv_expand_epilogue): Add assertion to check interrupt mode.
12927         (riscv_set_current_function): Extract getting interrupt type to new
12928         function.
12929         (riscv_get_interrupt_type): New function.
12930         (riscv_merge_decl_attributes): New function, checking interrupt type is
12931         same.
12932         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
12934 2018-07-12  Paul Koning  <ni1d@arrl.net>
12936         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
12937         directive.
12939 2018-07-12  Paul Koning  <ni1d@arrl.net>
12941         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
12942         zero reference, add doloop_end instead.
12943         * doc/md.texi (decrement_and_branch_until_zero): Remove.
12944         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
12945         detail for doloop_end.
12947 2018-07-12  Martin Sebor  <msebor@redhat.com>
12949         PR c/86453
12950         * attribs.c (decl_attributes): Reject conflicting attributes before
12951         calling attribute handlers.
12953 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12955         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
12956          parameter.
12957         (gcc::dump_manager::get_dump_file_name): likewise.
12958         (dump_begin): Likewise.
12959         * dumpfile.h (dump_begin): Update prototype.
12960         (gcc::dump_manager::get_dump_file_name,
12961         gcc::dump_manager::get_dump_file_name): Update prototype.
12963 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12965         * internal-fn.h (vectorizable_internal_fn_p): New function.
12966         * tree-vect-slp.c (compatible_calls_p): Likewise.
12967         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
12968         to internal functions.
12969         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
12971 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12973         * fold-const.h (inverse_conditions_p): Declare.
12974         * fold-const.c (inverse_conditions_p): New function.
12975         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
12976         that test the inverse condition of a conditional internal function.
12977         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
12978         * internal-fn.c (internal_fn_mask_index): Handle conditional
12979         internal functions.
12980         (vectorized_internal_fn_supported_p): New function.
12981         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
12982         (any_pred_load_store): Replace with...
12983         (need_to_predicate): ...this new variable.
12984         (redundant_ssa_names): New variable.
12985         (ifcvt_can_use_mask_load_store): Move initial checks to...
12986         (ifcvt_can_predicate): ...this new function.  Handle tree codes
12987         for which a conditional internal function exists.
12988         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
12989         instead of ifcvt_can_use_mask_load_store.  Update after variable
12990         name change.
12991         (predicate_load_or_store): New function, split out from
12992         predicate_mem_writes.
12993         (check_redundant_cond_expr): New function.
12994         (value_available_p): Likewise.
12995         (predicate_rhs_code): Likewise.
12996         (predicate_mem_writes): Rename to...
12997         (predicate_statements): ...this.  Use predicate_load_or_store
12998         and predicate_rhs_code.
12999         (combine_blocks, tree_if_conversion): Update after above name changes.
13000         (ifcvt_local_dce): Handle redundant_ssa_names.
13001         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
13002         general conditional functions.
13003         * tree-vect-stmts.c (vectorizable_call): Likewise.
13005 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13006             Alan Hayward  <alan.hayward@arm.com>
13007             David Sherwood  <david.sherwood@arm.com>
13009         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
13010         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
13011         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
13012         plus and minus and convert them into IFN_COND_FMA-based sequences.
13013         (convert_mult_to_fma): Handle conditional plus and minus.
13015 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13017         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
13018         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
13019         (cond_fnms_optab): New optabs.
13020         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
13021         internal functions.
13022         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
13023         * internal-fn.h (get_conditional_internal_fn): Declare.
13024         (get_unconditional_internal_fn): Likewise.
13025         * internal-fn.c (cond_ternary_direct): New macro.
13026         (expand_cond_ternary_optab_fn): Likewise.
13027         (direct_cond_ternary_optab_supported_p): Likewise.
13028         (FOR_EACH_COND_FN_PAIR): Likewise.
13029         (get_conditional_internal_fn): New function.
13030         (get_unconditional_internal_fn): Likewise.
13031         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
13032         (gimple_match_op::gimple_match_op): Add a new overload for 5
13033         operands.
13034         (gimple_match_op::set_op): Likewise.
13035         (gimple_resimplify5): Declare.
13036         * genmatch.c (decision_tree::gen): Generate simplifications for
13037         5 operands.
13038         * gimple-match-head.c (gimple_simplify): Define an overload for
13039         5 operands.  Handle calls with 5 arguments in the top-level overload.
13040         (convert_conditional_op): Handle conversions from unconditional
13041         internal functions to conditional ones.
13042         (gimple_resimplify5): New function.
13043         (build_call_internal): Pass a fifth operand.
13044         (maybe_push_res_to_seq): Likewise.
13045         (try_conditional_simplification): Try converting conditional
13046         internal functions to unconditional internal functions.
13047         Handle 3-operand unconditional forms.
13048         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
13049         Define ternary equivalents of the current rules for binary conditional
13050         internal functions.
13051         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
13052         ternary operations.
13053         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
13054         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
13055         (optab): Handle them.
13056         (SVE_COND_FP_TERNARY): New int iterator.
13057         (sve_fmla_op, sve_fmad_op): New int attributes.
13058         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
13059         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
13060         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
13062 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13064         * target.def (preferred_else_value): New target hook.
13065         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
13066         * doc/tm.texi: Regenerate.
13067         * targhooks.h (default_preferred_else_value): Declare.
13068         * targhooks.c (default_preferred_else_value): New function.
13069         * internal-fn.h (conditional_internal_fn_code): Declare.
13070         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
13071         (get_conditional_internal_fn): Use it.
13072         (conditional_internal_fn_code): New function.
13073         * gimple-match.h (gimple_match_cond): New struct.
13074         (gimple_match_op): Add a cond member function.
13075         (gimple_match_op::gimple_match_op): Update all forms to take a
13076         gimple_match_cond.
13077         * genmatch.c (expr::gen_transform): Use the same condition as res_op
13078         for the suboperation, but don't specify a particular else_value.
13079         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
13080         (visit_nary_op, visit_reference_op_load): Pass
13081         gimple_match_cond::UNCOND to the gimple_match_op constructor.
13082         * gimple-match-head.c: Include tree-eh.h
13083         (convert_conditional_op): New function.
13084         (maybe_resimplify_conditional_op): Likewise.
13085         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
13086         (gimple_resimplify2): Likewise.
13087         (gimple_resimplify3): Likewise.
13088         (gimple_resimplify4): Likewise.
13089         (maybe_push_res_to_seq): Return null for conditional operations.
13090         (try_conditional_simplification): New function.
13091         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
13092         constructor.
13093         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
13094         IFN_COND_* call.
13095         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
13096         function.
13097         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
13099 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
13101         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
13102         DECL_FCONTEXT
13103         (hash_tree): Do not hash DECL_FCONTEXT
13104         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
13105         Do not stream DECL_FCONTEXT.
13106         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
13107         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
13109 2018-07-12  Richard Biener  <rguenther@suse.de>
13111         PR debug/86462
13112         * dwarf2out.c (gen_block_die): Only output blocks when they have
13113         at least one !DECL_IGNORED_P variable.
13115 2018-07-12  Richard Biener  <rguenther@suse.de>
13117         PR target/84829
13118         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
13119         Remove -mieee-fp handling.
13121 2018-07-12  Richard Biener  <rguenther@suse.de>
13123         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
13124         left-over from last patch.
13126 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
13128         PR tree-optimization/86492
13129         * gimple-ssa-store-merging.c
13130         (imm_store_chain_info::coalesce_immediate_stores): Call
13131         check_no_overlap even for the merge_overlapping case.  Formatting fix.
13133 2018-07-12  Richard Biener  <rguenther@suse.de>
13135         PR middle-end/86479
13136         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
13137         move possibly trapping operations into the conditional.
13139 2018-07-12  Richard Biener  <rguenther@suse.de>
13141         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
13142         (vn_lookup_simplify_result): Remove recursion limit applied
13143         here.
13144         (vn_nary_build_or_lookup_1): Adjust.
13145         (try_to_simplify): Likewise.
13146         * gimple-match-head.c (gimple_resimplify1): Instead apply one
13147         here.
13148         (gimple_resimplify2): Likewise.
13149         (gimple_resimplify3): Likewise.
13150         (gimple_resimplify4): Likewise.
13152 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
13154         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
13155         Use __mmask64 type instead of __mmask8 for __M argument.
13156         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
13157         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
13158         __U argument.
13159         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
13160         __mmask16 for __M argument.
13161         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
13162         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
13163         to __mmask16 instead of __mmask8.
13164         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
13165         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
13166         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
13167         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
13168         instead of __mmask16 for __U argument.
13169         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
13170         __mmask16 instead of __mmask8 for __U argument.
13171         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
13172         __U argument.
13173         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
13174         __mmask16.
13175         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
13176         argument.
13177         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
13178         __U argument.
13179         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
13180         __mmask16.
13181         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
13182         of __mmask16.
13183         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
13184         __U argument.
13185         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
13186         __U argument.
13187         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
13188         __U argument.
13189         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
13190         __U argument.
13191         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
13192         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
13193         return type as well as __M argument type and all casts from __mmask8
13194         to __mmask32.
13195         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
13196         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
13197         return type as well as __M argument type and all casts from __mmask8
13198         to __mmask16.
13199         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
13200         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
13201         return type as well as __M argument type and all casts from __mmask8
13202         to __mmask32.
13203         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
13204         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
13205         return type as well as __M argument type and all casts from __mmask8
13206         to __mmask16.
13207         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
13208         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
13209         __mmask16.
13211 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
13213         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
13214         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
13215         for __U argument.
13217 2018-07-11  Paul Koning  <ni1d@arrl.net>
13219         * doc/md.texi (define_subst): Document how multiple occurrences of
13220         the same argument in the replacement pattern are handled.
13222 2018-07-11  Paul Koning  <ni1d@arrl.net>
13224         * doc/extend.texi (Common Variable Attributes): Move "mode" into
13225         alphabetical order.
13226         (Common Type Attributes): Add "mode" attribute.
13228 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
13230         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
13231         stream DECL_ORIGINAL_TYPE.
13232         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
13233         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
13234         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
13235         Do not walk original type.
13236         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
13237         external decls.
13238         (write_ts_decl_non_common_tree_pointers): Do not stream
13239         DECL_ORIGINAL_TYPE
13240         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
13241         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
13243 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
13245         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
13246         thread twice from the same starting edge.
13248 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
13250         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
13251         * gimple.c (gimple_call_nonnull_result_p): ...here...
13252         (gimple_call_nonnull_arg): ...and here.
13253         * gimple.h (gimple_call_nonnull_result_p): New.
13254         (gimple_call_nonnull_arg): New.
13256 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
13258         * config/arm/arm-cpus.in: Move information from fpu field of each
13259         cpu definition to the isa field.
13260         * config/arm/parsecpu.awk (fpu): Delete match rule.
13261         (gen_comm_data): Don't add bits from the CPU's FPU entry.
13263 2018-07-11  Richard Biener  <rguenther@suse.de>
13265         PR debug/86457
13266         * dwarf2out.c (init_sections_and_labels): Use
13267         output_asm_line_debug_info consistently.
13268         (dwarf2out_early_finish): Likewise.
13269         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
13270         type units.
13272 2018-07-11  Richard Biener  <rguenther@suse.de>
13274         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
13275         Rework father_bb setting in a way to avoid propagating constants
13276         multiple times on a loop body.
13278 2018-07-10  Mark Wielaard  <mark@klomp.org>
13280         PR debug/86459
13281         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
13283 2018-07-10  Richard Biener  <rguenther@suse.de>
13285         * hash-map.h (hash_map::iterator::operator*): Return
13286         references to key and value.
13288 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
13290         PR c++/86443
13291         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
13292         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
13293         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
13294         (gimplify_omp_for): For composite loops, move outer
13295         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
13296         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
13297         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
13298         TREE_LIST for both the original class iterator and the "last" helper
13299         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
13300         loop, remember has_decl_expr from outer composite loops for the
13301         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
13303 2018-07-09  Martin Sebor  <msebor@redhat.com>
13305         PR middle-end/77357
13306         PR middle-end/86428
13307         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
13308         accessing implicitly initialized array elements.
13309         * expr.c (string_constant): Handle string initializers of
13310         character arrays within aggregates.
13311         * gimple-fold.c (fold_array_ctor_reference): Add argument.
13312         Store element offset.  As a special case, handle zero size.
13313         (fold_nonarray_ctor_reference): Same.
13314         (fold_ctor_reference): Add argument.  Store subobject offset.
13315         * gimple-fold.h (fold_ctor_reference): Add argument.
13317 2018-07-09  Paul Koning  <ni1d@arrl.net>
13319         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
13320         (pdp11_insn_cost): New function.
13321         (pdp11_md_asm_adjust): New function.
13322         (TARGET_INVALID_WITHIN_DOLOOP): Define.
13323         (pdp11_rtx_costs): Update to match machine better.
13324         (output_addr_const_pdp11): Correct format mismatch warnings.
13325         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
13326         * config/pdp11/pdp11.md: General change to add base_cost and/or
13327         length attributes for use by new pdp11_insn_cost function.
13328         (MIN_BRANCH): Correct definition.
13329         (MIN_SOB): Ditto.
13330         (doloop_end): Use standard pattern name for looping pattern.
13331         (doloop_end_nocc): New.
13332         (movsf): Add another constraint alternative.
13333         (zero_extendqihi2): Add constraint alternatives for not in place
13334         extend.
13335         (zero_extendhisi2): Remove.
13336         (shift patterns): Add CC handling variants.
13337         (bswaphi2): New.
13338         (bswapsi2): New.
13339         (rothi3): New.
13340         (define_peephole2): New peephole to recognize mov that sets CC for
13341         subsequent test.
13343 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13345         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
13346         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
13347         wi::add.
13349 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
13351         PR c/86420
13352         * real.c (real_nextafter): Return true if result is denormal.
13354 2018-07-09  Martin Liska  <mliska@suse.cz>
13356         * common.opt: Add back wrongly removed attribute.
13358 2018-07-09  Richard Biener  <rguenther@suse.de>
13360         PR debug/86413
13361         * dwarf2out.c (gen_block_die): For an early generated DIE
13362         always output high/low PC attributes.
13364 2018-07-09  Tom de Vries  <tdevries@suse.de>
13366         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
13367         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
13368         onto VAR_DECL with abstract origin.
13370 2018-07-07  Jim Wilson  <jimw@sifive.com>
13372         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
13374 2018-07-07  Tom de Vries  <tdevries@suse.de>
13376         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
13378 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
13380         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
13381         overflow_type.
13382         (combine_bound): Use wide-int overflow calculation instead of
13383         rolling our own.
13384         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
13385         overflow_type.
13386         * fold-const.c (int_const_binop_2): Same.
13387         (extract_muldiv_1): Same.
13388         (fold_div_compare): Same.
13389         (fold_abs_const): Same.
13390         * match.pd: Same.
13391         * poly-int.h (add): Same.
13392         (sub): Same.
13393         (neg): Same.
13394         (mul): Same.
13395         * predict.c (predict_iv_comparison): Same.
13396         * profile-count.c (slow_safe_scale_64bit): Same.
13397         * simplify-rtx.c (simplify_const_binary_operation): Same.
13398         * tree-chrec.c (tree_fold_binomial): Same.
13399         * tree-data-ref.c (split_constant_offset_1): Same.
13400         * tree-if-conv.c (idx_within_array_bound): Same.
13401         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
13402         * tree-ssa-phiopt.c (minmax_replacement): Same.
13403         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
13404         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
13405         * vr-values.c (vr_values::adjust_range_with_scev): Same.
13406         * wide-int.cc (wi::add_large): Same.
13407         (wi::mul_internal): Same.
13408         (wi::sub_large): Same.
13409         (wi::divmod_internal): Same.
13410         * wide-int.h: Change overflow type to overflow_type for neg, add,
13411         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
13412         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
13413         mul_internal, divmod_internal.
13414         (overflow_type): New enum.
13415         (accumulate_overflow): New.
13417 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
13419         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
13420         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
13422 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
13424         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
13425         argument is checked for zero before entering loop, avoid checking again.
13427 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
13429         * gimplify.h (generic_expr_could_trap_p): Set as global function.
13430         * gimplify.h (generic_expr_could_trap_p): Likwise.
13431         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
13433 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
13435         PR tree-optimization/86401
13436         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
13437         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
13438         (fold_bit_and_mask): ... here.  New helper function for match.pd.
13439         * fold-const.h (fold_bit_and_mask): Declare.
13440         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
13442 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
13444         PR target/86324
13445         * target.def (translate_mode_attribute): New hook.
13446         * targhooks.h (default_translate_mode_attribute): Declare.
13447         * targhooks.c (default_translate_mode_attribute): New function.
13448         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
13449         * doc/tm.texi: Regenerate.
13450         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
13451         (rs6000_translate_mode_attribute): New function.
13453 2018-07-06  Paul Koning  <ni1d@arrl.net>
13455         * doc/md.texi (define_split): Document DONE and FAIL.
13456         (define_peephole2): Ditto.
13458 2018-07-05  Jeff Law  <law@redhat.com>
13460         PR tree-optimization/86010
13461         * tree-ssa-dse.c (compute_trims): More aggressively trim at
13462         both the head and tail of mem* and str* calls.
13464 2018-07-05  Jim Wilson  <jimw@sifive.com>
13466         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
13468 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
13470         * config/aarch64/aarch64-simd.md: correct flags text for
13471         MIN_EXPR replacement.
13473 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
13475         * configure: Regenerated.
13477 2018-07-05  Carl Love  <cel@us.ibm.com>
13479         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
13480         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
13481         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
13482         VSX_BUILTIN_DOUBLEL_V4SF.
13484 2018-07-05  Martin Sebor  <msebor@redhat.com>
13486         PR c++/86400
13487         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
13488         than its domain to compute its the upper bound of a char array.
13490 2018-07-05  Nathan Sidwell  <nathan@acm.org>
13492         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
13493         * doc/cpp.texi: Update comment.
13494         * doc/tm.texi: Rebuilt.
13495         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
13496         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
13497         * doc/extend.texi (Backwards Compatibility): Clarify it is system
13498         headers affected by extern "C".
13499         * system.h: Poison NO_IMPLICIT_EXTERN_C.
13500         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
13501         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
13502         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
13503         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
13504         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
13505         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
13506         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
13507         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
13508         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
13509         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
13510         config/sparc/sp64-elf.h, config/spu/spu.h,
13511         config/stormy16/stormy16.h, config/v850/v850.h,
13512         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
13513         define NO_IMPLICIT_EXTERN_C.
13514         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
13516 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13518         PR target/84711
13519         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
13520         instead of GET_MODE_SIZE when comparing Units.
13522 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13524         PR target/84711
13525         * rtlanal.c (set_noop_p): Constrain on mode change,
13526         include hard-reg-set.h
13528 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13530         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
13532 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
13534         Revert
13535         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13537         PR sanitizer/84250
13538         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13539         libasan.
13540         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13542 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13544         PR sanitizer/84250
13545         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13546         libasan.
13547         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13549 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
13551         PR middle-end/86380
13552         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
13554 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
13556         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
13557         neg_*_op* variables.
13559 2018-07-04  Martin Liska  <mliska@suse.cz>
13561         * tree-switch-conversion.c: Define
13562         max_ratio_for_speed and max_ratio_for_size constants.
13564 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13565             Martin Liska  <mliska@suse.cz>
13567         PR middle-end/66240
13568         PR target/45996
13569         PR c/84100
13570         * common.opt: Rename align options with 'str_' prefix.
13571         * common/config/i386/i386-common.c (set_malign_value): New
13572         function.
13573         (ix86_handle_option): Use it to set -falign-* options/
13574         * config/aarch64/aarch64-protos.h (struct tune_params): Change
13575         type from int to string.
13576         * config/aarch64/aarch64.c: Update default values from int
13577         to string.
13578         * config/alpha/alpha.c (alpha_override_options_after_change):
13579         Likewise.
13580         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
13581         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13582         max skip conditionally.
13583         * config/i386/freebsd.h (SUBALIGN_LOG): New.
13584         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13585         max skip conditionally.
13586         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13587         max skip conditionally.
13588         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
13589         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13590         max skip conditionally.
13591         * config/i386/i386.c (struct ptt): Change type from int to
13592         string.
13593         (ix86_default_align): Set default values.
13594         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
13595         max skip conditionally.
13596         * config/i386/iamcu.h (SUBALIGN_LOG): New.
13597         (ASM_OUTPUT_MAX_SKIP_ALIGN):
13598         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
13599         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13600         max skip conditionally.
13601         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
13602         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
13603         * config/i386/x86-64.h (SUBALIGN_LOG): New.
13604         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13605         max skip conditionally.
13606         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13607         * config/ia64/ia64.c (ia64_option_override): Set default values
13608         for alignment options.
13609         * config/m68k/m68k.c: Handle new str_align_* options.
13610         * config/mips/mips.c (mips_set_compression_mode): Change
13611         type of constants.
13612         (mips_option_override): Set default values for options.
13613         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
13614         Likewise.
13615         * config/rs6000/rs6000.c (rs6000_option_override_internal):
13616         Likewise.
13617         * config/rx/rx.c (rx_option_override): Likewise.
13618         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
13619         (LABEL_ALIGN): Use align_labels_log.
13620         (LOOP_ALIGN): Use align_loops_align.
13621         * config/s390/s390.c (s390_asm_output_function_label): Use new
13622         macros.
13623         * config/sh/sh.c (sh_override_options_after_change):
13624         Change type of constants.
13625         * config/spu/spu.c (spu_sched_init): Likewise.
13626         * config/sparc/sparc.c (sparc_option_override): Set default
13627         values for options.
13628         * config/visium/visium.c (visium_option_override): Likewise.
13629         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
13630         emit p2align format with last argument if it's not needed.
13631         * doc/invoke.texi: Document extended format of -falign-*.
13632         * final.c: Use align_labels alignment.
13633         * flags.h (struct target_flag_state): Change type to use
13634         align_flags.
13635         (struct align_flags_tuple): New.
13636         (struct align_flags): Likewise.
13637         (align_loops_log): Redefine macro to use new types.
13638         (align_loops_max_skip): Redefine macro to use new types.
13639         (align_jumps_log): Redefine macro to use new types.
13640         (align_jumps_max_skip): Redefine macro to use new types.
13641         (align_labels_log): Redefine macro to use new types.
13642         (align_labels_max_skip): Redefine macro to use new types.
13643         (align_functions_log): Redefine macro to use new types.
13644         (align_loops): Redefine macro to use new types.
13645         (align_jumps): Redefine macro to use new types.
13646         (align_labels): Redefine macro to use new types.
13647         (align_functions): Redefine macro to use new types.
13648         (align_functions_max_skip): Redefine macro to use new types.
13649         (align_loops_value): New macro.
13650         (align_jumps_value): New macro.
13651         (align_labels_value): New macro.
13652         (align_functions_value): New macro.
13653         * function.c (invoke_set_current_function_hook): Propagate
13654         alignment values from flags to global variables default in
13655         topleev.h.
13656         * ipa-icf.c (sem_function::equals_wpa): Use
13657         cl_optimization_option_eq instead of memcmp.
13658         * lto-streamer.h (cl_optimization_stream_out): Support streaming
13659         of string types.
13660         (cl_optimization_stream_in): Likewise.
13661         * optc-save-gen.awk: Support strings in cl_optimization.
13662         * opth-gen.awk: Likewise.
13663         * opts.c (finish_options): Remove error checking of invalid
13664         value ranges.
13665         (MAX_CODE_ALIGN): Remove.
13666         (MAX_CODE_ALIGN_VALUE): Likewise.
13667         (parse_and_check_align_values): New function.
13668         (check_alignment_argument): Likewise.
13669         (common_handle_option): Use check_alignment_argument.
13670         * opts.h (parse_and_check_align_values): Declare.
13671         * toplev.c (init_alignments): Remove.
13672         (read_log_maxskip): New.
13673         (parse_N_M): Likewise.
13674         (parse_alignment_opts): Likewise.
13675         (backend_init_target): Remove usage of init_alignments.
13676         * toplev.h (parse_alignment_opts): Declare.
13677         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
13678         argument.
13679         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
13680         * tree.c (cl_option_hasher::equal): New.
13681         * varasm.c: Use new global macros.
13683 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13685         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
13686         Use a simpler align directive also if MAXSKIP = ALIGN-1.
13687         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13688         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13689         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13690         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13691         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
13692         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
13693         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
13694         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13695         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13696         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13697         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13699 2018-07-04  Martin Liska  <mliska@suse.cz>
13700             Jonathan Wakely  <jwakely@redhat.com>
13702         * coverage.c: Use correct type.
13703         * doc/invoke.texi: Language correction.
13705 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
13707         PR target/85620
13708         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
13709         ENDBRANCH for non-tail call which may return via indirect branch.
13710         * doc/extend.texi: Document indirect_return attribute.
13712 2018-07-03  Martin Sebor  <msebor@redhat.com>
13714         PR tree-optimization/86274
13715         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
13716         precondition.
13717         (format_floating): Correct handling of infinities and NaNs.
13719 2018-07-03  Martin Sebor  <msebor@redhat.com>
13721         * print-tree.c (print_real_cst): New function.
13722         (print_node_brief): Call it.
13723         (print_node): Ditto.
13725 2018-07-03  Jeff Law  <law@redhat.com>
13727         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
13728         into a single pattern.
13730         * config/h8300/h8300.md (ors code_iterator): New.
13731         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
13732         a single pattern and single splitter.
13733         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
13734         (iorqi3_1, xorqi3_1): Likewise.
13735         (iorqi3, xorqi3 expanders): Similarly.
13737         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
13738         (movmd_internal) into a single pattern using the P mode iterator.
13739         (movmd splitters): Similarly.
13740         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
13741         (movsd splitters): Similarly.
13743         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
13744         ADDB, ADDW and ADDL into a single ADD attribute which selects the
13745         right table based on the size of the operand.
13746         * config/h8300/h8300.md (length_table): Corresponding changes. All
13747         references to "addb", "addw" and "addl" changed to "add".
13748         (btst patterns): Merge two variants into a single pattern.
13749         (tstqi, tsthi): Likewise.
13750         (addhi3_incdec, addsi3_incdec): Likewise.
13751         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
13752         (mulhi3, mulsi3): Likewise.
13753         (udivhi3, udivsi3): Likewise.
13754         (divhi3, divsi3): Likewise.
13755         (andorqi3, andorhi3, andorsi3): Likewise.
13757 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
13759         PR target/85694
13760         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
13761         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
13763 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13765         PR tree-optimization/85694
13766         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
13767         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
13768         UNSPEC_URHADD.
13769         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
13770         (<u>avg<mode>3_ceil): New patterns.
13772 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
13774         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
13775         scan-tree-dump directive.
13776         * gcc.dg/vect/slp-perm-2.c: Likewise.
13777         * gcc.dg/vect/slp-perm-3.c: Likewise.
13778         * gcc.dg/vect/slp-perm-5.c: Likewise.
13779         * gcc.dg/vect/slp-perm-6.c: Likewise.
13780         * gcc.dg/vect/slp-perm-7.c: Likewise.
13781         * gcc.dg/vect/slp-perm-8.c: Likewise.
13783 2018-07-03  Marek Polacek  <polacek@redhat.com>
13785         PR middle-end/86202
13786         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
13788 2018-07-03  Richard Biener  <rguenther@suse.de>
13790         PR ipa/86389
13791         * tree-ssa-structalias.c (find_func_clobbers): Properly
13792         handle indirect calls.
13794 2018-07-03  Jeff Law  <law@redhat.com>
13796         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
13797         (shifts): New code iterator.
13798         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
13799         expander.  Fix HImode handling on H8/SX.
13800         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
13801         (subqi3, subhi3, subsi3 expanders): Likewise.
13802         (andqi3, andhi3, andsi3 expanders): Likewise.
13803         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
13804         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
13805         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
13806         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
13807         (zero_extendqihi2, zero_extendqisi2): Likewise.
13808         (extendqihi2, extendqisi2): Likewise.
13809         (rotlqi3, rotlhi3, rotlsi3): Likewise.
13810         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
13811         (rotlqi3_1, rotlhi3_1): Likewise.
13812         (logicalhi3_sn, logicalsi3_sn): Likewise.
13813         (logicalhi3, logicalsi3): Likewise.
13815 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13817         * tree-vect-patterns.c (vect_recog_rotate_pattern)
13818         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13819         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
13820         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
13821         type to append_pattern_def_seq instead of creating a stmt_vec_info
13822         directly.
13823         (build_mask_conversion): Likewise.  Remove vinfo argument.
13824         (vect_add_conversion_to_patterm): Likewise, renaming to...
13825         (vect_add_conversion_to_pattern): ...this.
13826         (vect_recog_mask_conversion_pattern): Update call to
13827         build_mask_conversion.  Pass the vector type to
13828         append_pattern_def_seq here too.
13829         (vect_recog_gather_scatter_pattern): Update call to
13830         vect_add_conversion_to_pattern.
13832 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13834         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
13835         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13836         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
13837         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
13838         STMT_VINFO_PATTERN_DEF_SEQ to null here.
13839         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
13840         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
13841         append_pattern_def_seq instead of new_pattern_def_seq.
13842         (vect_recog_divmod_pattern): Do both of the above.
13843         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
13844         is null.
13846 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13848         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
13849         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
13850         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
13851         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
13852         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
13853         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13854         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
13855         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13856         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13857         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
13858         parameter with a single stmt_vec_info.
13859         (vect_recog_func_ptr): Likewise.
13860         (vect_recog_gather_scatter_pattern): Likewise, folding in...
13861         (vect_try_gather_scatter_pattern): ...this.
13862         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
13863         the stmt_vec_info of the statement to be matched.  Don't clear
13864         STMT_VINFO_RELATED_STMT.
13865         (vect_pattern_recog): Update call accordingly.
13867 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13869         PR tree-optimization/85694
13870         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
13871         (uavgM3_ceil): Document new optabs.
13872         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
13873         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
13874         functions.
13875         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
13876         (savg_ceil_optab): New optabs.
13877         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
13878         (vect_vect_recog_func_ptrs): Add it.
13879         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
13880         constant directly from the associated lhs.
13882 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13884         * tree-vect-patterns.c (vect_split_statement): New function.
13885         (vect_convert_input): Use it to try to split an existing cast.
13887 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13889         * poly-int.h (print_hex): New function.
13890         * dumpfile.h (dump_dec, dump_hex): Declare.
13891         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
13892         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
13893         min_input_precision, operation_precision and operation_sign.
13894         * tree-vect-patterns.c (vect_get_range_info): New function.
13895         (vect_same_loop_or_bb_p, vect_single_imm_use)
13896         (vect_operation_fits_smaller_type): Delete.
13897         (vect_look_through_possible_promotion): Add an optional
13898         single_use_p parameter.
13899         (vect_recog_over_widening_pattern): Rewrite to use new
13900         stmt_vec_info infomration.  Handle one operation at a time.
13901         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
13902         (vect_truncatable_operation_p, vect_set_operation_type)
13903         (vect_set_min_input_precision): New functions.
13904         (vect_determine_min_output_precision_1): Likewise.
13905         (vect_determine_min_output_precision): Likewise.
13906         (vect_determine_precisions_from_range): Likewise.
13907         (vect_determine_precisions_from_users): Likewise.
13908         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
13909         (vect_vect_recog_func_ptrs): Put over_widening first.
13910         Add cast_forwprop.
13911         (vect_pattern_recog): Call vect_determine_precisions.
13913 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13915         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
13916         statements that have been replaced by further pattern statements.
13917         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
13919 2018-07-03  Richard Biener  <rguenther@suse.de>
13921         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
13922         always set *dt.  Dump vectype in vectype overload.
13923         * dumpfile.h (dump_gimple_expr): New function.
13924         (dump_gimple_expr_loc): Likewise.
13925         * dumpfile.c (dump_gimple_expr): New function.
13926         (dump_gimple_expr_loc): Likewise.
13928 2018-07-02  Jeff Law  <law@redhat.com>
13930         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
13931         the H8/300, H8/300H and H8/S variants into a single pattern.
13932         (movhi_h8300, movqi_h8300hs): Similarly.
13933         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
13934         (QHI mode iterator): New.
13936         * config/h8300/h8300.md: Remove trailing whitespace.
13938 2018-07-02  Jim Wilson  <jimw@sifive.com>
13940         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
13941         instead of emit_insn for interrupt returns.
13942         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
13943         (riscv_sret, riscv_uret): Likewise.
13945 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13947         * pretty-print.c (selftest::test_pp_format): Move save and restore
13948         of quotes to class auto_fix_quotes, and add an instance.
13949         * selftest.c: Include "intl.h".
13950         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
13951         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
13952         * selftest.h (selftest::auto_fix_quotes): New class.
13954 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
13956         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
13957         (aarch64_sve_prepare_conditional_op): Remove.
13958         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
13959         Allow aarch64_simd_reg_or_zero as select operand; remove
13960         the aarch64_sve_prepare_conditional_op call.
13961         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
13962         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13963         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
13964         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
13965         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
13966         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
13967         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
13968         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
13969         and a splitters to match all of the *_any patterns.
13970         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
13972         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
13973         (SVE_COND_FP_BINARY_REV): Remove.
13974         (sve_int_op_rev, sve_fp_op_rev): New.
13975         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
13976         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
13977         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
13978         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
13979         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
13980         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
13981         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
13982         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
13983         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
13985         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
13986         Remove match_dup 1 from the inner unspec.
13987         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13989         * config/aarch64/aarch64.md (movprfx): New attr.
13990         (length): Default movprfx to 8.
13991         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
13992         (*madd<SVE_I>, *msub<SVE_I): Likewise.
13993         (*<su>mul<SVE_I>3_highpart): Likewise.
13994         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
13995         (*v<ASHIFT><SVE_I>3): Likewise.
13996         (*<su><MAXMIN><SVE_I>3): Likewise.
13997         (*<su><MAXMIN><SVE_F>3): Likewise.
13998         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
13999         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
14000         (*div<SVE_F>4): Likewise.
14002 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
14004         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
14005         in dump string.
14007 2018-07-02  Richard Biener  <rguenther@suse.de>
14009         PR tree-optimization/86363
14010         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
14011         memset argument refers to a non-variable address.
14013 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
14015         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
14016         {PLUS,MINUS}_EXPR code to...
14017         (adjust_symbolic_bound): ...here,
14018         (combine_bound): ...here,
14019         (set_value_range_with_overflow): ...and here.
14021 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
14023         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
14024         code...
14025         (extract_range_from_abs_expr): ...here.
14027 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
14029         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
14030         -fno-omit-frame-pointer when not optimizing.
14032 2018-07-02  Martin Liska  <mliska@suse.cz>
14034         PR ipa/86279
14035         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
14036         (propagate_nothrow): Likewise.
14038 2018-07-02  Martin Liska  <mliska@suse.cz>
14040         PR ipa/86323
14041         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
14043 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
14045         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
14046         function in r262149, changing "loc" param from source_location to
14047         const dump_location_t &.
14048         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
14049         declaration, as above.
14051 2018-07-01  Paul Koning  <ni1d@arrl.net>
14053         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14054         -munit-asm, -mgnu-asm, -mdec-asm.
14055         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
14056         (pdp11_output_labelref): New.
14057         (pdp11_output_def): New.
14058         (pdp11_output_addr_vec_elt): New.
14059         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
14060         %# and %@ format codes.
14061         (pdp11_option_override): New.
14062         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
14063         (pdp11_output_ident): New.
14064         (pdp11_asm_named_section): New.
14065         (pdp11_asm_init_sections): New.
14066         (pdp11_file_start): New.
14067         (pdp11_file_end): New.
14068         (output_ascii): Use .ascii/.asciz for -mdec-asm.
14069         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
14070         %o, like %c but octal.
14071         (pdp11_option_override): New.
14072         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
14073         -mdec-asm.
14074         (DATA_SECTION_ASM_OP): Ditto.
14075         (READONLY_DATA_SECTION_ASM_OP): New.
14076         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
14077         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
14078         (ASM_OUTPUT_LABELREF): Ditto.
14079         (ASM_OUTPUT_DEF): Ditto.
14080         (ASM_OUTPUT_EXTERNAL): New.
14081         (ASM_OUTPUT_SOURCE_FILENAME): New.
14082         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
14083         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
14084         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
14085         %# and %@ format codes.
14086         * config/pdp11/pdp11.opt (mgnu-asm): New.
14087         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
14088         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
14089         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
14091 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
14093         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
14094         dereferencing path[] beyond its length.
14095         (debug_path): New.
14096         (debug_all_paths): New.
14097         (rewire_first_differing_edge): New.
14098         (adjust_paths_after_duplication): New.
14099         (duplicate_thread_path): Call adjust_paths_after_duplication.
14100         Add new argument.
14101         (thread_through_all_blocks): Add new argument to
14102         duplicate_thread_path.
14104 2018-06-30  Jim Wilson  <jimw@sifive.com>
14106         * config/riscv/predicates.md (p2m1_shift_operand): New.
14107         (high_mask_shift_operand): New.
14108         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
14109         pattern using p2m1_shift_operand.
14110         (lshsi3_zero_extend_3+2): New combiner pattern using
14111         high_mask_shift_operand.
14113 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14115         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
14116         split out from...
14117         (vect_recog_rotate_pattern): ...here.
14118         (vect_convert_input): Try to insert casts of invariants in the
14119         preheader.
14120         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
14121         preheader to be empty.
14123 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14125         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
14126         vector type.  If given, install it in the new statement's
14127         STMT_VINFO_VECTYPE.
14128         (vect_element_precision): New function.
14129         (vect_unpromoted_value): New struct.
14130         (vect_unpromoted_value::vect_unpromoted_value): New function.
14131         (vect_unpromoted_value::set_op): Likewise.
14132         (vect_look_through_possible_promotion): Likewise.
14133         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
14134         (vect_widened_op_tree, vect_convert_input): Likewise.
14135         (vect_convert_inputs, vect_convert_output): Likewise.
14136         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
14137         to handle the optional cast of the multiplication result and
14138         vect_widened_op_tree to detect the widened multiplication itself.
14139         Do not require the input and output of promotion casts to have
14140         the same sign, but base the signedness of the operation on the
14141         input rather than the result.  If the pattern includes two
14142         promotions, check that those promotions have the same sign.
14143         Do not restrict the MULT_EXPR handling to a double-width result;
14144         handle quadruple-width results and wider.  Use vect_convert_inputs
14145         to convert the inputs to the common type.
14146         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
14147         to handle the optional cast of the ABS result.  Also allow a sign
14148         change or a sign extension between the ABS and MINUS.
14149         Use vect_widened_op_tree to detect the widened subtraction and use
14150         vect_convert_inputs to convert the inputs to the common type.
14151         (vect_handle_widen_op_by_const): Delete.
14152         (vect_recog_widen_op_pattern): New function.
14153         (vect_recog_widen_mult_pattern): Use it.
14154         (vect_recog_widen_shift_pattern): Likewise.
14155         (vect_recog_widen_sum_pattern): Use
14156         vect_look_through_possible_promotion to handle the promoted
14157         PLUS_EXPR operand.
14159 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14161         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
14162         the containing gimple_seq *.
14163         * gimple-iterator.h (gsi_for_stmt): Declare it.
14164         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
14165         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
14166         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
14167         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
14168         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
14169         checks.
14170         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
14171         split out from...
14172         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
14173         statement being replaced is part of an existing pattern
14174         definition sequence, inserting the new pattern statements before
14175         the original one.
14176         (vect_pattern_recog_1): Don't return a bool.  If the statement
14177         is already part of a pattern, instead apply pattern matching
14178         to the pattern definition statements.  Don't clear the
14179         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
14180         (vect_pattern_recog): Don't break after the first match;
14181         continue processing the pattern definition statements instead.
14182         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
14184 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14186         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14187         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14188         (vect_recog_widen_sum_pattern): Use it.
14190 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14192         * tree-vect-loop.c (vectorizable_reduction): Assert that the
14193         phi is not a pattern statement and has not been replaced by
14194         a pattern statement.
14195         * tree-vect-patterns.c (type_conversion_p): Don't check
14196         STMT_VINFO_IN_PATTERN_P.
14197         (vect_recog_vector_vector_shift_pattern): Likewise.
14198         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
14199         the pattern statement rather than the original statement; check
14200         directly for a WIDEN_MULT_EXPR here.
14201         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
14202         vect_is_simple_use to return the pattern statement rather
14203         than the original statement; use is_pattern_stmt_p to check
14204         for such a pattern statement.
14205         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
14206         to return the pattern statement rather than the original statement;
14207         don't do the same transformation here.
14208         (vect_is_simple_use): If the defining statement has been replaced
14209         by a pattern statement, return the pattern statement instead.
14210         Remove the corresponding (local) transformation from the vectype
14211         overload.
14213 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14215         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
14216         end and default to null.
14217         * tree-vect-loop.c (vect_create_epilog_for_reduction)
14218         (vectorizable_reduction): Update calls accordingly, dropping the
14219         gimple ** argument if the passed-back statement isn't needed.
14220         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
14221         (vect_recog_rotate_pattern): Likewise.
14222         (vect_recog_mask_conversion_pattern): Likewise.
14223         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
14224         (vect_mask_constant_operand_p): Likewise.
14225         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
14226         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
14227         (get_group_load_store_type, get_load_store_type): Likewise.
14228         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
14229         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
14230         (vectorizable_conversion, vectorizable_assignment): Likewise.
14231         (vectorizable_shift, vectorizable_operation): Likewise.
14232         (vectorizable_store, vect_is_simple_cond): Likewise.
14233         (vectorizable_condition, vectorizable_comparison): Likewise.
14234         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
14235         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
14236         and move it to the end.  Cope with null def_stmt_outs.
14238 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14240         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
14242 2018-06-29  Jeff Law  <law@redhat.com>
14244         * config/v850/v850.c (v850_legitimate_address_p): Handle large
14245         displacements for TARGET_V850E2V3 and newer.
14246         (TARGET_LRA_P): Remove.  Defaults to LRA now.
14247         * config/v850/v850.md (sign23byte_load): Remove.
14248         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
14249         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
14251 2018-06-29  Martin Liska  <mliska@suse.cz>
14253         PR lto/85759
14254         * coverage.c (coverage_init): Mangle full path name.
14255         * doc/invoke.texi: Document the change.
14256         * gcov-io.c (mangle_path): New.
14257         * gcov-io.h (mangle_path): Likewise.
14258         * gcov.c (mangle_name): Use mangle_path for path mangling.
14260 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14262         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
14263         if starting source register is not even.
14265 2018-06-29  Martin Liska  <mliska@suse.cz>
14267         PR tree-optimization/86263
14268         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
14269         Make edge redirection.
14271 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
14273         * dumpfile.c (dump_loc): Add indentation based on scope depth.
14274         (dump_scope_depth): New variable.
14275         (get_dump_scope_depth): New function.
14276         (dump_begin_scope): New function.
14277         (dump_end_scope): New function.
14278         * dumpfile.h (get_dump_scope_depth): New declaration.
14279         (dump_begin_scope): New declaration.
14280         (dump_end_scope): New declaration.
14281         (class auto_dump_scope): New class.
14282         (AUTO_DUMP_SCOPE): New macro.
14283         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
14284         AUTO_DUMP_SCOPE.
14286 2018-06-29  Richard Biener  <rguenther@suse.de>
14288         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
14289         compute_all_dependences succeeds.
14290         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
14291         exceed --param loop-max-datarefs-for-datadeps.
14293 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
14295         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
14297 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
14299         PR target/86348
14300         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
14301         alternative 0 in preferred_for_speed attribute.
14303 2018-06-28  Paul Koning  <ni1d@arrl.net>
14305         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
14306         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
14307         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
14308         * config/pdp11/pdp11.md: Correct "length" attribute calculation
14309         for shift insn patterns.
14311 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
14313         * cgraph.c (cgraph_node::get_body): Replace assignments to
14314         "dump_file" with calls to set_dump_file.
14315         * dumpfile.c (alt_dump_file): Make static, and group with...
14316         (alt_flags): ...this definition.
14317         (dumps_are_enabled): New variable.
14318         (refresh_dumps_are_enabled): New function.
14319         (set_dump_file): New function.
14320         (set_alt_dump_file): New function.
14321         (gcc::dump_manager::dump_start): Replace assignments to
14322         "dump_file" and "alt_dump_file" with calls to set_dump_file and
14323         set_alt_dump_file.
14324         (gcc::dump_manager::dump_finish): Likewise.
14325         * dumpfile.h (alt_dump_file): Delete decl.
14326         (dumps_are_enabled): New variable decl.
14327         (set_dump_file): New function decl.
14328         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
14329         global.
14330         * tree-nested.c (lower_nested_functions): Replace assignments to
14331         "dump_file" with calls to set_dump_file.
14333 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14335         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
14336         goto_locus of each outgoing edge of each basic block.
14338 2018-06-28  Richard Biener  <rguenther@suse.de>
14340         * dwarf2out.c (decl_scope_table): Remove.
14341         (push_decl_scope): Likewise.
14342         (pop_decl_scope): Likewise.
14343         (gen_type_die_for_member): Do not call push/pop_decl_scope.
14344         (gen_struct_or_union_type_die): Likewise.
14345         (gen_tagged_type_die): Likewise.
14346         (dwarf2out_init): Do not initialize decl_scope_table.
14347         (dwarf2out_c_finalize): Do not free it.
14349 2018-06-28  Richard Biener  <rguenther@suse.de>
14351         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
14352         deciding whether to not re-use a DIE.
14354 2018-06-28  Richard Biener  <rguenther@suse.de>
14356         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
14357         DW_AT_abstract_origin attribute.
14359 2018-06-28  Martin Liska  <mliska@suse.cz>
14361         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
14362         Use newly introduced constants.
14363         * tree-switch-conversion.h (struct jump_table_cluster):
14364         Define max_ratio_for_size and max_ratio_for_speed.
14366 2018-06-28  Martin Liska  <mliska@suse.cz>
14368         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14369         Add new checking assert to catch invalid state.
14370         (jump_table_cluster::can_be_handled): Handle single case
14371         clusters.
14372         (jump_table_cluster::is_beneficial): Bail out for such case.
14373         (bit_test_cluster::find_bit_tests):
14374         Add new checking assert to catch invalid state.
14375         (bit_test_cluster::can_be_handled): Handle single case
14376         clusters.
14377         (bit_test_cluster::is_beneficial): Bail out for such case.
14378         (switch_decision_tree::analyze_switch_statement):
14379         Fix comment.
14381 2018-06-28  Martin Liska  <mliska@suse.cz>
14383         * common.opt: Introduce -completion option.
14384         * gcc.c (driver_handle_option): Handle it.
14385         (driver::main): Print completions if completion
14386         is set.
14387         * opt-suggestions.c (option_proposer::get_completions):
14388         New function.
14389         (option_proposer::suggest_completion): Likewise.
14390         (option_proposer::find_param_completions): Likewise.
14391         (verify_autocompletions): Likewise.
14392         (test_completion_valid_options): Likewise.
14393         (test_completion_valid_params): Likewise.
14394         (in_completion_p): Likewise.
14395         (empty_completion_p): Likewise.
14396         (test_completion_partial_match): Likewise.
14397         (test_completion_garbage): Likewise.
14398         (opt_proposer_c_tests): Likewise.
14399         * opt-suggestions.h: Declare new functions.
14400         * opts.c (common_handle_option): Handle OPT__completion_.
14401         * selftest-run-tests.c (selftest::run_tests): Add
14402         opt_proposer_c_tests.
14403         * selftest.c (assert_str_startswith): New.
14404         * selftest.h (assert_str_startswith): Likewise.
14405         (opt_proposer_c_tests): New.
14406         (ASSERT_STR_STARTSWITH): Likewise.
14408 2018-06-28  Martin Liska  <mliska@suse.cz>
14410         * Makefile.in: Add opt-suggestions.o.
14411         * gcc-main.c: Include opt-suggestions.h.
14412         * gcc.c (driver::driver): Likewise.
14413         (driver::~driver): Remove m_option_suggestions.
14414         (driver::build_option_suggestions): Moved to option_proposer.
14415         (driver::suggest_option): Likewise.
14416         (driver::handle_unrecognized_options): Use option_proposer.
14417         * gcc.h (class driver): Add new memver m_option_proposer.
14418         * opt-suggestions.c: New file.
14419         * opt-suggestions.h: New file.
14421 2018-06-28  Martin Liska  <mliska@suse.cz>
14423         * vec.h (class auto_string_vec): New (moved from auto_argvec).
14424         (auto_string_vec::~auto_string_vec): Likewise.
14426 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14428         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
14429         prevent_decl_creation_for_types fields up and add reset_location field.
14430         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
14431         statement if id->reset_location is true.
14432         (copy_edges_for_bb): Do not set goto_locus on the new edges if
14433         id->reset_location is true.
14434         (copy_phis_for_bb): Force input_location on the arguments if
14435         id->reset_location is true.
14436         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
14437         is set on the function to be inlined.
14439 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
14441         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
14443 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
14445         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
14446         registers for Pmode.
14447         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
14448         hard registers for the clobbered pseudo.
14450 2018-06-27  Paul Koning  <ni1d@arrl.net>
14452         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14453         mutually exclusive options.
14454         * config/pdp11/constraints.md (h): New constraint.
14455         (O): Update definition to match shift code generation.
14456         (D): New constraint.
14457         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
14458         (CCFP): Remove.
14459         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
14460         function.
14461         (output_jump): Change arguments.
14462         (pdp11_fixed_cc_regs): New function.
14463         (pdp11_cc_mode): Ditto.
14464         (pdp11_expand_shift): Ditto.
14465         (pdp11_assemble_shift): Ditto.
14466         (pdp11_small_shift): Ditto.
14467         (pdp11_branch_cost): Remove.
14468         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
14469         from output.
14470         (pdp11_register_move_cost): Update for CC registers.
14471         (pdp11_rtx_costs): Add case for LSHIFTRT.
14472         (pdp11_output_jump): Add CCNZ mode conditional branches.
14473         (notice_update_cc_on_set): Remove.
14474         (pdp11_cc_mode): New function.
14475         (simple_memory_operand): Correct pre/post decrement case.
14476         (no_side_effect_operand): New function.
14477         (pdp11_regno_reg_class): Add CC_REGS class.
14478         (pdp11_fixed_cc_regs): New function.
14479         (pdp11_small_shift): New function.
14480         (pdp11_expand_shift): New function to expand shift insns.
14481         (pdp11_assemble_shift): New function to output shifts.
14482         (pdp11_branch_cost): Remove.
14483         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
14484         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
14485         (WCHAR_TYPE): Ditto.
14486         (PTRDIFF_TYPE): Ditto.
14487         (ADJUST_INSN_LENGTH): New macro.
14488         (FIXED_REGISTERS): Add CC registers.
14489         (CALL_USED_REGISTERS): Ditto.
14490         (reg_class): Ditto.
14491         (REG_CLASS_NAMES): Ditto.
14492         (REG_CLASS_CONTENTS): Ditto.
14493         (SELECT_CC_MODE): Use new function.
14494         (TARGET_FLAGS_REGNUM): New macro.
14495         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
14496         (cc0_reg_rtx): Remove.
14497         (CC_STATUS_MDEP): Remove.
14498         (CC_STATUS_MDEFP_INIT): Remove.
14499         (CC_IN_FPU): Remove.
14500         (NOTICE_UPDATE_CC): Remove.
14501         (REGISTER_NAMES): Add CC registers.
14502         (BRANCH_COST): Change to constant 1.
14503         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
14504         handling.
14505         * config/pdp11/pdp11.opt (mbcopy): Remove.
14506         (mbcopy-builtin): Remove.
14507         (mbranch-cheap): Remove.
14508         (mbranch-expensive): Remove.
14509         * config/pdp11/predicates.md (expand_shift_operand): Update to
14510         match shift code generation.
14511         (ccnz_operator): New predicate.
14512         * doc/invoke.texi (PDP-11 Options): Remove deleted options
14513         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
14514         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
14515         exclusive options.
14516         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
14517         description of O constraint.
14519 2018-06-27  Jeff Law  <law@redhat.com>
14520             Austin Law  <austinklaw@gmail.com>
14522         * config/v850/v850.md (addsi3_set_flags): New pattern.
14523         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
14524         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
14525         (zero_extendhisi2_v850_set_flags): Likewise.
14526         (zero_extendqisi2_v850_set_flags): Likewise.
14527         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
14528         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
14529         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
14531         * config/v850/v850-protos.h (notice_update_cc): Remove.
14532         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
14533         (v850_print_operand): Handle 'D' and "d".
14534         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
14535         Add handling of arithmetic/logical operations compared against zero.
14536         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
14537         Do not look at v850_compare_op, instead get mode from last argument.
14538         (v850_gen_compare): Remove
14539         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
14540         after reload for prologue insns.
14541         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
14542         patterns.
14543         (construct_save_jarl): Likewise.
14544         (TARGET_FLAGS_REGNUM): Define.
14545         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
14546         (NOTICE_UPDATE_CC): Remove.
14547         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
14548         than cc0.  Conditionalize on reload_completed.
14549         (cmpsi_insn, setfcc_insn): Likewise.
14550         (tst1 splitter): Turn into define_and_split which sets the flags
14551         after reload.
14552         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
14553         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
14554         (cstoresf4, cstoredf4): Clobber the flags.
14555         (cmpsi, cmpsf, cmpdf): Remove expanders.
14556         (setf_insn): Remove pattern.
14557         (addsi3): Turn into define_and_split which clobbers the flags after
14558         reload and a suitable pattern (addsi3_clobber_flags) for use after
14559         reload.
14560         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
14561         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
14562         (ashrsi3, ashrsi3_v850e2): Likewise.
14563         (bins): Clobber the flags.
14564         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
14565         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
14566         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
14567         (call_value_internal_short, call_value_internal_long): Likewise.
14568         (callt_save_interrupt, callt_return_interrupt): Likewise.
14569         (save_interrupt, return_interrupt): Likewise.
14570         (callt_save_all_interrupt, save_all_interrupt): Likewise.
14571         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
14572         (restore_all_interrupt, _restore_all_interrupt): Likewise.
14573         (All FP comparisons): Only allow after reload has completed.
14574         (trfsr): Likewise.
14575         (divh, divhu): Tweak output template.
14576         (branch_z_normal, branch_z_invert): Remove
14577         (branch_nz_normal, branch_nz_invert): Likewise.
14578         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
14580         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
14581         * config/v850/v850.c (notice_update_cc): Remove.
14582         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
14583         (CC_NO_CARRY): Likewise.
14584         (NOTICE_UPDATE_CC): Define to nothing.
14585         * config/v850/v850.md: Remove block comment on cc0 handling
14586         Remove "cc" attribute from all patterns.  Remove cc_status handling
14587         from all patterns.  Minor formatting fixes.
14589 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14591         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
14592         (cortex-a76.cortex-a55): Likewise.
14593         * config/aarch64/aarch64-tune.md: Regenerate.
14594         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
14595         cortex-a76.cortex-a55.
14597 2018-06-27  Jeff Law  <law@redhat.com>
14599         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
14600         (MULTILIB_DIRNAMES): Similarly.
14602 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
14604         * gimple.h (gimple_return_retbnd): Delete.
14605         (gimple_return_set_retbnd): Likewise.
14606         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
14607         gimple_return_set_retbnd.
14608         * gimple-pretty-print.c (dump_gimple_return): Remove call to
14609         gimple_return_retbnd and adjust.
14610         * tree-inline.h (struct copy_body_data): Remove retbnd field.
14611         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
14612         Explicitly return NULL in a couple more cases.  Move assertion
14613         on debug statements and remove unreachable code.
14614         (reset_debug_binding): Do not test id->retbnd.
14615         (expand_call_inline): Do not set it.
14617 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
14619         * configure.ac: Add --disable-gcov option.
14620         * configure: Regenerate.
14621         * Makefile.in: Honour @enable_gcov@.
14622         * doc/install.texi: Document --disable-gcov.
14624 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14626         * config/arm/arm-cpus.in (cortex-a76): New entry.
14627         (cortex-a76.cortex-a55): Likewise.
14628         * config/arm/arm-tables.opt: Regenerate.
14629         * config/arm/arm-tune.md: Likewise.
14630         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
14631         * doc/invoke.texi (ARM Options): Document cortex-a76 and
14632         cortex-a76.cortex-a55.
14634 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
14636         PR target/85769
14637         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
14639 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
14641         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
14642         comment.
14643         (EPILOGUE_USES): Likewise.
14645 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
14647         * tree-inline.c (remap_location): New function extracted from...
14648         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
14649         (copy_phis_for_bb): ...here.  Call remap_location.
14650         (copy_cfg_body): Adjust call to copy_edges_for_bb.
14652 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
14654         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
14655         unaligned vsx for 16B memset.
14657 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
14659         PR target/86285
14660         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
14661         ieee128_float_type_node to long_double_type_node unless
14662         TARGET_LONG_DOUBLE_128 is set.
14664 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
14666         * cfgloop.c (get_loop_location): Convert return type from
14667         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
14668         by implicit construction from rtx_insn *, and using
14669         dump_user_location_t::from_function_decl for the fallback case.
14670         * cfgloop.h (get_loop_location): Convert return type from
14671         location_t to dump_user_location_t.
14672         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
14673         dump_printf_loc to pass in a dump_location_t rather than a
14674         location_t, via the gimple stmt.
14675         * coverage.c (get_coverage_counts): Update calls to
14676         dump_printf_loc to pass in dump_location_t rather than a
14677         location_t.
14678         * doc/optinfo.texi (Dump types): Convert example of
14679         dump_printf_loc from taking "locus" to taking "insn".  Update
14680         description of the "_loc" calls to cover dump_location_t.
14681         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
14682         "selftest.h".
14683         (dump_user_location_t::dump_user_location_t): New constructors,
14684         from gimple *stmt and rtx_insn *.
14685         (dump_user_location_t::from_function_decl): New function.
14686         (dump_loc): Make static.
14687         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
14688         const dump_location_t &.
14689         (dump_generic_expr_loc): Delete.
14690         (dump_printf_loc): Convert param "loc" from location_t to
14691         const dump_location_t &.
14692         (selftest::test_impl_location): New function.
14693         (selftest::dumpfile_c_tests): New function.
14694         * dumpfile.h: Include "profile-count.h".
14695         (class dump_user_location_t): New class.
14696         (struct dump_impl_location_t): New struct.
14697         (class dump_location_t): New class.
14698         (dump_printf_loc): Convert 2nd param from source_location to
14699         const dump_location_t &.
14700         (dump_generic_expr_loc): Delete.
14701         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
14702         const dump_location_t &.
14703         * gimple-fold.c (fold_gimple_assign): Update call to
14704         dump_printf_loc to pass in a dump_location_t rather than a
14705         location_t, via the gimple stmt.
14706         (gimple_fold_call): Likewise.
14707         * gimple-loop-interchange.cc
14708         (loop_cand::analyze_iloop_reduction_var): Update for change to
14709         check_reduction_path.
14710         (tree_loop_interchange::interchange): Update for change to
14711         find_loop_location.
14712         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
14713         change in return-type of find_loop_location.
14714         (graphite_regenerate_ast_isl): Likewise.
14715         * graphite-optimize-isl.c (optimize_isl): Likewise.
14716         * graphite.c (graphite_transform_loops): Likewise.
14717         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
14718         pass in a dump_location_t rather than a location_t, via the
14719         gimple stmt.
14720         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14721         * ipa.c (walk_polymorphic_call_targets): Likewise.
14722         * loop-unroll.c (report_unroll): Convert "locus" param from
14723         location_t to dump_location_t.
14724         (decide_unrolling): Update for change to get_loop_location's
14725         return type.
14726         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
14727         location_t to dump_user_location_t.
14728         (grid_find_single_omp_among_assignments_1): Updates calls to
14729         dump_printf_loc to pass in a dump_location_t rather than a
14730         location_t, via the gimple stmt.
14731         (grid_parallel_clauses_gridifiable): Convert "tloc" from
14732         location_t to dump_location_t.  Updates calls to dump_printf_loc
14733         to pass in a dump_location_t rather than a location_t, via the
14734         gimple stmt.
14735         (grid_inner_loop_gridifiable_p): Likewise.
14736         (grid_dist_follows_simple_pattern): Likewise.
14737         (grid_gfor_follows_tiling_pattern): Likewise.
14738         (grid_target_follows_gridifiable_pattern): Likewise.
14739         (grid_attempt_target_gridification): Convert initialization
14740         of local "grid" from memset to zero-initialization; FIXME: does
14741         this require C++11?  Update call to dump_printf_loc to pass in a
14742         optinfo_location rather than a location_t, via the gimple stmt.
14743         * profile.c (read_profile_edge_counts): Updates call to
14744         dump_printf_loc to pass in a dump_location_t rather than a
14745         location_t
14746         (compute_branch_probabilities): Likewise.
14747         * selftest-run-tests.c (selftest::run_tests): Call
14748         dumpfile_c_tests.
14749         * selftest.h (dumpfile_c_tests): New decl.
14750         * tree-loop-distribution.c (pass_loop_distribution::execute):
14751         Update for change in return type of find_loop_location.
14752         * tree-parloops.c (parallelize_loops): Likewise.
14753         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
14754         "locus" from location_t to dump_user_location_t.
14755         (canonicalize_loop_induction_variables): Likewise.
14756         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
14757         for change in return type of find_loop_location.
14758         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
14759         to dump_printf_loc to pass in a dump_location_t rather than a
14760         location_t, via the stmt.
14761         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
14762         Likewise.
14763         * tree-vect-loop-manip.c (find_loop_location): Convert return
14764         type from source_location to dump_user_location_t.
14765         (vect_do_peeling): Update for above change.
14766         (vect_loop_versioning): Update for change in type of
14767         vect_location.
14768         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
14769         from location_t to dump_user_location_t.
14770         (vect_estimate_min_profitable_iters): Update for change in type
14771         of vect_location.
14772         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
14773         location_t to dump_location_t.
14774         (vect_slp_bb): Update for change in type of vect_location.
14775         * tree-vectorizer.c (vect_location): Convert from source_location
14776         to dump_user_location_t.
14777         (try_vectorize_loop_1): Update for change in vect_location's type.
14778         (vectorize_loops): Likewise.
14779         (increase_alignment): Likewise.
14780         * tree-vectorizer.h (vect_location): Convert from source_location
14781         to dump_user_location_t.
14782         (find_loop_location): Convert return type from source_location to
14783         dump_user_location_t.
14784         (check_reduction_path): Convert 1st param from location_t to
14785         dump_user_location_t.
14786         * value-prof.c (check_counter): Update call to dump_printf_loc to
14787         pass in a dump_user_location_t rather than a location_t; update
14788         call to error_at for change in type of "locus".
14789         (check_ic_target): Update call to dump_printf_loc to
14790         pass in a dump_user_location_t rather than a location_t, via the
14791         call_stmt.
14793 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
14795         * config/s390/s390.h (enum processor_flags): Do not use
14796         default tune parameter when -march was specified.
14798 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
14800         PR target/86314
14801         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
14802         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
14803         operands.
14805 2018-06-26  Richard Biener  <rguenther@suse.de>
14807         PR tree-optimization/86287
14808         PR bootstrap/86316
14809         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
14810         (vect_analyze_loop): Initialize n_stmts.
14812 2018-06-26  Richard Biener  <rguenther@suse.de>
14814         PR middle-end/86271
14815         * fold-const.c (fold_convertible_p): Pointer extension
14816         isn't valid.
14818 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
14820         PR debug/86064
14821         * dwarf2out.c (loc_list_has_views): Adjust comments.
14822         (dw_loc_list): Split single cross-partition range with
14823         nonzero locview.
14825 2018-06-25  Jeff Law  <law@redhat.com>
14827         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
14828         on -mbig-switch by default.
14830         * config/v850/predicates.md (const_float_1_operand): Fix match_code
14831         test.
14832         (const_float_0_operand): Remove unused predicate.
14833         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
14834         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
14835         (recipsf2): New expander.  Original pattern now called
14836         (recipsf2_insn).
14837         (recipdf2, recipdf2_insn): Similarly.
14838         (rsqrtsf2, rsqrtsf2_insn): Similarly
14839         (rsqrtdf2, rsqrtdf2_insn): Similarly
14841 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
14843         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
14844         Simplify logic for FreeBSD (twice).
14846 2018-06-25  Martin Sebor  <msebor@redhat.com>
14848         PR tree-optimization/86204
14849         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
14850         a strnlen result if it's less than the length of the string.
14852 2018-06-25  Martin Sebor  <msebor@redhat.com>
14854         PR tree-optimization/85700
14855         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
14856         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
14857         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
14859 2018-06-25  Martin Sebor  <msebor@redhat.com>
14861         * doc/extend.texi (Zero-length arrays): Update and clarify.
14863 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
14865         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
14866         added IEEE/IBM long double multilib support on PowerPC little
14867         endian Linux systems.
14868         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
14869         (MULTILIB_DEFAULTS): Likewise.
14870         * config/rs6000/rs6000.c (rs6000_option_override_internal):
14871         Likewise.
14872         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
14873         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
14874         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
14876 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
14878         PR middle-end/86311
14879         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
14880         (REORDER_45): Likewise.
14882 2018-06-25  Jeff Law  <law@redhat.com>
14884         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
14885         dividend to 32 bits.  Adjust length.
14886         (udivmodhi4): Cleanup output template.  Fix length.
14888 2018-06-25  Carl Love  <cel@us.ibm.com>
14890         * config/rs6000/vsx.md: Change word selector to prefered location.
14892 2018-06-25  Richard Biener  <rguenther@suse.de>
14894         PR tree-optimization/86304
14895         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
14896         epilogue-if-converted loops as well.
14898 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
14900         * lto-section-out.c (lto_begin_section): Do not print section
14901         name for noaddr and unnumbered dumps.
14903 2018-06-25  Richard Biener  <rguenther@suse.de>
14905         * tree-vectorizer.h (struct vec_info_shared): New structure
14906         with parts split out from struct vec_info and loop_nest from
14907         struct _loop_vec_info.
14908         (struct vec_info): Adjust accordingly.
14909         (struct _loop_vec_info): Likewise.
14910         (LOOP_VINFO_LOOP_NEST): Adjust.
14911         (LOOP_VINFO_DATAREFS): Likewise.
14912         (LOOP_VINFO_DDRS): Likewise.
14913         (struct _bb_vec_info): Likewise.
14914         (BB_VINFO_DATAREFS): Likewise.
14915         (BB_VINFO_DDRS): Likewise.
14916         (struct _stmt_vec_info): Add dr_aux member.
14917         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
14918         (DR_MISALIGNMENT_UNINITIALIZED): New.
14919         (set_dr_misalignment): Adjust.
14920         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
14921         (vect_analyze_loop): Adjust prototype.
14922         (vect_analyze_loop_form): Likewise.
14923         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14924         Compute dependences lazily.
14925         (vect_record_base_alignments): Use shared datarefs/ddrs.
14926         (vect_verify_datarefs_alignment): Likewise.
14927         (vect_analyze_data_refs_alignment): Likewise.
14928         (vect_analyze_data_ref_accesses): Likewise.
14929         (vect_analyze_data_refs): Likewise.
14930         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
14931         constructor parameter for shared part.
14932         (vect_analyze_loop_form): Pass in shared part and adjust.
14933         (vect_analyze_loop_2): Pass in storage for the number of
14934         stmts.  Move loop nest finding to the caller.  Compute
14935         datarefs lazily.
14936         (vect_analyze_loop): Pass in shared part.
14937         (vect_transform_loop): Verify shared datarefs are unchanged.
14938         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
14939         constructor parameter for shared part.
14940         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
14941         (vect_slp_bb): Verify shared datarefs are unchanged before
14942         transform.
14943         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
14944         change.
14945         (new_stmt_vec_info): Initialize DR_AUX misalignment to
14946         DR_MISALIGNMENT_UNINITIALIZED.
14947         * tree-vectorizer.c (vec_info::vec_info): Add constructor
14948         parameter for shared part.
14949         (vec_info::~vec_info): Adjust.
14950         (vec_info_shared::vec_info_shared): New.
14951         (vec_info_shared::~vec_info_shared): Likewise.
14952         (vec_info_shared::save_datarefs): Likewise.
14953         (vec_info_shared::check_datarefs): Likewise.
14954         (try_vectorize_loop_1): Construct shared part live for analyses
14955         of a single loop for multiple vector sizes.
14956         * tree-parloops.c (gather_scalar_reductions): Adjust.
14958 2018-06-25  Richard Biener  <rguenther@suse.de>
14960         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
14961         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
14962         (vect_analyze_data_refs): Remove similar code from here and
14963         simplify accordingly.
14965 2018-06-25  Richard Biener  <rguenther@suse.de>
14967         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
14968         for reverse storage order accesses rather than asserting
14969         they cannot happen here.
14971 2018-06-25  Tom de Vries  <tdevries@suse.de>
14973         PR debug/86257
14974         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
14975         Use data16 instead of .byte for insn prefix.
14977 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
14979         PR C++/86082
14980         * parser.c (make_char_string_pack): Pass this literal chars
14981         through cpp_interpret_string.
14982         (cp_parser_userdef_numeric_literal): Check the result of
14983         make_char_string_pack.
14985 2018-06-24  Maya Rashish  <coypu@sdf.org>
14987         * ginclude/stddef.h: Simplify conditions around avoiding
14988         re-definition of __size_t.
14990 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14992         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
14993         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
14995 2018-06-22  Maya Rashish  <coypu@sdf.org>
14997         * doc/invoke.texi (mno-fancy-math-387): Update for changes
14998         made to OpenBSD and NetBSD through the years.
15000 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15002         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
15003         behavior of vec_pack (vector double, vector double) to match
15004         behavior of vec_float2 (vector double, vector double).
15006 2018-06-22  Olivier Hainque  <hainque@adacore.com>
15008         * gimplify.c (gimplify_function_tree): Prevent creation
15009         of a trampoline for the address of the current function
15010         passed to entry/exit instrumentation hooks.
15012 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
15014         PR target/86222
15015         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
15016         correctly.
15018 2018-06-22  Martin Liska  <mliska@suse.cz>
15020         PR tree-optimization/86263
15021         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15022         Bail out if is_enabled is false.
15023         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
15024         New declaration.
15025         (jump_table_cluster::is_enabled): New function.
15027 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
15029         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
15030         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
15031         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
15032         (lto_input_ts_binfo_tree_pointers): Likewise.
15033         * tree-streamer-out.c (streamer_write_tree_bitfields,
15034         write_ts_binfo_tree_pointers): Likewise.
15035         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
15037 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
15039         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
15041 2018-06-22  Martin Liska  <mliska@suse.cz>
15043         * symbol-summary.h (get): Make it pure and inline move
15044         functionality from ::get function.
15045         (get): Remove and inline into ::get and ::get_create.
15046         (get_create): Move code from ::get function.
15048 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15050         PR target/85994
15051         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
15052         -x assembler-with-cpp.
15054 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15056         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
15057         _FILE_OFFSET_BITS=64 for C++.
15059 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
15061         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
15062         conversion insn that shows up when pr85657-3.c is compiled using
15063         IEEE 128-bit long double.
15064         (neg<mode>2_internal): Use the correct mode to check whether the
15065         mode is IBM extended.
15066         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
15067         multiply and divide external functions from being created more
15068         than once.
15070 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15072         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
15073         functions.
15074         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
15075         the edge can be forwarded.
15076         (cfg_layout_merge_blocks): Likewise.
15078 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15080         * except.c (finish_eh_generation): Commit edge insertions only after
15081         the EH edges have been redirected from post-landing to landing pads.
15083 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15085         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
15086         create_tmp_var_for to create the FRAME decl.
15087         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
15089 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15091         * tree-inline.c (copy_edges_for_bb): Minor tweak.
15092         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
15093         debug statement when resetting its value.
15094         (expand_call_inline): Copy the locus of the call onto the assignment
15095         of the return value, if any.  Use local variable in more cases.
15097 2018-06-21  Martin Liska  <mliska@suse.cz>
15099         * ipa-pure-const.c (propagate_nothrow): Use
15100         funct_state_summaries->get.
15101         (dump_malloc_lattice): Likewise.
15102         (propagate_malloc): Likewise.
15104 2018-06-21  Richard Biener  <rguenther@suse.de>
15106         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
15107         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
15108         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
15109         comment.
15110         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
15111         BLOCK_ABSTRACT_ORIGIN unconditionally.
15113 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
15115         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
15116         deleting it.
15117         * ipa-reference.c (ipa_reference_c_finalize): Delete
15118         ipa_ref_opt_sum_summaries and set it to NULL.
15120 2018-06-21  Tom de Vries  <tdevries@suse.de>
15122         PR tree-optimization/85859
15123         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
15124         test with comment from bb_no_side_effects_p.
15126 2018-06-21  Richard Biener  <rguenther@suse.de>
15128         PR tree-optimization/86232
15129         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
15130         max for constant niter.
15132 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15134         * config/aarch64/aarch64-simd.md
15135         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
15137 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15139         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
15140         Make opernads of the unspec commutative.
15142 2018-06-21  Richard Biener  <rguenther@suse.de>
15144         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
15145         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
15146         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
15147         (vect_analyze_data_ref_dependence): Re-order checks to deal with
15148         NULL DR_STEP.
15149         (vect_record_base_alignments): Do not record base alignment
15150         for gathers or scatters.
15151         (vect_compute_data_ref_alignment): Drop return value that is always
15152         true.  Bail out early for gathers or scatters.
15153         (vect_enhance_data_refs_alignment): Bail out early for gathers
15154         or scatters.
15155         (vect_find_same_alignment_drs): Likewise.
15156         (vect_analyze_data_refs_alignment): Remove dead code.
15157         (vect_slp_analyze_and_verify_node_alignment): Likewise.
15158         (vect_analyze_data_refs): For possible gathers or scatters do
15159         not create an alternate DR, just check their possible validity
15160         and mark them.  Adjust DECL_NONALIASED handling to not rely
15161         on DR_BASE_ADDRESS.
15162         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
15163         update inits of gathers or scatters.
15164         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
15165         Also copy gather/scatter flag to pattern vinfo.
15167 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15169         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
15170         behavior of vec_packsu (vector unsigned long long, vector unsigned
15171         long long) to match behavior of vec_packs with same signature.
15173 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
15174             Thomas Schwinge <thomas@codesourcery.com>
15175             Cesar Philippidis  <cesar@codesourcery.com>
15177         * gimplify.c (gimplify_scan_omp_clauses): Add support for
15178         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15179         (gimplify_adjust_omp_clauses): Likewise.
15180         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
15181         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
15182         (gimplify_omp_target_update): Update handling of acc update and
15183         enter/exit data.
15184         * omp-low.c (install_var_field): Remove unused parameter
15185         base_pointers_restrict.
15186         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
15187         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
15188         FINALIZE}
15189         (omp_target_base_pointers_restrict_p): Delete.
15190         (scan_omp_target): Update call to scan_sharing_clauses.
15191         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
15192         FINALIZE}.
15193         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
15194         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15195         (convert_local_omp_clauses): Likewise.
15196         * tree-pretty-print.c (dump_omp_clause): Likewise.
15197         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
15198         FINALIZE}.
15199         (omp_clause_code_name): Likewise.
15201 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
15203         PR debug/86194
15204         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
15205         be narrowed.
15207         PR tree-optimization/86231
15208         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
15209         anti-range don't overwrite *vr0min before using it to compute *vr0max.
15211 2018-06-20  Tom de Vries  <tdevries@suse.de>
15213         PR tree-optimization/86097
15214         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
15215         iv type if signedness of iv type is not the same as that of *nit.
15217 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
15219         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
15220         EDGE_EH edges, verify they are all EDGE_EH.
15222 2018-06-20  Maya Rashish  <coypu@sdf.org>
15224         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
15226 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15228         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
15229         * config/aarch64/aarch64.c (xgene1_tunings): Add
15230         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
15231         (aarch64_mode_valid_for_sched_fusion_p):
15232         Allow 16-byte modes.
15233         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
15234         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
15235         128-bit modes.
15236         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
15237         New pattern.
15238         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
15239         * config/aarch64/iterators.md (VQ2): New mode iterator.
15241 2018-06-20  Martin Liska  <mliska@suse.cz>
15243         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
15244         Change default ratio from 10 to 8.
15246 2018-06-20  Martin Liska  <mliska@suse.cz>
15248         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15249         New.
15250         (bit_test_cluster::find_bit_tests): Likewise.
15251         (switch_decision_tree::analyze_switch_statement): Find clusters.
15252         * tree-switch-conversion.h (struct jump_table_cluster): Document
15253         hierarchy.
15255 2018-06-20  Martin Liska  <mliska@suse.cz>
15257         * tree-switch-conversion.c (switch_conversion::collect):
15258         Record m_uniq property.
15259         (switch_conversion::expand): Bail out for special conditions.
15260         (group_cluster::~group_cluster): New.
15261         (group_cluster::group_cluster): Likewise.
15262         (group_cluster::dump): Likewise.
15263         (jump_table_cluster::emit): New.
15264         (switch_decision_tree::fix_phi_operands_for_edges): New.
15265         (struct case_node): Remove struct.
15266         (jump_table_cluster::can_be_handled): New.
15267         (case_values_threshold): Moved to header.
15268         (reset_out_edges_aux): Likewise.
15269         (jump_table_cluster::is_beneficial): New.
15270         (bit_test_cluster::can_be_handled): Likewise.
15271         (add_case_node): Remove.
15272         (bit_test_cluster::is_beneficial): New.
15273         (case_bit_test::cmp): New.
15274         (bit_test_cluster::emit): New.
15275         (expand_switch_as_decision_tree_p): Remove.
15276         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
15277         (fix_phi_operands_for_edge): Likewise.
15278         (switch_decision_tree::analyze_switch_statement): New.
15279         (compute_cases_per_edge): Move ...
15280         (switch_decision_tree::compute_cases_per_edge): ... here.
15281         (try_switch_expansion): Likewise.
15282         (switch_decision_tree::try_switch_expansion): Likewise.
15283         (record_phi_operand_mapping): Likewise.
15284         (switch_decision_tree::record_phi_operand_mapping): Likewise.
15285         (emit_case_decision_tree): Likewise.
15286         (switch_decision_tree::emit): Likewise.
15287         (balance_case_nodes): Likewise.
15288         (switch_decision_tree::balance_case_nodes): Likewise.
15289         (dump_case_nodes): Likewise.
15290         (switch_decision_tree::dump_case_nodes): Likewise.
15291         (emit_jump): Likewise.
15292         (switch_decision_tree::emit_jump): Likewise.
15293         (emit_cmp_and_jump_insns): Likewise.
15294         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
15295         (emit_case_nodes): Likewise.
15296         (switch_decision_tree::emit_case_nodes): Likewise.
15297         (conditional_probability): Remove.
15298         * tree-switch-conversion.h (enum cluster_type): New.
15299         (PRINT_CASE): New.
15300         (struct cluster): Likewise.
15301         (cluster::cluster): Likewise.
15302         (struct simple_cluster): Likewise.
15303         (simple_cluster::simple_cluster): Likewise.
15304         (struct group_cluster): Likewise.
15305         (struct jump_table_cluster): Likewise.
15306         (struct bit_test_cluster): Likewise.
15307         (struct min_cluster_item): Likewise.
15308         (struct case_tree_node): Likewise.
15309         (case_tree_node::case_tree_node): Likewise.
15310         (jump_table_cluster::case_values_threshold): Likewise.
15311         (struct case_bit_test): Likewise.
15312         (struct switch_decision_tree): Likewise.
15313         (struct switch_conversion): Likewise.
15314         (switch_decision_tree::reset_out_edges_aux): Likewise.
15316 2018-06-20  Martin Liska  <mliska@suse.cz>
15318         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
15319         (hoist_edge_and_branch_if_true): Likewise.
15320         (expand_switch_using_bit_tests_p): Likewise.
15321         (struct case_bit_test): Likewise.
15322         (case_bit_test_cmp): Likewise.
15323         (emit_case_bit_tests): Likewise.
15324         (switch_conversion::switch_conversion): New class.
15325         (struct switch_conv_info): Remove old struct.
15326         (collect_switch_conv_info): More to ...
15327         (switch_conversion::collect): ... this.
15328         (check_range): Likewise.
15329         (switch_conversion::check_range): Likewise.
15330         (check_all_empty_except_final): Likewise.
15331         (switch_conversion::check_all_empty_except_final): Likewise.
15332         (check_final_bb): Likewise.
15333         (switch_conversion::check_final_bb): Likewise.
15334         (create_temp_arrays): Likewise.
15335         (switch_conversion::create_temp_arrays): Likewise.
15336         (free_temp_arrays): Likewise.
15337         (gather_default_values): Likewise.
15338         (switch_conversion::gather_default_values): Likewise.
15339         (build_constructors): Likewise.
15340         (switch_conversion::build_constructors): Likewise.
15341         (constructor_contains_same_values_p): Likewise.
15342         (switch_conversion::contains_same_values_p): Likewise.
15343         (array_value_type): Likewise.
15344         (switch_conversion::array_value_type): Likewise.
15345         (build_one_array): Likewise.
15346         (switch_conversion::build_one_array): Likewise.
15347         (build_arrays): Likewise.
15348         (switch_conversion::build_arrays): Likewise.
15349         (gen_def_assigns): Likewise.
15350         (switch_conversion::gen_def_assigns): Likewise.
15351         (prune_bbs): Likewise.
15352         (switch_conversion::prune_bbs): Likewise.
15353         (fix_phi_nodes): Likewise.
15354         (switch_conversion::fix_phi_nodes): Likewise.
15355         (gen_inbound_check): Likewise.
15356         (switch_conversion::gen_inbound_check): Likewise.
15357         (process_switch): Use the newly created class.
15358         (switch_conversion::expand): New.
15359         (switch_conversion::~switch_conversion): New.
15360         * tree-switch-conversion.h: New file.
15362 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15364         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
15365         tree-vect-patterns.c.
15366         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
15367         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
15368         (vect_recog_sad_pattern): Likewise.
15369         (vect_recog_widen_sum_pattern): Likewise.
15370         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
15371         (vect_recog_widen_shift_pattern): Remove the type_in argument.
15372         (vect_recog_rotate_pattern): Likewise.
15373         (vect_recog_mult_pattern): Likewise.
15374         (vect_recog_vector_vector_shift_pattern): Likewise.
15375         (vect_recog_divmod_pattern): Likewise.
15376         (vect_recog_mixed_size_cond_pattern): Likewise.
15377         (vect_recog_bool_pattern): Likewise.
15378         (vect_recog_mask_conversion_pattern): Likewise.
15379         (vect_try_gather_scatter_pattern): Likewise.
15380         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
15381         (vect_recog_over_widening_pattern): Likewise.
15382         (vect_recog_gather_scatter_pattern): Likewise.
15383         (vect_recog_func_ptr): Move from tree-vectorizer.h
15384         (vect_vect_recog_func_ptrs): Move further down the file.
15385         (vect_recog_func): Likewise.  Remove the third argument.
15386         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
15387         (vect_pattern_recog_1): Expect the pattern function to do any
15388         necessary target tests.  Also expect it to provide a vector type.
15389         Remove the type_in handling.
15391 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15393         * tree-vect-patterns.c (vect_pattern_detected): New function.
15394         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
15395         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
15396         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
15397         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
15398         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
15399         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
15400         (vect_recog_mask_conversion_pattern)
15401         (vect_try_gather_scatter_pattern): Likewise.
15403 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15405         * tree-vect-patterns.c (vect_get_internal_def): New function.
15406         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
15407         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
15408         (search_type_for_mask_1): Use it.
15410 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15412         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15413         redundant WIDEN_SUM_EXPR handling.
15414         (vect_recog_sad_pattern): Likewise.
15416 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15418         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15419         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
15420         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
15421         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
15422         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
15424 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15426         * tree-vect-stmts.c (vectorizable_call): Make sure that we
15427         use the stmt_vec_info of the original bb statement for the
15428         new zero assignment, even if the call is part of a pattern.
15430 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15432         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
15433         that the sequence is attached to the original statement rather
15434         than the pattern statement.
15435         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
15436         PATTERN_DEF_SEQ from the original statement rather than
15437         the main pattern statement.
15438         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
15439         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
15440         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
15442 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15444         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
15445         definition statements before the early exit for statements that aren't
15446         live or relevant.
15447         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
15448         split out from...
15449         (vect_transform_loop): ...here.  Process pattern definition
15450         statements without first checking whether the main pattern
15451         statement is live or relevant.
15453 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15455         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
15456         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
15458 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
15460         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
15461         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
15462         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
15463         (expand_block_compare): Change select_block_compare_mode call.
15464         (expand_strncmp_align_check): Use new functions, fix comment.
15465         (emit_final_str_compare_gpr): New function.
15466         (expand_strn_compare): Refactor and clean up code.
15467         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
15469 2018-06-19  Tony Reix  <tony.reix@atos.com>
15470             Damien Bergamini  <damien.bergamini@atos.com>
15471             David Edelsohn  <dje.gcc@gmail.com>
15473         * collect2.c (static_obj): New variable.
15474         (static_libs): New variable.
15475         (is_in_list): Uncomment declaration.
15476         (main): Track AIX libraries linked statically.
15477         (is_in_list): Uncomment definition.
15478         (scan_prog_file): Don't add AIX shared libraries initializer
15479         to constructor list if linking statically.
15481 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
15483         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
15484         constant.
15485         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
15487 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
15489         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
15490         blocks.
15492 2018-06-19  Martin Liska  <mliska@suse.cz>
15494         * config/i386/i386.c (ix86_can_inline_p): Do not use
15495         ipa_fn_summaries::get_create.
15496         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
15497         get.
15498         (devirtualization_time_bonus): Likewise.
15499         (ipcp_propagate_stage): Likewise.
15500         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
15501         (edge_set_predicate): Likewise.
15502         (evaluate_conditions_for_known_args): Likewise.
15503         (evaluate_properties_for_edge): Likewise.
15504         (ipa_call_summary::reset): Tranform to ...
15505         (ipa_call_summary::~ipa_call_summary): ... this.
15506         (ipa_fn_summary::reset): Transform to ...
15507         (ipa_fn_summary::~ipa_fn_summary): ... this.
15508         (ipa_fn_summary_t::remove): Rename to ...
15509         (ipa_fn_summary_t::remove_callees): ... this.
15510         (ipa_fn_summary_t::duplicate): Use placement new
15511         instead of memory copy.
15512         (ipa_call_summary_t::duplicate): Likewise.
15513         (ipa_call_summary_t::remove): Remove.
15514         (dump_ipa_call_summary): Change get_create to get.
15515         (ipa_dump_fn_summary): Dump only when summary exists.
15516         (analyze_function_body): Use symbol_summary::get instead
15517         of get_create.
15518         (compute_fn_summary): Likewise.
15519         (estimate_edge_devirt_benefit): Likewise.
15520         (estimate_edge_size_and_time): Likewise.
15521         (inline_update_callee_summaries): Likewise.
15522         (remap_edge_change_prob): Likewise.
15523         (remap_edge_summaries): Likewise.
15524         (ipa_merge_fn_summary_after_inlining): Likewise.
15525         (write_ipa_call_summary): Likewise.
15526         (ipa_fn_summary_write): Likewise.
15527         (ipa_free_fn_summary): Likewise.
15528         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
15529         (struct ipa_call_summary): Likewise.
15530         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
15531         of get_create.
15532         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
15533         (estimate_size_after_inlining): Likewise.
15534         (estimate_growth): Likewise.
15535         (growth_likely_positive): Likewise.
15536         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15537         (inline_call): Likewise.
15538         * ipa-inline.c (caller_growth_limits): Likewise.
15539         (can_inline_edge_p): Likewise.
15540         (can_inline_edge_by_limits_p): Likewise.
15541         (compute_uninlined_call_time): Likewise.
15542         (compute_inlined_call_time): Likewise.
15543         (want_inline_small_function_p): Likewise.
15544         (edge_badness): Likewise.
15545         (update_caller_keys): Likewise.
15546         (update_callee_keys): Likewise.
15547         (inline_small_functions): Likewise.
15548         (inline_to_all_callers_1): Likewise.
15549         (dump_overall_stats): Likewise.
15550         (early_inline_small_functions): Likewise.
15551         (early_inliner): Likewise.
15552         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15553         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15554         * ipa-pure-const.c (malloc_candidate_p): Likewise.
15555         * ipa-split.c (execute_split_functions): Likewise.
15556         * symbol-summary.h: Likewise.
15557         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15559 2018-06-19  Richard Biener  <rguenther@suse.de>
15561         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
15562         (vectorize_loops): ... here.  Fix dbgcnt handling.
15563         (try_vectorize_loop): Wrap try_vectorize_loop_1.
15565 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
15567         PR target/86197
15568         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
15569         ieee128 argument takes up only one (vector) register, not two (floating
15570         point) registers.
15572 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15574         * gimplify.c (gimplify_init_constructor): Really never clear for an
15575         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
15577 2018-06-19  Richard Biener  <rguenther@suse.de>
15579         PR tree-optimization/86179
15580         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
15581         after failed recognition.
15583 2018-06-18  Martin Sebor  <msebor@redhat.com>
15585         PR middle-end/85602
15586         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
15587         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
15588         Handle integer subtraction.
15589         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
15590         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
15592 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15594         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
15595         param from rtx to rtx_insn *.
15596         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
15597         param.
15598         (frv_ifcvt_modify_insn): Likwise.
15599         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
15600         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
15601         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
15602         as_a <rtx_insn *> cast to local "unprotected_region" once
15603         it's been established that it's not NULL or pc_rtx.
15604         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
15605         param "sethi" from rtx to rtx_insn *.
15606         (nds32_group_float_insns): Likewise for param "insn".
15607         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
15608         param.
15609         (vax_output_int_subtract): Likewise.
15610         * config/vax/vax.c (vax_output_int_add): Likewise for param
15611         "insn".
15612         (vax_output_int_subtract): Likewise.
15613         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
15614         (emit_pattern_after): Likewise for param "after".
15615         (emit_insn_after): Likewise.
15616         (emit_jump_insn_after): Likewise.
15617         (emit_call_insn_after): Likewise.
15618         (emit_debug_insn_after): Likewise.
15619         (emit_pattern_before): Likewise for param "before".
15620         (emit_insn_before): Likewise.
15621         (emit_jump_insn_before): Likewise.
15622         * final.c (get_insn_template): Likewise for param "insn", removing
15623         a cast.
15624         * output.h (get_insn_template): Likewise for 2nd param.
15625         * rtl.h (emit_insn_before): Likewise.
15626         (emit_jump_insn_before): Likewise.
15627         (emit_debug_insn_before_noloc): Likewise.
15628         (emit_insn_after): Likewise.
15629         (emit_jump_insn_after): Likewise.
15630         (emit_call_insn_after): Likewise.
15631         (emit_debug_insn_after): Likewise.
15632         (set_insn_deleted): Likewise for param.
15634 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
15636         PR target/85358
15637         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
15638         floating point modes, so that IFmode is numerically greater than
15639         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
15640         to declare the ordering.  This prevents IFmode from being
15641         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
15642         machine.  Include rs6000-modes.h to share the fractional values
15643         between genmodes* and the rest of the compiler.
15644         (IFmode): Likewise.
15645         (KFmode): Likewise.
15646         (TFmode): Likewise.
15647         * config/rs6000/rs6000-modes.h: New file.
15648         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
15649         meaning of rs6000_long_double_size so that 126..128 selects an
15650         appropriate 128-bit floating point type.
15651         (rs6000_option_override_internal): Likewise.
15652         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
15653         (TARGET_LONG_DOUBLE_128): Change the meaning of
15654         rs6000_long_double_size so that 126..128 selects an appropriate
15655         128-bit floating point type.
15656         (LONG_DOUBLE_TYPE_SIZE): Update comment.
15657         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
15658         source and destination to match the standard usage.
15659         (truncifkf2): Likewise.
15660         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
15661         ISA 2.07 to use an explicit clobber, instead of passing in a
15662         temporary.
15663         (copysign<mode>3_soft): Likewise.
15665 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15667         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
15668         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
15669         (vect_slp_analyze_instance_dependence): Likewise.
15670         (vect_enhance_data_refs_alignment): Likewise.
15671         (vect_analyze_data_refs_alignment): Likewise.
15672         (vect_slp_analyze_and_verify_instance_alignment
15673         (vect_analyze_data_ref_accesses): Likewise.
15674         (vect_prune_runtime_alias_test_list): Likewise.
15675         (vect_analyze_data_refs): Likewise.
15676         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
15677         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15678         (vect_analyze_scalar_cycles_1): Likewise.
15679         (vect_get_loop_niters): Likewise.
15680         (vect_analyze_loop_form_1): Likewise.
15681         (vect_update_vf_for_slp): Likewise.
15682         (vect_analyze_loop_operations): Likewise.
15683         (vect_analyze_loop): Likewise.
15684         (vectorizable_induction): Likewise.
15685         (vect_transform_loop): Likewise.
15686         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
15687         * tree-vect-slp.c (vect_analyze_slp): Likewise.
15688         (vect_make_slp_decision): Likewise.
15689         (vect_detect_hybrid_slp): Likewise.
15690         (vect_slp_analyze_operations): Likewise.
15691         (vect_slp_bb): Likewise.
15692         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
15693         (vectorizable_bswap): Likewise.
15694         (vectorizable_call): Likewise.
15695         (vectorizable_simd_clone_call): Likewise.
15696         (vectorizable_conversion): Likewise.
15697         (vectorizable_assignment): Likewise.
15698         (vectorizable_shift): Likewise.
15699         (vectorizable_operation): Likewise.
15700         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
15702 2018-06-18  Martin Sebor  <msebor@redhat.com>
15704         PR tree-optimization/81384
15705         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
15706         * builtins.c (expand_builtin_strnlen): New function.
15707         (expand_builtin): Call it.
15708         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
15709         * builtins.def (BUILT_IN_STRNLEN): New.
15710         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
15711         Warn for bounds in excess of maximum object size.
15712         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
15713         single-value ranges.  Handle strnlen.
15714         (handle_builtin_strlen): Handle strnlen.
15715         (strlen_check_and_optimize_stmt): Same.
15716         * doc/extend.texi (Other Builtins): Document strnlen.
15718 2018-06-18  Maya Rashish  <coypu@sdf.org>
15720         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
15721         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
15722         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
15724         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
15725         here to ...
15726         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
15728 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15730         * tree.c (escaped_string::escape): Replace cast to char * by
15731         const_cast<char *> (unescaped).
15733 2018-06-18  Nick Clifton  <nickc@redhat.com>
15735         PR 84195
15736         * tree.c (escaped_string): New class.  Converts an unescaped
15737         string into its escaped equivalent.
15738         (warn_deprecated_use): Use the new class to convert the
15739         deprecation message, if present.
15740         (test_escaped_strings): New self test.
15741         (test_c_tests): Add test_escaped_strings.
15742         * doc/extend.texi (deprecated): Add a note that the
15743         deprecation message is affected by the -fmessage-length
15744         option, and that control characters will be escaped.
15745         (#pragma GCC error): Document this pragma.
15746         (#pragma GCC warning): Likewise.
15747         * doc/invoke.texi (-fmessage-length): Document this option's
15748         effect on the #warning and #error preprocessor directives and
15749         the deprecated attribute.
15751 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
15753         * tree.c (decl_value_expr_lookup): Revert latest change.
15754         (decl_value_expr_insert): Likewise.
15756 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
15758         * gimplify.c (nonlocal_vlas): Delete.
15759         (nonlocal_vla_vars): Likewise.
15760         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
15761         referenced VLAs.
15762         (gimplify_body): Do not create and destroy nonlocal_vlas.
15763         * tree-nested.c: Include diagnostic.h.
15764         (use_pointer_in_frame): Tweak.
15765         (lookup_field_for_decl): Add assertion and declare the transformation.
15766         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
15767         internal error when the reference is in a wrong context.  Do not
15768         create a debug decl by default.
15769         (note_nonlocal_block_vlas): Delete.
15770         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
15771         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
15772         create a debug decl by default.
15773         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
15774         call is in a wrong context.
15775         (fixup_vla_decls): New function.
15776         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
15777         debug variables were created.
15778         * tree.c (decl_value_expr_lookup): Add checking assertion.
15779         (decl_value_expr_insert): Likewise.
15781 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15783         PR middle-end/82479
15784         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
15785         * tree-scalar-evolution.c (interpret_expr): Likewise.
15786         (expression_expensive_p): Likewise.
15787         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
15788         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
15789         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
15790         (ssa_defined_by_minus_one_stmt_p): New.
15792 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15794         PR middle-end/64946
15795         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
15796         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15797         * dojump.c (do_jump): Likewise.
15798         * expr.c (expand_expr_real_2): Check operand type's sign.
15799         * fold-const.c (const_unop): Handle ABSU_EXPR.
15800         (fold_abs_const): Likewise.
15801         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
15802         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
15803         (strip_sign_op_1): Likesise.
15804         * match.pd: Add new pattern to generate ABSU_EXPR.
15805         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
15806         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15807         * tree-eh.c (operation_could_trap_helper_p): Likewise.
15808         * tree-inline.c (estimate_operator_cost): Likewise.
15809         * tree-pretty-print.c (dump_generic_node): Likewise.
15810         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
15811         * tree.def (ABSU_EXPR): New.
15813 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
15815         PR middle-end/86095
15816         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
15817         documented as preserved for backward compatibility only.
15818         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
15820         PR rtl-optimization/86108
15821         * bb-reorder.c (create_forwarder_block): Renamed to ...
15822         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
15823         jump from new landing pad to the second part.
15824         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
15825         Adjust callers.
15827 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
15829         PR middle-end/85878
15830         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
15831         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
15832         Only call store_expr for halves if the mode is the same.
15834         PR middle-end/86123
15835         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
15836         Fix up comment formatting.
15838 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15840         * typed-splay-tree.h (typed_splay_tree::remove): New function.
15841         (typed_splay_tree::closure,
15842         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
15843         (typed_splay_tree::typed_splay_tree,
15844         typed_splay_tree::operator =): Declared private.
15845         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
15846         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
15847         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
15848         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
15849         typed_splay_tree::splay_tree_splay,
15850         typed_splay_tree::splay_tree_foreach_helper,
15851         typed_splay_tree::splay_tree_insert,
15852         typed_splay_tree::splay_tree_remove,
15853         typed_splay_tree::splay_tree_lookup,
15854         typed_splay_tree::splay_tree_predecessor,
15855         typed_splay_tree::splay_tree_successor,
15856         typed_splay_tree::splay_tree_min,
15857         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
15858         (typed_splay_tree::root, typed_splay_tree::comp,
15859         typed_splay_tree::delete_key,
15860         typed_splay_tree::delete_value): New data members.
15861         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
15862         typed_splay_tree::remove.
15864 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
15866         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
15867         -mginv and -mno-ginv to the assembler.
15868         * config/mips/mips.opt (-mcrc): New option.
15869         (-mginv): Likewise.
15870         * doc/invoke.text (-mcrc): Document.
15871         (-mginv): Likewise.
15873 2018-06-15  Nick Clifton  <nickc@redhat.com>
15875         PR 84195
15876         * tree.c (escaped_string): New class.  Converts an unescaped
15877         string into its escaped equivalent.
15878         (warn_deprecated_use): Use the new class to convert the
15879         deprecation message, if present.
15880         (test_escaped_strings): New self test.
15881         (test_c_tests): Add test_escaped_strings.
15882         * doc/extend.texi (deprecated): Add a note that the
15883         deprecation message is affected by the -fmessage-length
15884         option, and that control characters will be escaped.
15885         (#pragma GCC error): Document this pragma.
15886         (#pragma GCC warning): Likewise.
15887         * doc/invoke.texi (-fmessage-length): Document this option's
15888         effect on the #warning and #error preprocessor directives and
15889         the deprecated attribute.
15891 2018-06-15  Richard Biener  <rguenther@suse.de>
15893         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
15894         here, also noting vector size used.
15895         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
15896         size used in MSG_OPTIMIZED_LOCATIONS dump.
15897         (pass_slp_vectorize::execute): Adjust.
15899 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
15901         PR target/85968
15902         * config/arc/arc.c (arc_return_address_register): Fix
15903         if-condition.
15905 2018-06-15  Richard Biener  <rguenther@suse.de>
15907         PR middle-end/86159
15908         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
15909         leave useless conversion stripping to force_gimple_operand_gsi.
15910         (gimplify_build2): Likewise.
15911         (gimplify_build1): Likewise.
15913 2018-06-15  Richard Biener  <rguenther@suse.de>
15915         PR middle-end/86076
15916         * tree-cfg.c (move_stmt_op): unshare invariant addresses
15917         before adjusting their block.
15919 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15921         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
15922         multilibs for *-*-rtems*.
15923         * config/riscv/t-rtems: New file.
15925 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
15927         PR middle-end/86122
15928         * match.pd ((A +- CST1) +- CST2): Punt if last resort
15929         unsigned_type_for returns NULL.
15931         PR target/85945
15932         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
15933         subregs of multi-word pseudos unless the float mode has word size.
15935 2018-06-14  Richard Biener  <rguenther@suse.de>
15937         PR middle-end/86139
15938         * tree-vect-generic.c (build_word_mode_vector_type): Remove
15939         duplicate and harmful type_hash_canon.
15940         * tree.c (type_hash_canon): Assert we didn't find ourselves.
15942 2018-06-14  Richard Biener  <rguenther@suse.de>
15944         PR ipa/86124
15945         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
15946         NULL cgraph_node.
15948 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15950         * config/rtems.h (STDINT_LONG32): Define.
15952 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
15953             Prachi Godbole  <prachi.godbole@imgtec.com>
15955         * config/mips/mips-cpus.def: Define P6600.
15956         * config/mips/mips-tables.opt: Regenerate.
15957         * config/mips/mips.c (mips_ucbranch_type): New enum.
15958         (mips_rtx_cost_data): Add support for P6600.
15959         (mips_issue_rate): Likewise.
15960         (mips_multipass_dfa_lookahead): Likewise.
15961         (mips_avoid_hazard): Likewise.
15962         (mips_reorg_process_insns): Likewise.
15963         (mips_classify_branch_p6600): New function.
15964         * config/mips/mips.h (TUNE_P6600): New define.
15965         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
15966         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
15967         * config/mips/mips.md: Include p6600.md.
15968         (processor): Add p6600.
15969         * config/mips/p6600.md: New file.
15970         * doc/invoke.texi: Add p6600 to supported architectures.
15972 2018-06-13  Martin Sebor  <msebor@redhat.com>
15974         PR tree-optimization/86114
15975         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
15976         of integer types.
15977         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
15979 2018-06-13  Richard Biener  <rguenther@suse.de>
15981         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
15982         Properly set vector type of the intermediate stmt.
15983         * tree-vect-stmts.c (vectorizable_operation): The destination
15984         var always has vectype_out type.
15986 2018-06-13  Jeff Law  <law@redhat.com>
15988         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
15989         integer 0 for argument to print_rtl_with_bb.
15990         (rl78_reorg): Likewise.
15992 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
15994         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
15995         from rtx to rtx_insn *.
15996         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
15997         "label".
15998         (add_sched_insns_for_speculation): Likewise for local "target",
15999         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
16000         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
16001         from rtx_insn ** to rtx_code_label **.
16002         (reorg_emit_nops): Likewise.
16003         (c6x_reorg): Likewise for local "call_labels".
16004         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
16005         rtx to rtx_insn *.
16006         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
16007         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
16008         the loops over LABEL_REFS.
16009         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
16010         braf_label.
16011         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
16012         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
16013         (split_branches): Strengthen local "olabel" from rtx to
16014         rtx_insn *, adding a safe_as_a cast.
16015         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
16016         to "rtx_insn *".
16017         (add_insn_after): Likewise for first two params.
16018         (add_insn_before): Likewise.
16019         (remove_insn): Likewise for param.
16020         (emit_pattern_before_noloc): Likewise for second and third params.
16021         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
16022         (emit_call_insn_before_noloc): Likewise.
16023         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
16024         to "rtx_insn *".
16025         (emit_barrier_before): Likewise.
16026         (emit_label_before): Strengthen "label" param from "rtx" to
16027         "rtx_code_label *".  Strengthen "before" param from "rtx" to
16028         "rtx_insn *".
16029         (emit_insn_after_1): Strengthen "after" param from "rtx" to
16030         "rtx_insn *".
16031         (emit_pattern_after_noloc): Likewise.
16032         (emit_insn_after_noloc): Likewise.
16033         (emit_jump_insn_after_noloc): Likewise.
16034         (emit_call_insn_after_noloc): Likewise.
16035         (emit_debug_insn_after_noloc): Likewise.
16036         (emit_barrier_after): Likewise.
16037         (emit_label_after): Likewise for both params.
16038         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
16039         "loc" param from "int" to "location_t".
16040         (emit_insn_after_setloc): Likewise.
16041         (emit_jump_insn_after_setloc): Likewise.
16042         (emit_call_insn_after_setloc): Likewise.
16043         (emit_debug_insn_after_setloc): Likewise.
16044         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
16045         "loc" param from "int" to "location_t".
16046         (emit_pattern_before): Convert NULL_RTX to NULL.
16047         (emit_insn_before_setloc): Convert "loc" param from "int" to
16048         "location_t".
16049         (emit_jump_insn_before_setloc): Likewise.
16050         (emit_call_insn_before_setloc): Likewise.
16051         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
16052         rtx_insn *.  Convert "loc" param from "int" to "location_t".
16053         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
16054         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
16055         Convert 3rd param from "int" to "location_t".
16056         (emit_barrier_before, emit_barrier_after, next_real_insn):
16057         Strengthen param from rtx to rtx_insn *.
16058         (emit_label_before): Strengthen 1st param from "rtx" to
16059         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
16060         "rtx_insn *".
16061         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
16062         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
16063         Strengthen 2nd param from "rtx" to "rtx_insn *".
16064         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
16065         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
16066         Likewise. Convert 3rd param from "int" to "location_t".
16067         (emit_label_after): Strengthen 1st param from "rtx" to
16068         "rtx_code_label *".
16069         (next_real_insn, remove_insn): Strengthen param from "rtx" to
16070         "rtx_insn *".
16071         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
16072         from "rtx" to "rtx_insn *".
16074 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
16076         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
16077         bodies streamed in with -Q.
16078         * dumpfile.c (dump_files): Add lto-stream-out dump file.
16079         * dumpfile.h (tree_dump_index): Add lto_stream_out.
16080         * gimple-streamer-out.c: Include gimple-pretty-print.h
16081         (output_bb): Dump stmts streamed.
16082         * lto-section-out.c: Include print-tree.h
16083         (lto_begin_section): Dump sections created.
16084         (lto_output_decl_index): Dump decl encoded.
16085         * lto-streamer-out.c: Include print-tree.h
16086         (create_output_block): Dump output block created.
16087         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
16088         (output_function): Dump function output.
16089         (output_constructor): Dump constructor streamed.
16090         (write_global_stream): Output indexes encoded.
16091         (produce_asm_for_decls): Dump streams encoded.
16092         * lto-streamer.c (streamer_dump_file): New global var.
16093         * lto-streamer.h (streamer_dump_file): Declare.
16094         * passes.c (ipa_write_summaries): Initialize streamer dump.
16095         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
16096         in.
16098 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
16100         PR target/86048
16101         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
16102         offsets for register save directives.  Emit a second batch of save
16103         directives, if need be, when the function accesses prior frames.
16105 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16107         * config/arc/fpu.md (fmasf4): Force operand to register.
16108         (fnmasf4): Likewise.
16110 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16112         * config/arc/arc-protos.h (arc_pad_return): Remove.
16113         * config/arc/arc.c (machine_function): Remove force_short_suffix
16114         and size_reason.
16115         (arc_print_operand): Adjust printing of '&'.
16116         (arc_verify_short): Remove conditional printing of short suffix.
16117         (arc_final_prescan_insn): Remove reference to size_reason.
16118         (pad_return): New function.
16119         (arc_reorg): Call pad_return.
16120         (arc_pad_return): Remove.
16121         (arc_init_machine_status): Remove reference to force_short_suffix.
16122         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
16123         (attr length): When attribute iscompact is true force to 2
16124         regardless; in the case of maybe check if we want to force the
16125         instruction to have 4 bytes length.
16126         (nopv): Change it to generate 4 byte long nop as well.
16127         (blockage): New pattern.
16128         (simple_return): Remove call to arc_pad_return.
16129         (p_return_i): Likewise.
16131 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16133         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
16135 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16137         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
16138         ARC cores.
16140 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16142         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
16143         for ARC700 and ARCv2.
16145 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
16147         PR target/86076
16148         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
16149         operands[2] instead of operands[1].
16152 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
16154         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
16155         case, check whether the outer register overlaps an unallocatable
16156         register, not just whether it fits the required class.
16158 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
16160         * poly-int.h (can_div_trunc_p): Add new overload in which all values
16161         are poly_ints.
16162         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
16163         (memrefs_conflict_p): Likewise.
16164         (init_alias_analysis): Likewise.
16165         * cfgexpand.c (expand_debug_expr): Likewise.
16166         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
16167         * cse.c (fold_rtx): Likewise.
16168         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
16169         * expr.c (emit_block_move_hints): Likewise.
16170         (clear_storage_hints, push_block, emit_push_insn): Likewise.
16171         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
16172         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
16173         (emit_group_store): Likewise.
16174         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
16175         to read the PRE/POST_MODIFY increment.
16176         * calls.c (store_one_arg): Use strip_offset.
16177         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
16178         poly_int_rtx_p.
16179         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
16180         by a VEC_SELECT.
16181         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
16182         (simplify_binary_operation_1): Extend CONST_INT handling to
16183         poly_int_rtx_p.
16184         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
16185         than a HOST_WIDE_INT.
16186         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
16187         poly_int64.
16188         (adjust_mems, add_stores): Update accodingly.
16189         (vt_canonicalize_addr): Track polynomial offsets.
16190         (emit_note_insn_var_location): Likewise.
16191         (vt_add_function_parameter): Likewise.
16192         (vt_initialize): Likewise.
16194 2018-06-12  Jeff Law  <law@redhat.com>
16196         * config.gcc (alpha*-*-freebsd*): Remove.
16197         * config/alpha/freebsd.h: Remove.
16199 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
16201         PR other/69968
16202         * spellcheck-tree.c (levenshtein_distance): Rename to...
16203         (get_edit_distance): ...this, and update for underlying renaming.
16204         * spellcheck-tree.h (levenshtein_distance): Rename to...
16205         (get_edit_distance): ...this.
16206         * spellcheck.c (levenshtein_distance): Rename to...
16207         (get_edit_distance): ...this.  Convert from Levenshtein distance
16208         to Damerau-Levenshtein distance by supporting transpositions of
16209         adjacent characters.  Rename "v1" to "v_next" and "v0" to
16210         "v_one_ago".
16211         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
16212         (selftest::test_edit_distance_unit_test_oneway): ...this, and
16213         update for underlying renaming.
16214         (selftest::levenshtein_distance_unit_test): Rename to...
16215         (selftest::test_get_edit_distance_unit): ...this, and update for
16216         underlying renaming.
16217         (selftest::test_find_closest_string): Add example from PR 69968
16218         where transposition helps
16219         (selftest::test_metric_conditions): Update for renaming.
16220         (selftest::test_metric_conditions): Likewise.
16221         (selftest::spellcheck_c_tests): Likewise.
16222         * spellcheck.h (levenshtein_distance): Rename both overloads to...
16223         (get_edit_distance): ...this.
16224         (best_match::consider): Update for renaming.
16226 2018-06-12  Martin Sebor  <msebor@redhat.com>
16228         PR tree-optimization/85259
16229         * builtins.c (compute_objsize): Handle constant offsets.
16230         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
16231         true iff a warning has been issued.
16232         * gimple.h (gimple_nonartificial_location): New function.
16233         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
16234         gimple_nonartificial_location and handle -Wno-system-headers.
16235         (handle_builtin_stxncpy): Same.
16237 2018-06-12  Martin Sebor  <msebor@redhat.com>
16239         PR c/85931
16240         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
16242 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
16244         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16245         BUILTIN_VEC_XST entries for pointer to double and long long.
16247 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
16249         PR target/85990
16250         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16251         Update comments.
16252         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16253         Likewise.
16255 2018-06-12  Martin Liska  <mliska@suse.cz>
16257         * doc/options.texi: Document IntegerRange.
16259 2018-06-12  Martin Liska  <mliska@suse.cz>
16261         * config/i386/i386.opt: Make MPX-related options as Deprecated.
16262         * opt-functions.awk: Handle Deprecated flag.
16263         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
16264         and report error.
16265         (read_cmdline_option): Report warning for a deprecated option.
16266         * opts.h (struct cl_option): Add new field cl_deprecated.
16267         (CL_ERR_DEPRECATED): New.
16269 2018-06-12  Martin Liska  <mliska@suse.cz>
16271         * doc/options.texi: Document Deprecated option flag.
16273 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16275         * config/arc/arc-arch.h (arc_extras): New enum.
16276         (arc_cpu_t):Add field extra.
16277         (arc_cpu_types): Consider the extras.
16278         * config/arc/arc-cpus.def: Add extras info.
16279         * config/arc/arc-opts.h (processor_type): Consider extra field.
16280         * config/arc/arc.c (arc_override_options): Handle extra field.
16282 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16284         * config/arc/arc-arch.h: Update ARC_OPTX macro.
16285         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
16286         field.
16287         * config/arc/arc.c (arc_init): Update pic warning.
16288         (irq_range): Update irq range parsing warnings.
16289         (arc_override_options): Update various warning messages.
16290         (arc_handle_aux_attribute): Likewise.
16292 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
16294         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
16296 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16298         * doc/sourcebuild.texi: Document usage of line number 0 in verify
16299         compiler messages directives.
16301 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
16303         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
16304         * config/mips/mips-tables.opt: Regenerate.
16305         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
16306         mips64r6.
16307         * doc/invoke.texi: Document -march=i6500.
16309 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
16311         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
16312         (i6400_gpmul): Add cpu_unit.
16313         (i6400_gpdiv): Likewise.
16314         (i6400_msa_add_d): Update reservations.
16315         (i6400_msa_int_add) Likewise.
16316         (i6400_msa_short_logic3) Likewise.
16317         (i6400_msa_short_logic2) Likewise.
16318         (i6400_msa_short_logic) Likewise.
16319         (i6400_msa_move) Likewise.
16320         (i6400_msa_cmp) Likewise.
16321         (i6400_msa_short_float2) Likewise.
16322         (i6400_msa_div_d) Likewise.
16323         (i6400_msa_long_logic1) Likewise.
16324         (i6400_msa_long_logic2) Likewise.
16325         (i6400_msa_mult) Likewise.
16326         (i6400_msa_long_float2) Likewise.
16327         (i6400_msa_long_float4) Likewise.
16328         (i6400_msa_long_float5) Likewise.
16329         (i6400_msa_long_float8) Likewise.
16330         (i6400_fpu_fadd): Include frint type.
16331         (i6400_fpu_store): New define_insn_reservation.
16332         (i6400_fpu_load): Likewise.
16333         (i6400_fpu_move): Likewise.
16334         (i6400_fpu_fcmp): Likewise.
16335         (i6400_fpu_fmadd): Likewise.
16336         (i6400_int_mult): Include imul3nc type and update reservation.
16337         (i6400_int_div): Include idiv3 type and update reservation.
16338         (i6400_int_load): Update to check type not move_type.
16339         (i6400_int_store): Likewise.
16340         (i6400_int_prefetch): Set zero latency.
16342 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
16344         * gcc.c: Document new %@{...} sequence.
16345         (LINK_COMMAND_SPEC): Use it for the -L switches.
16346         (cpp_unique_options): Use it for the -I switches.
16347         (at_file_argbuf): New global variable.
16348         (in_at_file): Likewise.
16349         (alloc_args): Create at_file_argbuf.
16350         (clear_args): Truncate at_file_argbuf.
16351         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
16352         (open_at_file): New function.
16353         (close_at_file): Likewise.
16354         (create_at_file): Delete.
16355         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
16356         <'o'>: Likewise.
16357         <'@'>: New case.
16358         (validate_switches_from_spec): Deal with %@{...} sequence.
16359         (validate_switches): Likewise.
16360         (driver::finalize): Call clear_args.
16362 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
16364         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
16366 2018-06-11  Martin Sebor  <msebor@redhat.com>
16368         * doc/invoke.texi (-Wall): List -Wc++17-compat.
16369         (Wno-class-memaccess): Add @opindex.
16370         (Wno-templates, Wno-multiple-inheritance): Same.
16371         (Wno-virtual-inheritance, Wno-namespaces): Same.
16372         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
16373         (Wno-format-overflow, Wno-format-truncation): Same.
16374         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
16375         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
16376         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
16377         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
16378         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
16379         (Wno-misspelled-isr): Same.
16381 2018-06-11  Martin Sebor  <msebor@redhat.com>
16383         * PR tree-optimization/86083
16384         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
16386 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
16388         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
16390 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
16392         PR target/85755
16393         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
16394         on the correct operand.
16395         (*movdi_internal64): Ditto.
16397 2018-06-11  Martin Liska  <mliska@suse.cz>
16399         PR tree-optimization/86089
16400         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
16402 2018-06-11  Julia Koval  <julia.koval@intel.com>
16404         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
16405         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
16406         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
16408 2018-06-11  Olivier Hainque  <hainque@adacore.com>
16410         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
16411         for Ada with strict dwarf2.
16413 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
16415         PR target/85755
16416         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
16417         addresses.
16419 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
16421         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
16423 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
16425         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
16426         TARGET_ELF.
16428 2018-06-08  Martin Liska  <mliska@suse.cz>
16430         * tree-cfg.h (debug_function): Fix argument type to match
16431         implementation.
16433 2018-06-08  Martin Liska  <mliska@suse.cz>
16435         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
16436         Remove usage of MPX-related (and removed) fields.
16437         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
16439 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
16441         * cfg.c (debug): Use TDF_NONE rather than 0.
16442         * cfghooks.c (debug): Likewise.
16443         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
16444         (struct dump_option_value_info): Convert to...
16445         (struct kv_pair): ...this template type.
16446         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
16447         rather than 0.
16448         (optinfo_verbosity_options): Likewise.
16449         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
16450         OPTGROUP_NONE.
16451         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
16452         than int for "optgroup_flags" param.
16453         (dump_generic_expr_loc): Use dump_flags_t rather than int for
16454         "dump_kind" param.
16455         (dump_dec): Likewise.
16456         (dump_finish): Use TDF_NONE rather than 0.
16457         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
16458         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
16459         than 0.  Update for change to option_ptr.
16460         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
16461         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
16462         0.  Update for changes to optinfo_verbosity_options and
16463         optgroup_options.
16464         (opt_info_switch_p): Convert optgroup_flags from int to
16465         optgroup_flags_t.
16466         (dump_basic_block): Use dump_flags_t rather than int
16467         for "dump_kind" param.
16468         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
16469         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
16470         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
16471         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
16472         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
16473         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
16474         TDF_NONE): Convert from macros to...
16475         (enum dump_flag): ...this new enum.
16476         (dump_flags_t): Update to use enum.
16477         (operator|, operator&, operator~, operator|=, operator&=):
16478         Implement for dump_flags_t.
16479         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
16480         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
16481         Convert from macros to...
16482         (enum optgroup_flag): ...this new enum.
16483         (optgroup_flags_t): New typedef.
16484         (operator|, operator|=): Implement for optgroup_flags_t.
16485         (struct dump_file_info): Convert field "alt_flags" to
16486         dump_flags_t.  Convert field "optgroup_flags" to
16487         optgroup_flags_t.
16488         (dump_basic_block): Use dump_flags_t rather than int for param.
16489         (dump_generic_expr_loc): Likewise.
16490         (dump_dec): Likewise.
16491         (dump_register): Convert param "optgroup_flags" to
16492         optgroup_flags_t.
16493         (opt_info_enable_passes): Likewise.
16494         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
16495         than 0.
16496         * gimple-pretty-print.c (debug): Likewise.
16497         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
16498         (merged_store_group::apply_stores): Likewise.
16499         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
16500         * gimple.c (verify_gimple_pp): Likewise.
16501         * graphite-poly.c (print_pbb_body): Likewise.
16502         * passes.c (pass_manager::register_one_dump_file): Convert
16503         local "optgroup_flags" to optgroup_flags_t.
16504         * print-tree.c (print_node): Use TDF_NONE rather than 0.
16505         (debug): Likewise.
16506         (debug_body): Likewise.
16507         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
16508         to optgroup_flags_t.
16509         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
16510         than 0.
16511         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
16512         (convert_mult_to_fma): Likewise.
16513         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
16514         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
16515         * tree-vect-data-refs.c (dump_lower_bound): Convert param
16516         "dump_kind" to dump_flags_t.
16518 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16520         * config/rs6000/rs6000.c (min, max): Delete.
16522 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16524         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
16525         -mabi=spe and -mabi=no-spe.
16527 2018-06-08  Martin Liska  <mliska@suse.cz>
16529         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
16530         where we expect an existing summary.
16532 2018-06-08  Martin Liska  <mliska@suse.cz>
16534         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
16535         * ipa-inline.h (estimate_edge_growth): Likewise.
16537 2018-06-08  Martin Liska  <mliska@suse.cz>
16539         * cgraph.c (function_version_hasher::hash): Use
16540         cgraph_node::get_uid ().
16541         (function_version_hasher::equal):
16542         * cgraph.h (cgraph_node::get_uid): New method.
16543         * ipa-inline.c (update_caller_keys): Use
16544         cgraph_node::get_uid ().
16545         (update_callee_keys): Likewise.
16546         * ipa-utils.c (searchc): Likewise.
16547         (ipa_reduced_postorder): Likewise.
16548         * lto-cgraph.c (input_node): Likewise.
16549         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
16550         * symbol-summary.h (symtab_insertion): Likewise.
16551         (symtab_removal): Likewise.
16552         (symtab_duplication): Likewise.
16553         * tree-pretty-print.c (dump_function_header): Likewise.
16554         * tree-sra.c (convert_callers_for_node): Likewise.
16556 2018-06-08  Martin Liska  <mliska@suse.cz>
16558         * cgraph.c (symbol_table::create_edge): Always assign a new
16559         unique number.
16560         (symbol_table::free_edge): Do not recycle numbers.
16561         * cgraph.h (cgraph_edge::get): New method.
16562         * symbol-summary.h (symtab_removal): Use it.
16563         (symtab_duplication): Likewise.
16564         (call_summary::hashable_uid): Remove.
16566 2018-06-08  Martin Liska  <mliska@suse.cz>
16568         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
16569         (initialize_growth_caches): Remove.
16570         (free_growth_caches): Likewise.
16571         (do_estimate_edge_time): Use edge_growth_cache.
16572         (do_estimate_edge_size): Likewise.
16573         (do_estimate_edge_hints): Likewise.
16574         * ipa-inline.c (reset_edge_caches): Likewise.
16575         (recursive_inlining): Likewise.
16576         (inline_small_functions): Likewise.
16577         * ipa-inline.h (initialize_growth_caches): Remove.
16578         (estimate_edge_size): Likewise.
16579         (estimate_edge_time): Likewise.
16580         (estimate_edge_hints): Likewise.
16581         (reset_edge_growth_cache): Likewise.
16582         * symbol-summary.h (call_summary::remove): New method.
16584 2018-06-08  Martin Liska  <mliska@suse.cz>
16586         * ipa-cp.c (class edge_clone_summary): New summary.
16587         (grow_edge_clone_vectors): Remove.
16588         (ipcp_edge_duplication_hook): Remove.
16589         (class edge_clone_summary_t): New call_summary class.
16590         (ipcp_edge_removal_hook): Remove.
16591         (edge_clone_summary_t::duplicate): New function.
16592         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
16593         (create_specialized_node): Likewise.
16594         (ipcp_driver): Initialize edge_clone_summaries and do not
16595         register hooks.
16597 2018-06-08  Martin Liska  <mliska@suse.cz>
16599         * symbol-summary.h (get): New function.
16600         (call_summary::m_initialize_when_cloning): New class member.
16602 2018-06-08  Martin Liska  <mliska@suse.cz>
16604         * cgraph.c (cgraph_node::remove): Do not recycle uid.
16605         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
16606         (symbol_table::allocate_cgraph_symbol): Do not set uid.
16607         * passes.c (uid_hash_t): Record removed_nodes by their uids.
16608         (remove_cgraph_node_from_order): Use the removed_nodes set.
16609         (do_per_function_toporder): Likwise.
16610         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
16611         instead of summary_uid.
16612         (symtab_removal): Likewise.
16613         (symtab_duplication): Likewise.
16615 2018-06-08  Martin Liska  <mliska@suse.cz>
16617         * ipa-cp.c (ipcp_store_bits_results): Use
16618         ipcp_transformation_sum.
16619         (ipcp_store_vr_results): Likewise.
16620         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
16621         to ...
16622         (ipcp_transformation_initialize): ... this.
16623         (ipa_set_node_agg_value_chain):
16624         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
16625         (write_ipcp_transformation_info): Likewise.
16626         (read_ipcp_transformation_info): Likewise.
16627         (ipcp_update_bits): Likewise.
16628         (ipcp_update_vr): Likewise.
16629         (ipcp_transform_function): Likewise.
16630         * ipa-prop.h: Rename ipcp_transformation_summary to
16631         ipcp_transformation.
16632         (class ipcp_transformation_t): New function summary.
16633         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
16634         (ipa_get_agg_replacements_for_node): Likewise.
16636 2018-06-08  Martin Liska  <mliska@suse.cz>
16638         * ipa-pure-const.c (struct funct_state_d): Do it class instead
16639         of struct.
16640         (class funct_state_summary_t): New function_summary class.
16641         (has_function_state): Remove.
16642         (get_function_state): Likewise.
16643         (set_function_state): Likewise.
16644         (add_new_function): Likewise.
16645         (funct_state_summary_t::insert): New function.
16646         (duplicate_node_data): Remove.
16647         (remove_node_data): Remove.
16648         (funct_state_summary_t::duplicate): New function.
16649         (register_hooks): Create new funct_state_summaries.
16650         (pure_const_generate_summary): Use it.
16651         (pure_const_write_summary): Likewise.
16652         (pure_const_read_summary): Likewise.
16653         (propagate_pure_const): Likewise.
16654         (propagate_nothrow): Likewise.
16655         (dump_malloc_lattice): Likewise.
16656         (propagate_malloc): Likewise.
16657         (execute): Do not register hooks, just remove summary
16658         instead.
16659         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
16660         constructor.
16662 2018-06-08  Martin Liska  <mliska@suse.cz>
16664         * ipa-reference.c (remove_node_data): Remove.
16665         (duplicate_node_data): Likewise.
16666         (class ipa_ref_var_info_summary_t): New class.
16667         (class ipa_ref_opt_summary_t): Likewise.
16668         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
16669         (get_reference_optimization_summary): Use
16670         ipa_ref_opt_sum_summaries.
16671         (set_reference_vars_info): Remove.
16672         (set_reference_optimization_summary): Likewise.
16673         (ipa_init): Create summaries.
16674         (init_function_info): Use function summary.
16675         (ipa_ref_opt_summary_t::duplicate): New function.
16676         (ipa_ref_opt_summary_t::remove): New function.
16677         (get_read_write_all_from_node): Fix GNU coding style.
16678         (propagate): Use function summary.
16679         (write_node_summary_p): Fix GNU coding style.
16680         (stream_out_bitmap): Likewise.
16681         (ipa_reference_read_optimization_summary): Use function summary.
16682         (ipa_reference_c_finalize): Do not release hooks.
16684 2018-06-08  Martin Liska  <mliska@suse.cz>
16686         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
16687         (analyze_function_body): Extract multiple calls of get_create.
16688         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16689         * ipa-inline.c (recursive_inlining): Use ::get method.
16690         * ipa-inline.h (estimate_edge_growth): Likewise.
16692 2018-06-08  Martin Liska  <mliska@suse.cz>
16694         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
16695         HSA_INVALID.
16696         (hsa_function_summary::hsa_function_summary): Use the new enum
16697         value.
16698         (hsa_gpu_implementation_p): Use hsa_summaries::get.
16699         * hsa-gen.c (hsa_get_host_function): Likewise.
16700         (get_brig_function_name): Likewise.
16701         * ipa-hsa.c (process_hsa_functions): Likewise.
16702         (ipa_hsa_write_summary): Likewise.
16703         * symbol-summary.h (symtab_duplication): Use ::get function/
16704         (get): New function.
16706 2018-06-08  Martin Liska  <mliska@suse.cz>
16708         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
16709         of get.
16710         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
16711         (hsa_register_kernel): Likewise.
16712         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
16713         * hsa-gen.c (hsa_get_host_function): Likewise.
16714         (get_brig_function_name): Likewise.
16715         (generate_hsa): Likewise.
16716         (pass_gen_hsail::execute): Likewise.
16717         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
16718         (devirtualization_time_bonus): Likewise.
16719         (ipcp_propagate_stage): Likewise.
16720         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
16721         (edge_set_predicate): Likewise.
16722         (evaluate_conditions_for_known_args): Likewise.
16723         (evaluate_properties_for_edge): Likewise.
16724         (ipa_fn_summary::reset): Likewise.
16725         (ipa_fn_summary_t::duplicate): Likewise.
16726         (dump_ipa_call_summary): Likewise.
16727         (ipa_dump_fn_summary): Likewise.
16728         (analyze_function_body): Likewise.
16729         (compute_fn_summary): Likewise.
16730         (estimate_edge_devirt_benefit): Likewise.
16731         (estimate_edge_size_and_time): Likewise.
16732         (estimate_calls_size_and_time): Likewise.
16733         (estimate_node_size_and_time): Likewise.
16734         (inline_update_callee_summaries): Likewise.
16735         (remap_edge_change_prob): Likewise.
16736         (remap_edge_summaries): Likewise.
16737         (ipa_merge_fn_summary_after_inlining): Likewise.
16738         (ipa_update_overall_fn_summary): Likewise.
16739         (read_ipa_call_summary): Likewise.
16740         (inline_read_section): Likewise.
16741         (write_ipa_call_summary): Likewise.
16742         (ipa_fn_summary_write): Likewise.
16743         (ipa_free_fn_summary): Likewise.
16744         * ipa-hsa.c (process_hsa_functions): Likewise.
16745         (ipa_hsa_write_summary): Likewise.
16746         (ipa_hsa_read_section): Likewise.
16747         * ipa-icf.c (sem_function::merge): Likewise.
16748         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16749         (do_estimate_edge_time): Likewise.
16750         (estimate_size_after_inlining): Likewise.
16751         (estimate_growth): Likewise.
16752         (growth_likely_positive): Likewise.
16753         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
16754         (inline_call): Likewise.
16755         * ipa-inline.c (caller_growth_limits): Likewise.
16756         (can_inline_edge_p): Likewise.
16757         (can_inline_edge_by_limits_p): Likewise.
16758         (compute_uninlined_call_time): Likewise.
16759         (compute_inlined_call_time): Likewise.
16760         (want_inline_small_function_p): Likewise.
16761         (edge_badness): Likewise.
16762         (update_caller_keys): Likewise.
16763         (update_callee_keys): Likewise.
16764         (recursive_inlining): Likewise.
16765         (inline_small_functions): Likewise.
16766         (inline_to_all_callers_1): Likewise.
16767         (dump_overall_stats): Likewise.
16768         (early_inline_small_functions): Likewise.
16769         (early_inliner): Likewise.
16770         * ipa-inline.h (estimate_edge_growth): Likewise.
16771         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
16772         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16773         * ipa-prop.h (IPA_NODE_REF): Likewise.
16774         (IPA_EDGE_REF): Likewise.
16775         * ipa-pure-const.c (malloc_candidate_p): Likewise.
16776         (propagate_malloc): Likewise.
16777         * ipa-split.c (execute_split_functions): Likewise.
16778         * symbol-summary.h: Rename get to get_create.
16779         (get): Likewise.
16780         (get_create): Likewise.
16781         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
16783 2018-06-08  Martin Liska  <mliska@suse.cz>
16785         * symbol-summary.h (release): Move definition out of class
16786         declaration.
16787         (symtab_removal): Likewise.
16788         (symtab_duplication): Likewise.
16790 2018-06-08  Martin Liska  <mliska@suse.cz>
16792         * symbol-summary.h (function_summary): Move constructor
16793         implementation out of class declaration.
16794         (release): Likewise.
16795         (symtab_insertion): Likewise.
16796         (symtab_removal): Likewise.
16797         (symtab_duplication): Likewise.
16798         (get): Likewise.
16800 2018-06-08  Martin Liska  <mliska@suse.cz>
16802         * Makefile.in: Remove support for MPX (macros, related functions,
16803         fields in cgraph_node, ...).
16804         * builtin-types.def (BT_BND): Likewise.
16805         (BT_FN_BND_CONST_PTR): Likewise.
16806         (BT_FN_CONST_PTR_BND): Likewise.
16807         (BT_FN_VOID_PTR_BND): Likewise.
16808         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
16809         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
16810         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
16811         (expand_builtin_mempcpy_with_bounds): Likewise.
16812         (expand_builtin_memset_with_bounds): Likewise.
16813         (expand_builtin_memset_args): Likewise.
16814         (std_expand_builtin_va_start): Likewise.
16815         (expand_builtin): Likewise.
16816         (expand_builtin_with_bounds): Likewise.
16817         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
16818         (DEF_LIB_BUILTIN_CHKP): Likewise.
16819         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
16820         (DEF_CHKP_BUILTIN): Likewise.
16821         (BUILT_IN_MEMCPY): Likewise.
16822         (BUILT_IN_MEMMOVE): Likewise.
16823         (BUILT_IN_MEMPCPY): Likewise.
16824         (BUILT_IN_MEMSET): Likewise.
16825         (BUILT_IN_STPCPY): Likewise.
16826         (BUILT_IN_STRCAT): Likewise.
16827         (BUILT_IN_STRCHR): Likewise.
16828         (BUILT_IN_STRCPY): Likewise.
16829         (BUILT_IN_STRLEN): Likewise.
16830         (BUILT_IN_MEMCPY_CHK): Likewise.
16831         (BUILT_IN_MEMMOVE_CHK): Likewise.
16832         (BUILT_IN_MEMPCPY_CHK): Likewise.
16833         (BUILT_IN_MEMSET_CHK): Likewise.
16834         (BUILT_IN_STPCPY_CHK): Likewise.
16835         (BUILT_IN_STRCAT_CHK): Likewise.
16836         (BUILT_IN_STRCPY_CHK): Likewise.
16837         * calls.c (store_bounds): Likewise.
16838         (emit_call_1): Likewise.
16839         (special_function_p): Likewise.
16840         (maybe_warn_nonstring_arg): Likewise.
16841         (initialize_argument_information): Likewise.
16842         (finalize_must_preallocate): Likewise.
16843         (compute_argument_addresses): Likewise.
16844         (expand_call): Likewise.
16845         * cfgexpand.c (expand_call_stmt): Likewise.
16846         (expand_return): Likewise.
16847         (expand_gimple_stmt_1): Likewise.
16848         (pass_expand::execute): Likewise.
16849         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
16850         (cgraph_node::remove): Likewise.
16851         (cgraph_node::dump): Likewise.
16852         (cgraph_node::verify_node): Likewise.
16853         * cgraph.h (chkp_function_instrumented_p): Likewise.
16854         (symtab_node::get_alias_target): Likewise.
16855         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
16856         (cgraph_local_p): Likewise.
16857         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
16858         (cgraph_edge::rebuild_references): Likewise.
16859         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
16860         (walk_polymorphic_call_targets): Likewise.
16861         (cgraph_node::expand_thunk): Likewise.
16862         (symbol_table::output_weakrefs): Likewise.
16863         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
16864         (ix86_handle_option): Likewise.
16865         * config/i386/constraints.md: Likewise.
16866         * config/i386/i386-builtin-types.def (BND): Likewise.
16867         (VOID): Likewise.
16868         (PVOID): Likewise.
16869         (ULONG): Likewise.
16870         * config/i386/i386-builtin.def (BDESC_END): Likewise.
16871         (BDESC_FIRST): Likewise.
16872         (BDESC): Likewise.
16873         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
16874         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
16875         * config/i386/i386.c (enum reg_class): Likewise.
16876         (ix86_target_string): Likewise.
16877         (ix86_option_override_internal): Likewise.
16878         (ix86_conditional_register_usage): Likewise.
16879         (ix86_valid_target_attribute_inner_p): Likewise.
16880         (ix86_set_indirect_branch_type): Likewise.
16881         (ix86_set_current_function): Likewise.
16882         (ix86_function_arg_regno_p): Likewise.
16883         (init_cumulative_args): Likewise.
16884         (ix86_function_arg_advance): Likewise.
16885         (ix86_function_arg): Likewise.
16886         (ix86_pass_by_reference): Likewise.
16887         (ix86_function_value_regno_p): Likewise.
16888         (ix86_function_value_1): Likewise.
16889         (ix86_function_value_bounds): Likewise.
16890         (ix86_return_in_memory): Likewise.
16891         (ix86_setup_incoming_vararg_bounds): Likewise.
16892         (ix86_va_start): Likewise.
16893         (indirect_thunk_need_prefix): Likewise.
16894         (print_reg): Likewise.
16895         (ix86_print_operand): Likewise.
16896         (ix86_expand_call): Likewise.
16897         (ix86_output_function_return): Likewise.
16898         (reg_encoded_number): Likewise.
16899         (BDESC_VERIFYS): Likewise.
16900         (ix86_init_mpx_builtins): Likewise.
16901         (ix86_init_builtins): Likewise.
16902         (ix86_emit_cmove): Likewise.
16903         (ix86_emit_move_max): Likewise.
16904         (ix86_expand_builtin): Likewise.
16905         (ix86_builtin_mpx_function): Likewise.
16906         (ix86_get_arg_address_for_bt): Likewise.
16907         (ix86_load_bounds): Likewise.
16908         (ix86_store_bounds): Likewise.
16909         (ix86_load_returned_bounds): Likewise.
16910         (ix86_store_returned_bounds): Likewise.
16911         (ix86_class_likely_spilled_p): Likewise.
16912         (ix86_hard_regno_mode_ok): Likewise.
16913         (x86_order_regs_for_local_alloc): Likewise.
16914         (ix86_mitigate_rop): Likewise.
16915         (ix86_bnd_prefixed_insn_p): Likewise.
16916         (ix86_mpx_bound_mode): Likewise.
16917         (ix86_make_bounds_constant): Likewise.
16918         (ix86_initialize_bounds): Likewise.
16919         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
16920         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
16921         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
16922         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
16923         (TARGET_CHKP_BOUND_MODE): Likewise.
16924         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
16925         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
16926         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
16927         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
16928         * config/i386/i386.h (TARGET_MPX): Likewise.
16929         (TARGET_MPX_P): Likewise.
16930         (VALID_BND_REG_MODE): Likewise.
16931         (FIRST_BND_REG): Likewise.
16932         (LAST_BND_REG): Likewise.
16933         (enum reg_class): Likewise.
16934         (BND_REG_P): Likewise.
16935         (BND_REGNO_P): Likewise.
16936         (BNDmode): Likewise.
16937         (ADJUST_INSN_LENGTH): Likewise.
16938         * config/i386/i386.md: Likewise.
16939         * config/i386/i386.opt: Likewise.
16940         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
16941         (defined): Likewise.
16942         (LINK_MPX): Likewise.
16943         (MPX_SPEC): Likewise.
16944         (LIBMPX_SPEC): Likewise.
16945         (LIBMPXWRAPPERS_SPEC): Likewise.
16946         (CHKP_SPEC): Likewise.
16947         * config/i386/predicates.md: Likewise.
16948         * dbxout.c (dbxout_type): Likewise.
16949         * doc/extend.texi: Likewise.
16950         * doc/invoke.texi: Likewise.
16951         * doc/md.texi: Likewise.
16952         * doc/tm.texi: Likewise.
16953         * doc/tm.texi.in: Likewise.
16954         * dwarf2out.c (is_base_type): Likewise.
16955         (gen_formal_types_die): Likewise.
16956         (gen_subprogram_die): Likewise.
16957         (gen_type_die_with_usage): Likewise.
16958         (gen_decl_die): Likewise.
16959         (dwarf2out_late_global_decl): Likewise.
16960         * expr.c (expand_assignment): Likewise.
16961         (emit_storent_insn): Likewise.
16962         (store_expr_with_bounds): Likewise.
16963         (store_expr): Likewise.
16964         (expand_expr_real_1): Likewise.
16965         * expr.h (store_expr_with_bounds): Likewise.
16966         * function.c (use_register_for_decl): Likewise.
16967         (struct bounds_parm_data): Likewise.
16968         (assign_parms_augmented_arg_list): Likewise.
16969         (assign_parm_find_entry_rtl): Likewise.
16970         (assign_parm_is_stack_parm): Likewise.
16971         (assign_parm_load_bounds): Likewise.
16972         (assign_bounds): Likewise.
16973         (assign_parms): Likewise.
16974         (expand_function_start): Likewise.
16975         * gcc.c (CHKP_SPEC): Likewise.
16976         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16977         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
16978         (wrestrict_dom_walker::check_call): Likewise.
16979         * gimple.c (gimple_build_call_from_tree): Likewise.
16980         * gimple.h (enum gf_mask): Likewise.
16981         (gimple_call_with_bounds_p): Likewise.
16982         (gimple_call_set_with_bounds): Likewise.
16983         * gimplify.c (gimplify_init_constructor): Likewise.
16984         * ipa-cp.c (initialize_node_lattices): Likewise.
16985         (propagate_constants_across_call): Likewise.
16986         (find_more_scalar_values_for_callers_subset): Likewise.
16987         * ipa-hsa.c (process_hsa_functions): Likewise.
16988         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
16989         * ipa-icf.c (sem_function::merge): Likewise.
16990         * ipa-inline.c (early_inliner): Likewise.
16991         * ipa-pure-const.c (warn_function_noreturn): Likewise.
16992         (warn_function_cold): Likewise.
16993         (propagate_pure_const): Likewise.
16994         * ipa-ref.h (enum GTY): Likewise.
16995         * ipa-split.c (find_retbnd): Likewise.
16996         (consider_split): Likewise.
16997         (split_function): Likewise.
16998         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
16999         * ipa.c (walk_polymorphic_call_targets): Likewise.
17000         (symbol_table::remove_unreachable_nodes): Likewise.
17001         (process_references): Likewise.
17002         (cgraph_build_static_cdtor_1): Likewise.
17003         * lto-cgraph.c (lto_output_node): Likewise.
17004         (output_refs): Likewise.
17005         (compute_ltrans_boundary): Likewise.
17006         (input_overwrite_node): Likewise.
17007         (input_node): Likewise.
17008         (input_cgraph_1): Likewise.
17009         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
17010         * passes.c (pass_manager::execute_early_local_passes): Likewise.
17011         (class pass_chkp_instrumentation_passes): Likewise.
17012         (make_pass_chkp_instrumentation_passes): Likewise.
17013         * passes.def: Likewise.
17014         * rtl.h (struct GTY): Likewise.
17015         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
17016         * stor-layout.c (layout_type): Likewise.
17017         * symtab.c: Likewise.
17018         * target.def: Likewise.
17019         * targhooks.c (default_chkp_bound_type): Likewise.
17020         (default_chkp_bound_mode): Likewise.
17021         (default_builtin_chkp_function): Likewise.
17022         (default_chkp_function_value_bounds): Likewise.
17023         (default_chkp_make_bounds_constant): Likewise.
17024         (default_chkp_initialize_bounds): Likewise.
17025         * targhooks.h (default_chkp_bound_type): Likewise.
17026         (default_chkp_bound_mode): Likewise.
17027         (default_builtin_chkp_function): Likewise.
17028         (default_chkp_function_value_bounds): Likewise.
17029         (default_chkp_make_bounds_constant): Likewise.
17030         (default_chkp_initialize_bounds): Likewise.
17031         * toplev.c (compile_file): Likewise.
17032         (process_options): Likewise.
17033         * tree-core.h (DEF_BUILTIN): Likewise.
17034         (DEF_BUILTIN_CHKP): Likewise.
17035         * tree-inline.c (declare_return_variable): Likewise.
17036         (remap_gimple_stmt): Likewise.
17037         (copy_bb): Likewise.
17038         (initialize_inlined_parameters): Likewise.
17039         (expand_call_inline): Likewise.
17040         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
17041         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
17042         (make_pass_ipa_chkp_produce_thunks): Likewise.
17043         (make_pass_chkp): Likewise.
17044         (make_pass_chkp_opt): Likewise.
17045         (make_pass_chkp_instrumentation_passes): Likewise.
17046         * tree-pretty-print.c (dump_generic_node): Likewise.
17047         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
17048         * tree-ssa-dce.c (propagate_necessity): Likewise.
17049         (eliminate_unnecessary_stmts): Likewise.
17050         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
17051         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
17052         * tree-ssa-sccvn.h: Likewise.
17053         * tree-ssa-strlen.c (get_string_length): Likewise.
17054         (valid_builtin_call): Likewise.
17055         (adjust_last_stmt): Likewise.
17056         (handle_builtin_strchr): Likewise.
17057         (handle_builtin_strcpy): Likewise.
17058         (handle_builtin_stxncpy): Likewise.
17059         (handle_builtin_memcpy): Likewise.
17060         (handle_builtin_strcat): Likewise.
17061         (strlen_check_and_optimize_stmt): Likewise.
17062         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
17063         * tree-streamer-in.c: Likewise.
17064         * tree-streamer.c (record_common_node): Likewise.
17065         * tree.c (tree_code_size): Likewise.
17066         (wide_int_to_tree_1): Likewise.
17067         (type_contains_placeholder_1): Likewise.
17068         (build_common_tree_nodes): Likewise.
17069         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
17070         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
17071         (POINTER_BOUNDS_P): Likewise.
17072         (BOUNDED_TYPE_P): Likewise.
17073         (BOUNDED_P): Likewise.
17074         (CALL_WITH_BOUNDS_P): Likewise.
17075         (pointer_bounds_type_node): Likewise.
17076         * value-prof.c (gimple_ic): Likewise.
17077         * var-tracking.c (vt_add_function_parameters): Likewise.
17078         * varasm.c (make_decl_rtl): Likewise.
17079         (assemble_start_function): Likewise.
17080         (output_constant): Likewise.
17081         (maybe_assemble_visibility): Likewise.
17082         * varpool.c (ctor_for_folding): Likewise.
17083         * chkp-builtins.def: Remove.
17084         * ipa-chkp.c: Remove.
17085         * ipa-chkp.h: Remove.
17086         * rtl-chkp.c: Remove.
17087         * rtl-chkp.h: Remove.
17088         * tree-chkp-opt.c: Remove.
17089         * tree-chkp.c: Remove.
17090         * tree-chkp.h: Remove.
17092 2018-06-07  Carl Love  <cel@us.ibm.com>
17094         * config/rs6000/vsx.md (vextract_fp_from_shorth,
17095         vextract_fp_from_shortl): Add BE support.
17097 2018-06-07  Paul Koning  <ni1d@arrl.net>
17099         * compare-elim.c (try_merge_compare): Don't merge compare if
17100         address contains a side effect.
17101         (try_eliminate_compare): Likewise.
17103 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
17105         * config.gcc: Support "tremont".
17106         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
17107         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17108         PROCESSOR_TREMONT.
17109         * config/i386/i386.c (m_TREMONT): Define.
17110         (processor_target_table): Add "tremont".
17111         (PTA_TREMONT): Define.
17112         (ix86_lea_outperforms): Add TARGET_TREMONT.
17113         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
17114         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
17115         and M_INTEL_GOLDMONT_PLUS.
17116         (fold_builtin_cpu): Add "tremont".
17117         (ix86_add_stmt_cost): Add TARGET_TREMONT.
17118         (ix86_option_override_internal): Add "tremont".
17119         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
17120         (processor_type): Add PROCESSOR_TREMONT.
17121         * config/i386/x86-tune.def: Add m_TREMONT.
17122         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
17124 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17126         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
17127         symbol defined for msp430i* devices to be lower case.
17129 2018-06-07  Richard Biener  <rguenther@suse.de>
17131         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
17132         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
17133         Properly wrap signed arithmetic if overflow wraps.
17135 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
17137         PR tree-optimization/69615
17138         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
17139         of a cast from a same precision integral SSA_NAME in a bb dominated
17140         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
17141         cast to utype if rhs2 has already a compatible type.
17143 2018-06-07  Richard Biener  <rguenther@suse.de>
17145         PR tree-optimization/85935
17146         * graphite-scop-detection.c (find_params_in_bb): Analyze
17147         condition operands with respect to the correct loop.  Assert
17148         the analysis doesn't fail.
17150 2018-06-04  Carl Love  <cel@us.ibm.com>
17152         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
17153         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
17154         as it is slightly cheaper.
17155         (first_match_or_eos_index_<mode>):
17156         Calculate index using natural element order.
17157         (first_match_index_<mode>):
17158         Calculate index using natural element order.
17159         (first_match_or_eos_index_<mode>):
17160         Calculate index using natural order.
17161         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
17162         for BE and LE modes.
17163         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
17164         P9V_BUILTIN_VCLZLSBB_V16QI.
17165         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
17166         specific.
17168 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17170         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
17171         indentation and line wrap for many prototypes.  Add missing
17172         @smallexample directives around block of prototypes for vec_xl and
17173         vec_xst.
17175 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
17177         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
17178         track if we pass or return IEEE 128-bit floating point.
17179         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
17180         C++ mangling that is compatible with GCC 8.1.
17181         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
17182         (init_cumulative_args): Note if we pass or return IEEE 128-bit
17183         floating point types.
17184         (rs6000_function_arg_advance_1): Likewise.
17185         (rs6000_mangle_type): Optionally generate mangled names that match
17186         what GCC 8.1 generated for IEEE 128-bit floating point types.
17187         (rs6000_globalize_decl_name): If we have an external function that
17188         passes or returns IEEE 128-bit types, generate a weak reference
17189         from the mangled name used in GCC 8.1 to the current mangled
17190         name.
17191         (rs6000_init_builtins): Make __ibm128 use the long double type if
17192         long double is IBM extended double.  Make __float128 use the long
17193         double type if long double is IEEE 128-bit.
17195         PR target/85657
17196         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
17197         macro for __ibm128 built-in functions.
17198         (PACK_IF): Add __ibm128 pack/unpack functions.
17199         (UNPACK_IF): Likewise.
17200         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
17201         enable long double built-in functions if long double is IEEE
17202         128-bit floating point.
17203         (rs6000_invalid_builtin): Update long double built-in function
17204         error message.
17205         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
17206         functions, adjust the built-in function to use the long double
17207         built-in function if __ibm128 and long double are the same type.
17208         * doc/extend.texi (PowerPC builtins): Update documention for
17209         __builtin_{,un}pack_longdouble.  Add documentation for
17210         __builtin_{,un}pack_ibm128.
17212 2018-06-06  Jim Wilson  <jimw@sifive.com>
17214         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
17215         (struct machine_function): New field interrupt_mode.
17216         (riscv_handle_type_attribute): New function.  Add forward declaration.
17217         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
17218         (riscv_expand_epilogue): Check interrupt_mode field.
17219         (riscv_set_current_function): Check interrupt attribute args and
17220         set interrupt_mode field.
17221         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
17222         (riscv_sret, riscv_uret): New.
17223         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
17224         new arguments to interrupt attribute.
17226 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
17228         PR target/63177
17229         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
17230         Don't handle -mcpu=power8 if -mpower9-vector is also used.
17232 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17234         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
17235         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
17236         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17237         several redundant entries.
17239 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
17241         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
17242         type from "rtx" to "rtx_insn *".
17243         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
17244         for local "call_insn", removing cast.
17245         (ix86_expand_call): Likewise, introducing a "call_insn" local.
17247 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
17249         PR tree-optimization/86066
17250         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
17251         for BIT_INSERT_EXPR stores.
17253 2018-06-06  Richard Biener  <rguenther@suse.de>
17255         PR tree-optimization/86062
17256         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
17257         component refs ontop
17258         of to be offsetted base.
17260 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17262         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
17263         to be static and remove check on interrupt attribute name.
17265 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17267         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17268         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
17270 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
17272         PR target/79924
17273         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
17274         second argument.
17275         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
17276         Remove second argument, change how error is called.
17277         (aarch64_layout_arg): Remove second argument from
17278         aarch64_err_no_fpadvsimd call.
17279         (aarch64_init_cumulative_args): Ditto.
17280         (aarch64_gimplify_va_arg_expr): Ditto.
17281         * config/aarch64/aarch64.md (mov<mode>): Ditto.
17283 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
17285         * config/i386/i386.md (simple_return_indirect_internal): New expander.
17286         (*simple_return_indirect_internal<mode>): Rename from
17287         simple_return_indirect_internal.  Use W mode iterator.
17288         (rstorssp): New expander.
17289         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
17290         (clrssbsy): New expander.
17291         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
17293 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17295         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
17296         __typeof__.
17297         (cmse_check_pointed_object): Likewise.
17299 2018-06-05  Martin Liska  <mliska@suse.cz>
17301         PR gcov-profile/47618
17302         * doc/invoke.texi: Document how -fprofile-dir format
17303         is extended.
17305 2018-06-05  Richard Biener  <rguenther@suse.de>
17307         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
17308         removal pretend DOM info isn't available so we do not update
17309         it and only remove edges, not dominated blocks.  Actually free
17310         DOM info in case we removed something.  Remove unreachable blocks.
17311         (mfb_keep_latches): Work with either DOM info or marked backedges.
17312         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
17313         first.  Mark backedges if DOM info isn't available.
17314         (Re-)compute DOM info after cleanup_control_flow_pre.
17316 2018-06-05  Richard Biener  <rguenther@suse.de>
17318         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
17319         (locus_discrim_hasher::hash): Adjust.
17320         (locus_discrim_hasher::equal): Likewise.
17321         (next_discriminator_for_locus): Work on line directly.
17322         (same_line_p): Pass in expanded locus1 as well.
17323         (assign_discriminators): Avoid redundant location expansions.
17325 2018-06-05  Richard Biener  <rguenther@suse.de>
17327         PR tree-optimization/86046
17328         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
17329         if required after clearing TREE_ADDRESSABLE.
17331 2018-06-05  Richard Biener  <rguenther@suse.de>
17333         PR tree-optimization/86047
17334         * tree-ssa-loop.c (for_each_index): Glob handling of all
17335         decls and constants and really handle all of them.
17337 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17339         PR target/81497
17340         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
17341         qualifier_void_pointer and qualifier_const_void_pointer.
17342         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
17343         (arm_init_builtins): Handle the above.
17344         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
17345         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
17346         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
17347         void intrinsics.
17349 2018-06-05  Martin Liska  <mliska@suse.cz>
17351         * auto-profile.c (read_autofdo_file): Do not use
17352         gcov_ctr_summary struct.
17353         (afdo_callsite_hot_enough_for_early_inline): Likewise.
17354         * coverage.c (struct counts_entry): Likewise.
17355         (read_counts_file): Read just single summary entry.
17356         (get_coverage_counts): Use gcov_summary struct.
17357         * coverage.h (get_coverage_counts): Likewise.
17358         * gcov-dump.c (dump_working_sets): Likewise.
17359         (tag_summary): Dump just single summary.
17360         * gcov-io.c (gcov_write_summary): Write just histogram
17361         summary.
17362         (gcov_read_summary): Read just single summary.
17363         (compute_working_sets): Use gcov_summary struct.
17364         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
17365         of GCOV_COUNTERS_SUMMABLE.
17366         (GCOV_COUNTERS_SUMMABLE): Remove.
17367         (GCOV_FIRST_VALUE_COUNTER): Replace with
17368         GCOV_COUNTER_V_INTERVAL.
17369         (struct gcov_ctr_summary): Remove.
17370         (struct gcov_summary): Directly use fields of former
17371         gcov_ctr_summary.
17372         (compute_working_sets): Use gcov_summary struct.
17373         * gcov.c (read_count_file): Do not use ctrs fields.
17374         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
17375         struct.
17376         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
17377         struct.
17378         * profile.c: Likewise.
17379         * profile.h: Likewise.
17381 2018-06-05  Martin Liska  <mliska@suse.cz>
17383         PR gcov-profile/84846
17384         * gcov.c (output_lines): Print working directory only
17385         in intermediate format.
17387 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
17389         * config/s390/s390-builtin-types.def: Add void function type.
17390         * config/s390/s390-builtins.def: Use the function type for the
17391         tbeginc builtin.
17393 2018-06-04  Jim Wilson  <jimw@sifive.com>
17395         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
17396         to int.
17397         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
17398         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
17399         handle EH_RETURN_DATA_REGNO registers properly.
17400         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
17401         (riscv_expand_epilogue): Update comment.  Change argument name and
17402         type.  Update code to use new name and type.  Pass new args to
17403         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
17404         EXCEPTION_RETURN.
17405         * config/riscv/riscv.md (NORMAL_RETURN): New.
17406         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
17407         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
17408         (eh_return): Call gen_eh_return_internal and emit barrier.
17409         (eh_return_internal): Call riscv_expand_epilogue.
17411 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
17413         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
17414         bit_insertion field and declare can_be_merged_into method.
17415         (merged_store_group::can_be_merged_into): New method.
17416         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
17417         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
17418         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
17420 2018-06-04  Richard Biener  <rguenther@suse.de>
17422         PR tree-optimization/85955
17423         * builtins.c (fold_builtin_sincos): Convert pointers to
17424         destination to appropriate type before dereferencing.
17426 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
17428         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
17430 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
17432         * expr.c (expand_expr_real_1): Force the operand into memory if
17433         its TYPE_MODE is BLKmode and if there is no integer mode for
17434         the number of bits being extracted.
17436 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17438         PR target/85832
17439         PR target/86036
17440         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17441         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
17443 2018-06-04  Richard Biener  <rguenther@suse.de>
17445         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
17446         (cleanup_tree_cfg_noloop): ... single caller.  Do
17447         start_recording_case_labels later.
17449 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
17451         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
17452         to _IMMINTRIN_H_INCLUDED.
17453         * config/i386/pconfigintrin.h: Ditto.
17454         * config/i386/waitpkgintrin.h: Ditto.
17455         * config/i386/immintrin.h: Add includes for sgxintrin.h,
17456         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
17457         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
17458         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
17459         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
17460         waitpkgintrin.h and cldemoteintrin.h.
17462 2018-06-04  Richard Biener  <rguenther@suse.de>
17464         PR tree-optimization/86038
17465         * tracer.c (find_best_successor): Check probability for
17466         being initialized, bail out if not.
17468 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
17470         PR target/86003
17471         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
17472         of bits to ignore when comparing architectures.
17474 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17476         PR tree-optimization/69615
17477         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
17478         maximum or minimum of the type, try to merge it also as if
17479         range1 is + [-, x - 1] or + [x + 1, -].
17481         PR c++/86025
17482         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
17484 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
17486         PR tree-optimization/86034
17487         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
17488         the unsigned bitfield type in a bit insertion sequence if it does not
17489         have a larger precision than the bitfield size.
17490         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
17492 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
17494         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
17496 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17498         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
17499         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
17500         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
17501         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
17503 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17505         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
17506         Disable -fdelete-null-pointer-checks for ELF toolchain.
17508 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17509             Kito Cheng  <kito.cheng@gmail.com>
17511         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
17512         (nds32le-*-*, nds32be-*-*): Integrate checking process.
17513         (nds32*-*-*): Add glibc and uclibc conditions.
17514         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
17515         (TARGET_EXCEPT_UNWIND_INFO): Define.
17516         * config/nds32/elf.h: New file.
17517         * config/nds32/linux.h: New file.
17518         * config/nds32/nds32-elf.opt: New file.
17519         * config/nds32/nds32-linux.opt: New file.
17520         * config/nds32/nds32-fp-as-gp.c
17521         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
17522         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
17523         TARGET_LINUX_ABI.
17524         (nds32_asm_file_end): Ditto.
17525         (nds32_print_operand): Ditto.
17526         (nds32_insert_attributes): Ditto.
17527         (nds32_init_libfuncs): New function.
17528         (TARGET_HAVE_TLS): Define.
17529         (TARGET_INIT_LIBFUNCS): Define.
17530         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
17531         spec content.
17532         (TARGET_ELF): Apply different mcmodel setting.
17533         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
17534         been migrated into elf.h and linux.h files.
17535         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
17536         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
17537         (mcmodel): The content has been migrated into nds32-elf.opt and
17538         nds32-linux.opt files.
17539         * config/nds32/t-elf: New file.
17540         * config/nds32/t-linux: New file.
17542 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17543             Shiva Chen  <shiva0217@gmail.com>
17545         * config/nds32/constants.md (unspec_volatile_element): Add
17546         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
17547         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
17548         optimization.
17549         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
17550         (make_pass_nds32_fp_as_gp): Declare.
17551         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
17552         optmization pass.
17553         (nds32_asm_function_end_prologue): Remove unused asm output.
17554         (nds32_asm_function_begin_epilogue): Remove unused asm output.
17555         (nds32_asm_file_start): Output necessary fp_as_gp information.
17556         (nds32_option_override): Adjust register usage.
17557         (nds32_expand_prologue): Consider fp_as_gp situation.
17558         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
17559         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
17560         (epilogue): Ditto.
17561         (return): Ditto.
17562         (simple_return): Ditto.
17563         (omit_fp_begin): Output special directive for fp_as_gp.
17564         (omit_fp_end): Output special directive for fp_as_gp.
17565         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
17566         mforbid-fp-as-gp): New options.
17568 2018-06-01  Mark Wielaard  <mark@klomp.org>
17570         * dwarf2out.c (dwarf2out_finish): Remove generation of
17571         DW_AT_loclists_base.
17573 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
17575         * gimple-ssa-store-merging.c: Include gimple-fold.h.
17576         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
17577         (struct merged_store_group): Add bit_insertion field.
17578         (dump_char_array): Use standard hexadecimal format.
17579         (merged_store_group::merged_store_group): Set bit_insertion to false.
17580         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
17581         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
17582         also print the mask in the dump file.
17583         (pass_store_merging::gate): Minor tweak.
17584         (imm_store_chain_info::coalesce_immediate): Fix wrong association
17585         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
17586         stores with INTEGER_CST stores.
17587         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
17588         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
17589         and use it throughout.  Generate bit insertion sequences if need be.
17590         (pass_store_merging::process_store): Remove redundant condition.
17591         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
17593 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
17595         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
17596         the 128-bit floating point types.  Fix function comment.
17598 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17600         * config/aarch64/aarch64-simd.md
17601         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
17602         mnemonics.
17603         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
17604         mnemonics.
17606 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17608         PR tree-optimization/85989
17609         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
17610         variable.
17611         (backprop::intersect_uses): Check it when deciding whether this
17612         is a backedge reference.
17613         (backprop::process_block): Add each phi to m_visited_phis
17614         after visiting it, then clear it at the end.
17616 2018-06-01  Richard Biener  <rguenther@suse.de>
17618         * tree-vectorizer.h (vect_dr_stmt): New function.
17619         (vect_get_load_cost): Adjust.
17620         (vect_get_store_cost): Likewise.
17621         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
17622         Use vect_dr_stmt instead of DR_SMTT.
17623         (vect_record_base_alignments): Likewise.
17624         (vect_calculate_target_alignment): Likewise.
17625         (vect_compute_data_ref_alignment): Likewise and make static.
17626         (vect_update_misalignment_for_peel): Likewise.
17627         (vect_verify_datarefs_alignment): Likewise.
17628         (vector_alignment_reachable_p): Likewise.
17629         (vect_get_data_access_cost): Likewise.  Pass down
17630         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
17631         (vect_get_peeling_costs_all_drs): Likewise.
17632         (vect_peeling_hash_get_lowest_cost): Likewise.
17633         (vect_enhance_data_refs_alignment): Likewise.
17634         (vect_find_same_alignment_drs): Likewise.
17635         (vect_analyze_data_refs_alignment): Likewise.
17636         (vect_analyze_group_access_1): Likewise.
17637         (vect_analyze_group_access): Likewise.
17638         (vect_analyze_data_ref_access): Likewise.
17639         (vect_analyze_data_ref_accesses): Likewise.
17640         (vect_vfa_segment_size): Likewise.
17641         (vect_small_gap_p): Likewise.
17642         (vectorizable_with_step_bound_p): Likewise.
17643         (vect_prune_runtime_alias_test_list): Likewise.
17644         (vect_analyze_data_refs): Likewise.
17645         (vect_supportable_dr_alignment): Likewise.
17646         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
17647         (vect_gen_prolog_loop_niters): Likewise.
17648         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
17649         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
17650         modify DR_STMT.
17651         (vect_recog_mask_conversion_pattern): Likewise.
17652         (vect_try_gather_scatter_pattern): Likewise.
17653         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
17654         to vect_get_store_cost.
17655         (vect_get_store_cost): Get stmt_info instead of DR.
17656         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
17657         (vect_get_load_cost): Get stmt_info instead of DR.
17659 2018-06-01  Richard Biener  <rguenther@suse.de>
17661         PR middle-end/86017
17662         * gimple-fold.c (var_decl_component_p): Also allow offsetted
17663         vars wrapped in MEM_REFs.
17665 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17667         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17668         Fix subreg tests so that we only return a choice between
17669         GENERAL_REGS and FP_REGS if the original classes included both.
17671 2018-06-01  Richard Biener  <rguenther@suse.de>
17673         PR ipa/85960
17674         * tree-ssa-structalias.c (get_function_part_constraint):
17675         Handle NULL fi->decl.
17676         (find_func_aliases_for_call): Properly handle indirect
17677         fi from direct call.
17678         (find_func_clobbers): Likewise.
17679         (ipa_pta_execute): Likewise.
17680         (create_variable_info_for): For functions that are ifunc_resolver
17681         resolve to a varinfo that contains the result of the resolver call.
17682         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
17683         aliases.
17685 2018-05-31  Michael Collison  <michael.collison@arm.com>
17687         * config/aarch64/aarch64.md:
17688         (*fix_to_zero_extenddfdi2): New pattern.
17689         * gcc.target/aarch64/fix_extend1.c: New testcase.
17691 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
17693         PR middle-end/78809
17694         PR middle-end/83026
17695         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
17696         and BUILT_IN_STRNCMP_EQ.
17697         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
17698         BUILT_IN_STRNCMP_EQ.
17699         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
17700         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17701         (gimple_fold_builtin): Likewise.
17702         * tree-ssa-strlen.c (compute_string_length): New function.
17703         (determine_min_obsize): New function.
17704         (handle_builtin_string_cmp): New function to handle calls to
17705         string compare functions.
17706         (strlen_optimize_stmt): Add handling to builtin string compare
17707         calls.
17708         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
17709         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17710         * tree.c (build_common_builtin_nodes): Add new defines of
17711         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
17713 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
17715         PR target/85984
17716         * bb-reorder.c (pass_partition_blocks::gate): Return false for
17717         functions with naked attribute.
17719 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
17721         * config/i386/sse.md (avx_vec_concat<mode>):
17722         Substitute concat_tg_mode mode attribute with xtg_mode.
17723         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
17724         (concat_tg_mode): Remove mode attribute.
17726 2018-05-31  Martin Sebor  <msebor@redhat.com>
17728         PR c/82063
17729         * calls.c (alloc_max_size): Correct a logic error/typo.
17730         Treat excessive arguments as infinite.  Warn for invalid arguments.
17731         * doc/invoke.texi (-Walloc-size-larger-than): Update.
17733 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
17735         PR target/85829
17736         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
17737         and movx for Haswell.
17739 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
17740             Cesar Philippidis  <cesar@codesourcery.com>
17742         PR middle-end/85879
17743         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
17744         when emitting error on private/firstprivate reductions.
17745         * omp-low.c (lower_omp_target): Avoid reference-type processing
17746         on pointers for firstprivate clause.
17748 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17750         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
17751         (st1x2): Likewise.
17752         (st1x3): Likewise.
17753         * config/aarch64/aarch64-simd.md
17754         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
17755         (aarch64_ld1_x3_<mode>): Likewise
17756         (aarch64_st1x2<VALLDIF:mode>): Likewise
17757         (aarch64_st1_x2_<mode>): Likewise
17758         (aarch64_st1x3<VALLDIF:mode>): Likewise
17759         (aarch64_st1_x3_<mode>): Likewise
17760         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
17761         (vld1_s8_x3): Likewise.
17762         (vld1_u16_x3): Likewise.
17763         (vld1_s16_x3): Likewise.
17764         (vld1_u32_x3): Likewise.
17765         (vld1_s32_x3): Likewise.
17766         (vld1_u64_x3): Likewise.
17767         (vld1_s64_x3): Likewise.
17768         (vld1_f16_x3): Likewise.
17769         (vld1_f32_x3): Likewise.
17770         (vld1_f64_x3): Likewise.
17771         (vld1_p8_x3): Likewise.
17772         (vld1_p16_x3): Likewise.
17773         (vld1_p64_x3): Likewise.
17774         (vld1q_u8_x3): Likewise.
17775         (vld1q_s8_x3): Likewise.
17776         (vld1q_u16_x3): Likewise.
17777         (vld1q_s16_x3): Likewise.
17778         (vld1q_u32_x3): Likewise.
17779         (vld1q_s32_x3): Likewise.
17780         (vld1q_u64_x3): Likewise.
17781         (vld1q_s64_x3): Likewise.
17782         (vld1q_f16_x3): Likewise.
17783         (vld1q_f32_x3): Likewise.
17784         (vld1q_f64_x3): Likewise.
17785         (vld1q_p8_x3): Likewise.
17786         (vld1q_p16_x3): Likewise.
17787         (vld1q_p64_x3): Likewise.
17788         (vst1_s64_x2): Likewise.
17789         (vst1_u64_x2): Likewise.
17790         (vst1_f64_x2): Likewise.
17791         (vst1_s8_x2): Likewise.
17792         (vst1_p8_x2): Likewise.
17793         (vst1_s16_x2): Likewise.
17794         (vst1_p16_x2): Likewise.
17795         (vst1_s32_x2): Likewise.
17796         (vst1_u8_x2): Likewise.
17797         (vst1_u16_x2): Likewise.
17798         (vst1_u32_x2): Likewise.
17799         (vst1_f16_x2): Likewise.
17800         (vst1_f32_x2): Likewise.
17801         (vst1_p64_x2): Likewise.
17802         (vst1q_s8_x2): Likewise.
17803         (vst1q_p8_x2): Likewise.
17804         (vst1q_s16_x2): Likewise.
17805         (vst1q_p16_x2): Likewise.
17806         (vst1q_s32_x2): Likewise.
17807         (vst1q_s64_x2): Likewise.
17808         (vst1q_u8_x2): Likewise.
17809         (vst1q_u16_x2): Likewise.
17810         (vst1q_u32_x2): Likewise.
17811         (vst1q_u64_x2): Likewise.
17812         (vst1q_f16_x2): Likewise.
17813         (vst1q_f32_x2): Likewise.
17814         (vst1q_f64_x2): Likewise.
17815         (vst1q_p64_x2): Likewise.
17816         (vst1_s64_x3): Likewise.
17817         (vst1_u64_x3): Likewise.
17818         (vst1_f64_x3): Likewise.
17819         (vst1_s8_x3): Likewise.
17820         (vst1_p8_x3): Likewise.
17821         (vst1_s16_x3): Likewise.
17822         (vst1_p16_x3): Likewise.
17823         (vst1_s32_x3): Likewise.
17824         (vst1_u8_x3): Likewise.
17825         (vst1_u16_x3): Likewise.
17826         (vst1_u32_x3): Likewise.
17827         (vst1_f16_x3): Likewise.
17828         (vst1_f32_x3): Likewise.
17829         (vst1_p64_x3): Likewise.
17830         (vst1q_s8_x3): Likewise.
17831         (vst1q_p8_x3): Likewise.
17832         (vst1q_s16_x3): Likewise.
17833         (vst1q_p16_x3): Likewise.
17834         (vst1q_s32_x3): Likewise.
17835         (vst1q_s64_x3): Likewise.
17836         (vst1q_u8_x3): Likewise.
17837         (vst1q_u16_x3): Likewise.
17838         (vst1q_u32_x3): Likewise.
17839         (vst1q_u64_x3): Likewise.
17840         (vst1q_f16_x3): Likewise.
17841         (vst1q_f32_x3): Likewise.
17842         (vst1q_f64_x3): Likewise.
17843         (vst1q_p64_x3): Likewise.
17845 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17847         * config/msp430/msp430.c (msp430_output_labelref): Prepend
17848         user_label_prefix to name.
17850         * tree-core.h: Update comment about the format of NAME string
17851         passed to handler in attribute_spec.
17853         * config/msp430/msp430.md: Remove erroneous subreg expression from
17854         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
17855         zero_extend{q,h}isi2.
17857 2018-05-30  Borislav Petkov  <bp@suse.de>
17859         * doc/extend.texi: Document some architecture specific
17860         constraints and sort entries.
17862 2018-05-30  Martin Sebor  <msebor@redhat.com>
17864         PR middle-end/85369
17865         * builtins.c (expand_builtin_stpcpy_1): New function.
17866         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
17867         only if the former succeeds.
17869 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17871         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
17872         in saphira.
17874 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17876         * doc/invoke.texi (-flinker-output): Document
17878 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17880         * passes.c (ipa_write_summaries): Only modify statements if body
17881         is in memory.
17882         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
17883         incrementally linking.
17884         (ipa_passes): Likewise.
17885         * lto-cgraph.c (lto_output_node): When incrementally linking do not
17886         pass down resolution info.
17887         * common.opt (flag_incremental_link): Update info.
17888         * gcc.c (plugin specs): Turn flinker-output=* to
17889         -plugin-opt=-linker-output-known
17890         * toplev.c (compile_file): Also cut compilation when doing incremental
17891         link.
17892         * flag-types. (enum lto_partition_model): Add
17893         LTO_LINKER_OUTPUT_NOLTOREL.
17894         (invoke.texi): Add -flinker-output docs.
17895         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
17896         link same way as WPA; do not stream in dead initializers.
17898         * dwarf2out.c (dwarf2out_die_ref_for_decl,
17899         darf2out_register_external_decl): Support incremental link.
17901 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17903         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
17905 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17907         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
17908         it down to simple_object_copy_lto_debug_sections.
17909         (run_gcc): Determine incremental LTO link time and configure
17910         lto1 into non-wpa mode, disable renaming of debug sections.
17912 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17914         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17915         descriptions of various incorrectly documented functions.
17917 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17919         Revert:
17920         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17921         address check not strict.
17923 2018-05-30  Richard Biener  <rguenther@suse.de>
17925         PR tree-optimization/85964
17926         * tracer.c (better_p): Drop initialized count check, we only
17927         call the function with initialized counts now.
17928         (find_best_successor): Do find a best edge if one
17929         has uninitialized count.
17930         (find_best_predecessor): Likewise.  Do BB frequency check only
17931         if count is initialized.
17933 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
17935         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
17936         (aarch64_ldrstr_offset_compare): New.
17937         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
17938         load/store orderings.
17939         (aarch64_gen_adjusted_ldpstp): Likewise.
17941 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
17943         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17944         Check for subset of GENERAL_REGS and FP_REGS.
17945         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
17946         r=w alternative.
17948 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
17950         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
17951         and wi::to_poly_offset.  Add the current offset and then check
17952         whether the sum fits, rather than using an unchecked addition of
17953         a checked term.  Check for a shwi rather than a uhwi.
17954         * expr.c (get_bit_range): Use tree_to_poly_uint64.
17955         (store_constructor): Use poly_int_tree_p.
17956         (expand_expr_real_1): Likewise.
17957         * function.c (assign_temp): Likewise.
17958         * fold-const.c (const_binop): Use poly_int_tree_p and
17959         wi::to_poly_offset.
17960         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
17961         division.
17962         * ipa-icf-gimple.c (func_checker::compare_operand): Use
17963         to_poly_offset for MEM offsets.
17964         * ipa-icf.c (sem_variable::equals): Likewise.
17965         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
17966         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
17967         wi::to_poly_offset for BIT_FIELD_REF offsets.
17968         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
17969         wi::to_poly_offset.
17970         * var-tracking.c (emit_note_insn_var_location): Use
17971         tree_to_poly_uint64.
17973 2018-05-29  Jim Wilson  <jimw@sifive.com>
17975         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
17977 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
17979         PR target/85950
17980         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
17981         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
17982         sequence.
17983         (sse4_1_round<mode>2): Use nonimmediate_operand
17984         for operand 1 predicate.
17986 2018-05-29  Martin Sebor  <msebor@redhat.com>
17987             Richard Biener  <rguenther@suse.de>
17989         PR testsuite/85888
17990         * calls.c (get_size_range): Call determine_value_range instead
17991         of get_value_range..
17992         * tree-vrp.h (determine_value_range): Declared new function.
17993         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
17995 2018-05-29  Richard Biener  <rguenther@suse.de>
17997         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
17998         sure to use non-pattern stmts for get_earlier_stmt arguments.
17999         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
18000         called on pattern stmts.
18001         (get_later_stmt): Likewise.
18003 2018-05-29  Martin Liska  <mliska@suse.cz>
18005         PR gcov-profile/85759
18006         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
18007         env variables.
18009 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
18011         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
18012         VEC_UNPACK_*_EXPR.
18013         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
18014         VEC_PACK_*_EXPR.
18016         PR target/85918
18017         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
18018         VEC_PACK_FLOAT_EXPR): New tree codes.
18019         * tree-pretty-print.c (op_code_prio): Handle
18020         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
18021         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
18022         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
18023         * tree-inline.c (estimate_operator_cost): Likewise.
18024         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
18025         * fold-const.c (const_binop): Likewise.
18026         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
18027         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
18028         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
18029         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
18030         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
18031         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
18032         * expr.c (expand_expr_real_2): Likewise.
18033         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
18034         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
18035         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
18036         optabs.
18037         * optabs.c (expand_widen_pattern_expr): For
18038         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
18039         sign from result type rather than operand's type.
18040         (expand_binop_directly): For vec_packu_float_optab and
18041         vec_packs_float_optab allow result type to be different from operand's
18042         type.
18043         * optabs-tree.c (optab_for_tree_code): Handle
18044         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
18045         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
18046         * tree-vect-generic.c (expand_vector_operations_1):  Handle
18047         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
18048         VEC_PACK_FLOAT_EXPR.
18049         * tree-vect-stmts.c (supportable_widening_operation): Handle
18050         FIX_TRUNC_EXPR.
18051         (supportable_narrowing_operation): Handle FLOAT_EXPR.
18052         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
18053         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
18054         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
18055         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
18056         mode attributes.
18057         (vec_pack<floatprefix>_float_<mode>): New expander.
18058         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
18059         attributes.
18060         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
18061         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
18062         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
18063         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
18064         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
18065         Document.
18066         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
18067         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
18068         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
18069         VEC_PACK_FLOAT_EXPR): Document.
18071 2018-05-29  Richard Biener  <rguenther@suse.de>
18073         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
18074         member.
18075         (stmt_vec_info_vec): Make pointer.
18076         (init_stmt_vec_info_vec): Remove.
18077         (free_stmt_vec_info_vec): Likewise.
18078         (set_stmt_vec_info_vec): New function.
18079         (free_stmt_vec_infos): Likewise.
18080         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
18081         (set_vinfo_for_stmt): Likewise.
18082         (get_earlier_stmt): Likewise.
18083         (get_later_stmt): Likewise.
18084         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
18085         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
18086         (vec_info::~vec_info): Free stmt_vec_infos.
18087         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
18088         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
18089         (pass_slp_vectorize::execute): Likewise.
18090         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
18091         (free_stmt_vec_info_vec): Likewise.
18092         (set_stmt_vec_info_vec): New function.
18093         (free_stmt_vec_infos): Likewise.
18094         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
18095         the global stmt_vec_info_vec.
18096         * tree-parloops.c (gather_scalar_reductions): Use
18097         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
18098         vector.
18100 2018-05-29  Richard Biener  <rguenther@suse.de>
18102         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
18104 2018-05-29  Martin Liska  <mliska@suse.cz>
18105             David Malcolm  <dmalcolm@redhat.com>
18107         * vec.c (test_reverse): New.
18108         (vec_c_tests): Add new test.
18109         * vec.h (vl_ptr>::reverse): New function.
18111 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
18113         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
18115         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
18116         and later.
18118 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
18120         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
18122 2018-05-28  Richard Biener  <rguenther@suse.de>
18124         PR tree-optimization/85933
18125         * tree-vect-data-refs.c (vect_record_base_alignments): Only
18126         look at stmts marked as vectorizable.
18128 2018-05-28  Richard Biener  <rguenther@suse.de>
18130         PR tree-optimization/85934
18131         * tree-vect-generic.c (expand_vector_operations_1): Hoist
18132         vector boolean check before scalar optimization.
18134 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
18136         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
18137         for armv5te.
18139 2018-05-28  Mark Wielaard  <mark@klomp.org>
18141         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
18142         if it is an expression containing a minus sign.
18144 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
18146         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
18148 2018-05-27  Paul Koning  <ni1d@arrl.net>
18150         * config/pdp11/pdp11.md (truncsihi2): Remove.
18152 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
18153             Chung-Ju Wu  <jasonwucj@gmail.com>
18155         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
18156         implementation.
18157         (unaligned_store_dw): Ditto.
18158         * config/nds32/nds32-memory-manipulation.c
18159         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
18160         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
18161         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
18162         (emit_setmem_word_loop): Rename to ...
18163         (emit_setmem_doubleword_loop): ... this.
18164         (nds32_gen_dup_4_byte_to_word_value): New function.
18165         (nds32_gen_dup_8_byte_to_double_word_value): New function.
18166         (nds32_expand_setmem_loop): Refine implementation.
18167         (nds32_expand_setmem_loop_v3m): Ditto.
18168         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
18169         pattern.
18171 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18173         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
18175 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18177         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
18178         (nds32_init_machine_status): Initialize machine->attr_naked_p and
18179         machine->attr_no_prologue_p.
18180         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
18181         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
18182         (nds32_expand_epilogue): Consider attr_naked_p.
18183         (nds32_expand_epilogue_v3pop): Likewise.
18184         (nds32_can_use_return_insn): Likewise.
18185         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
18186         attr_no_prologue_p fields.
18187         * config/nds32/nds32.opt (mret-in-naked-func): New option.
18189 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
18191         PR target/85918
18192         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
18193         attributes.
18194         * config/i386/sse.md
18195         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
18196         Rename to ...
18197         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
18198         ... this.
18199         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
18200         Rename to ...
18201         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
18202         ... this.
18203         (*<floatsuffix>floatv2div2sf2): Rename to ...
18204         (*float<floatunssuffix>v2div2sf2): ... this.
18205         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
18206         (float<floatunssuffix>v2div2sf2_mask): ... this.
18207         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
18208         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
18209         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
18210         to ...
18211         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
18212         ... this.
18213         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18214         Rename to ...
18215         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18216         ... this.
18217         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18218         Rename to ...
18219         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18220         ... this.
18221         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
18222         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
18223         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
18224         gen_ufix_truncv8dfv8si2.
18225         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
18226         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
18227         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
18228         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
18229         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
18230         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
18231         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
18232         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
18234 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18236         PR target/85900
18237         PR target/85345
18238         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
18240 2018-05-25  Jim Wilson  <jimw@sifive.com>
18242         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
18243         * config/riscv/riscv.c (struct machine_function): Add
18244         interrupt_handler_p and attribute_checked_p fields.
18245         (riscv_attribute_table): Add interrupt.
18246         (riscv_interrupt_type_p): New.
18247         (riscv_save_reg_p): Save extra regs for interrupt handler.
18248         (riscv_use_save_libcall): Return false  for interrupt handler.
18249         (riscv_first_stack_step): Add forward declaration.
18250         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
18251         for interrupt handler with large frame.  Use it for saved reg list.
18252         (riscv_expand_prologue): Move flag_stack_usage_info support to
18253         eliminate duplication.
18254         (riscv_expand_epilogue): Generate mret for interrupt handler.
18255         (riscv_epilogue_uses): New.
18256         (riscv_can_use_return_insn): Return false for interrupt handler.
18257         (riscv_function_ok_for_sibcall): Likewise.
18258         (riscv_set_current_function): Add interrupt handler support.
18259         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
18260         * config/riscv/riscv.md (UNSPECV_MRET): New.
18261         (GP_REGNUM): New.
18262         (riscv_frflags, riscv_fsflags): Use tab after opcode.
18263         (riscv_mret): New.
18264         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
18266 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
18268         PR tree-optimization/85712
18269         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
18270         this candidate has already been replaced in-situ by a copy.
18272 2018-05-25  Jason Merrill  <jason@redhat.com>
18274         PR c++/80485 - inline function non-zero address.
18275         * symtab.c (nonzero_address): Check DECL_COMDAT.
18277 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
18279         PR target/83628
18280         * config/alpha/alpha.md (ashlsi3): New insn pattern.
18281         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
18282         extension of SImode operation.  Use const123_operand predicate.
18283         (*saddsi_1): Remove.
18284         (*saddl_se_1): Ditto.
18285         (*ssubsi_1): Ditto.
18286         (*ssubl_se_1): Ditto.
18287         * config/alpha/predicates.md (const123_operand): New predicate.
18288         * config/alpha/constraints.md (P): Use IN_RANGE.
18290 2018-05-25  Richard Biener  <rguenther@suse.de>
18292         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
18293         defaulted to true.
18294         (ref_maybe_used_by_stmt_p): Likewise.
18295         (stmt_may_clobber_ref_p): Likewise.
18296         (stmt_may_clobber_ref_p_1): Likewise.
18297         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
18298         and pass it along.
18299         (ref_maybe_used_by_stmt_p): Likewise.
18300         (stmt_may_clobber_ref_p): Likewise.
18301         (stmt_may_clobber_ref_p_1): Likewise.
18302         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
18303         the alias oracle to disambiguate DRs with stmts DR analysis
18304         couldn't handle.
18305         (vect_analyze_data_refs): Do not give up on not analyzable
18306         DRs for BB vectorization.  Remove code truncating the dataref
18307         vector.
18309 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
18311         PR target/85832
18312         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
18313         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
18314         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
18316 2018-05-25  Richard Biener  <rguenther@suse.de>
18318         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
18319         function, combining stmt data ref gathering and fatal analysis
18320         parts.
18321         (vect_analyze_data_refs): Remove now redudnant code and simplify.
18322         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
18323         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
18324         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
18325         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
18327 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18329         PR tree-optimization/85720
18330         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
18331         SCC if all partitions are builtins.
18332         (version_loop_by_alias_check): New parameter.  Generate cancelable
18333         runtime alias check if all partitions are builtins.
18334         (distribute_loop): Update call to above function.
18336 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18338         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
18339         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
18340         (parm_default_def_partition_arg): Ditto.
18341         (set_parm_default_def_partition): Ditto.
18342         (get_parm_default_def_partitions): Ditto and make it static.
18343         (get_undefined_value_partitions): Ditto and make it static.
18344         (remove_ssa_form): Refactor call to init_var_map here.
18345         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
18346         computation for loop region.
18347         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
18348         (register_default_def): Delete.
18349         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
18350         (parm_default_def_partition_arg): Ditto.
18351         (set_parm_default_def_partition): Ditto.
18352         (get_parm_default_def_partitions): Ditto and make it static.
18353         (get_undefined_value_partitions): Ditto and make it static.
18354         (coalesce_with_default, coalesce_with_default): Update comment.
18355         (create_coalesce_list_for_region): New func factored out from
18356         create_outofssa_var_map.
18357         (populate_coalesce_list_for_outofssa): New func factored out from
18358         create_outofssa_var_map and coalesce_ssa_name.
18359         (create_outofssa_var_map): Delete.
18360         (coalesce_ssa_name): Refactor to support live range computation.
18361         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
18362         (get_parm_default_def_partitions): Delete.
18363         (get_undefined_value_partitions): Ditto.
18364         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
18365         computation for loop region.
18366         (new_tree_live_info, loe_visit_block): Ditto.
18367         (live_worklist, set_var_live_on_entry): Ditto.
18368         (calculate_live_on_exit, verify_live_on_entry): Ditto.
18369         * tree-ssa-live.h (struct _var_map): New fields.
18370         (init_var_map): Change decl.
18371         (region_contains_p): New.
18373 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18375         * tree-ssa-live.h (live_merge_and_clear): Delete.
18377 2018-05-25  Richard Biener  <rguenther@suse.de>
18379         PR c++/85912
18380         * tree-dump.c (dequeue_and_dump): Remove access to removed
18381         operand 2 of a SWITCH_EXPR.
18383 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18385         * doc/sourcebuild.texi (vect_double_cond_arith): Include
18386         multiplication and division.
18387         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
18388         (cond_udiv@var{m}, cond_umod@var{m}): Document.
18389         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
18390         (cond_udiv_optab, cond_umod_optab): New optabs.
18391         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
18392         (IFN_COND_RDIV): New internal functions.
18393         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
18394         TRUNC_MOD_EXPR and RDIV_EXPR.
18395         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
18396         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
18397         New unspecs.
18398         (SVE_INT_BINARY): Include mult.
18399         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
18400         (optab, sve_int_op): Handle mult.
18401         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
18402         UNSPEC_COND_DIV.
18403         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
18404         for SVE_INT_BINARY_SD.
18406 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18408         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
18409         (optab, sve_int_op): Handle div and udiv.
18410         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
18411         for SVE_INT_BINARY_SD.
18412         (*<optab><mode>3): New insn for the same.
18414 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18416         * tree-vect-patterns.c: Include predict.h.
18417         (vect_recog_divmod_pattern): Restrict check for division support
18418         to when optimizing for size.
18420 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18422         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
18423         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
18424         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
18425         (gimple_match_op::set_op): Likewise.
18426         (gimple_resimplify4): Declare.
18427         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
18428         (expr::gen_transform): Likewise.
18429         (decision_tree::gen): Generate a simplification routine for 4 operands.
18430         * gimple-match-head.c (gimple_simplify): Add an overload for
18431         4 operands.  In the top-level function, handle up to 4 call
18432         arguments and call gimple_resimplify4.
18433         (gimple_resimplify4): New function.
18434         (build_call_internal): Pass a fourth operand.
18435         (maybe_push_to_seq): Likewise.
18436         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
18437         Fold VEC_COND_EXPRs of an operation and a default value into
18438         an IFN_COND_* function if possible.
18439         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
18440         New unspecs.
18441         (SVE_COND_FP_BINARY): Include them.
18442         (optab, sve_fp_op): Handle them.
18443         (SVE_INT_BINARY_REV): New code iterator.
18444         (SVE_COND_FP_BINARY_REV): New int iterator.
18445         (commutative): New int attribute.
18446         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
18447         Declare.
18448         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
18449         function.
18450         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
18451         (*cond_<optab><mode>): New patterns for reversed operands.
18453 2018-05-25  Richard Biener  <rguenther@suse.de>
18455         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
18456         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
18457         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
18458         (STMT_VINFO_GROUPED_ACCESS): Adjust.
18459         * tree-vect-data-refs.c (everywhere): Adjust users.
18460         * tree-vect-loop.c (everywhere): Likewise.
18461         * tree-vect-slp.c (everywhere): Likewise.
18462         * tree-vect-stmts.c (everywhere): Likewise.
18463         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
18465 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18467         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
18468         Rename to...
18469         (gcc_cv_as_section_exclude): ... this.
18470         Try Solaris as #exclude syntax.
18471         * configure: Regenerate.
18472         * config.in: Regenerate.
18473         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
18474         SECTION_EXCLUDE.
18475         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
18476         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
18478         * varasm.c (default_elf_asm_named_section): Don't check if
18479         HAVE_GAS_SECTION_EXCLUDE is defined.
18481 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18483         * doc/md.texi: Update the documentation of the cond_* optabs
18484         to mention the new final operand.  Fix GET_MODE_NUNITS call.
18485         Describe the scalar case too.
18486         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
18487         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
18488         instead of 2.
18489         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
18490         (get_conditional_internal_fn): Update comment.
18491         * tree-vect-loop.c (vectorizable_reduction): Pass the original
18492         accumulator value as a final argument to conditional functions.
18493         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
18494         a define_expand and add an "else" operand.  Assert for now that
18495         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
18496         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
18497         (*cond_<optab><mode>): New patterns.
18498         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
18499         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
18500         (UNSPEC_COND_EOR): Delete.
18501         (optab): Remove associated mappings.
18502         (SVE_INT_BINARY): New code iterator.
18503         (sve_int_op): Remove int attribute and add "minus" to the code
18504         attribute.
18505         (SVE_COND_INT_OP): Delete.
18506         (SVE_COND_FP_OP): Rename to...
18507         (SVE_COND_FP_BINARY): ...this.
18509 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18511         * optabs.c (can_reuse_operands_p): New function.
18512         (maybe_legitimize_operands): Try to reuse the results for
18513         earlier operands.
18515 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18517         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
18518         Add {q} suffix to insn mnemonic.
18520 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18522         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
18523         (msp430_warn_func_return): New.
18525 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
18527         * fold-const.c (tree_nonzero_bits): New function.
18528         * fold-const.h (tree_nonzero_bits): Likewise.
18529         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
18530         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
18532 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18534         PR target/85900
18535         PR target/85345
18536         * varasm.c (assemble_alias): Check ifunc_resolver only on
18537         FUNCTION_DECL.
18539 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18541         PR target/85903
18542         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
18543         when memory input operand is handled.
18545 2018-05-24  Luis Machado  <luis.machado@linaro.org>
18547         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
18548         global.
18549         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
18551 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18553         * match.pd: Delay FMA folds until after vectorization.
18555 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18557         PR target/83009
18558         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
18559         address check not strict.
18561 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18563         * gimple-match.h (gimple_match_op): New class.
18564         (mprts_hook): Replace parameters with a gimple_match_op *.
18565         (maybe_build_generic_op): Likewise.
18566         (gimple_simplified_result_is_gimple_val): Replace parameters with
18567         a const gimple_match_op *.
18568         (gimple_simplify): Replace code_helper * and tree * parameters with
18569         a gimple_match_op * parameter.
18570         (gimple_resimplify1): Replace code_helper *, tree and tree *
18571         parameters with a gimple_match_op * parameter.
18572         (gimple_resimplify2): Likewise.
18573         (gimple_resimplify3): Likewise.
18574         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
18575         parameters with a gimple_match_op * parameter.
18576         * gimple-match-head.c (gimple_simplify): Change prototypes of
18577         auto-generated functions to take a gimple_match_op * instead of
18578         separate code_helper * and tree * parameters.  Make the same
18579         change in the top-level overload and update calls to the
18580         gimple_resimplify routines.  Update calls to the auto-generated
18581         functions and to maybe_push_res_to_seq in the publicly-facing
18582         operation-specific gimple_simplify overloads.
18583         (gimple_match_op::MAX_NUM_OPS): Define.
18584         (gimple_resimplify1): Replace rcode and ops with a single res_op
18585         parameter.  Update call to gimple_simplify.
18586         (gimple_resimplify2): Likewise.
18587         (gimple_resimplify3): Likewise.
18588         (mprts_hook): Replace parameters with a gimple_match_op *.
18589         (maybe_build_generic_op): Likewise.
18590         (build_call_internal): Replace type, nargs and ops with
18591         a gimple_match_op *.
18592         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
18593         with a single gimple_match_op *.  Update calls to mprts_hook,
18594         build_call_internal and gimple_simplified_result_is_gimple_val.
18595         Factor out code that is common to the tree_code and combined_fn cases.
18596         * genmatch.c (expr::gen_transform): Replace tem_code and
18597         tem_ops with a gimple_match_op called tem_op.  Update calls
18598         to the gimple_resimplify functions and maybe_push_res_to_seq.
18599         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
18600         res_ops.  Update call to the gimple_resimplify functions.
18601         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
18602         (decision_tree::gen): Make the functions take a gimple_match_op *
18603         called res_op instead of separate res_code and res_ops parameters.
18604         Update call accordingly.
18605         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
18606         and ops with a single res_op parameter.  Update calls to
18607         maybe_build_generic_op and maybe_push_res_to_seq.
18608         (fold_stmt_1): Update calls to gimple_simplify and
18609         replace_stmt_with_simplification.
18610         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
18611         and gimple_simplified_result_is_gimple_val.
18612         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
18613         gimple_simplify.
18614         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
18615         with a gimple_match_op *.
18616         (vn_nary_build_or_lookup): Likewise.  Update call to
18617         vn_nary_build_or_lookup_1.
18618         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
18619         gimple_match_op *.  Update calls to the gimple_resimplify routines
18620         and to gimple_simplified_result_is_gimple_val.
18621         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
18622         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
18623         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
18624         (visit_nary_op): Likewise.
18625         (visit_reference_op_load): Likewise.
18627 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18629         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
18630         modifier for printing the step amount.
18632 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
18634         PR target/78849
18635         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
18636         types.
18638 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
18640         * doc/sourcebuild.texi (Endianness): New subsubsection.
18642 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18644         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18645         <prefetch_dynamic_strides>: New const bool field.
18646         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18647         prefetch_dynamic_strides.
18648         (exynosm1_prefetch_tune): Likewise.
18649         (thunderxt88_prefetch_tune): Likewise.
18650         (thunderx_prefetch_tune): Likewise.
18651         (thunderx2t99_prefetch_tune): Likewise.
18652         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
18653         false.
18654         (aarch64_override_options_internal): Update to set
18655         PARAM_PREFETCH_DYNAMIC_STRIDES.
18656         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18657         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18658         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18659         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18660         prefetch-dynamic-strides setting.
18662 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18664         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18665         <minimum_stride>: New const int field.
18666         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18667         minimum_stride field defaulting to -1.
18668         (exynosm1_prefetch_tune): Likewise.
18669         (thunderxt88_prefetch_tune): Likewise.
18670         (thunderx_prefetch_tune): Likewise.
18671         (thunderx2t99_prefetch_tune): Likewise.
18672         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
18673         <default_opt_level>: Set to 3.
18674         (aarch64_override_options_internal): Update to set
18675         PARAM_PREFETCH_MINIMUM_STRIDE.
18676         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18677         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18678         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18679         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18680         stride is constant and is below the minimum stride threshold.
18682 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18684         * config/arm/arm-cpus.in (mode26): Delete.
18685         (armv4): Delete mode26 reference.
18686         * config/arm/arm.c (arm_configure_build_target): Delete use of
18687         isa_bit_mode26.
18689 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18691         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
18692         New insn pattern.
18693         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
18694         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
18695         for non-SSE modes.
18696         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
18697         (floatunsdidf2): Ditto.
18699 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18701         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
18702         (fixuns_trunc<mode>si2_avx512f): Ditto.
18703         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
18704         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
18705         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
18707 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18709         PR rtl-optimization/79985
18710         * df-scan.c (df_insn_refs_collect): Remove special case for
18711         global registers and asm statements.
18713 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18715         * extend.texi (Global Register Variables): Rewrite the bullet list.
18716         Note that the register is available for allocation. Note that access
18717         via inline asm must use constraints. Add note about async-signal
18718         handlers. Remove paragraph about automagic register selection.
18720 2018-05-23  Richard Biener  <rguenther@suse.de>
18722         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
18723         of fixed offset from memset VN.
18725 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
18727         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
18728         first_interp field.
18729         (alloc_cand_and_find_basis): Initialize first_interp field.
18730         (slsr_process_mul): Modify first_interp field.
18731         (slsr_process_add): Likewise.
18732         (slsr_process_cast): Modify first_interp field for each new
18733         interpretation.
18734         (slsr_process_copy): Likewise.
18735         (dump_candidate): Dump first_interp field.
18736         (replace_mult_candidate): Process all interpretations, not just
18737         subsequent ones.
18738         (replace_rhs_if_not_dup): Likewise.
18739         (replace_one_candidate): Likewise.
18741 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
18743         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
18744         Add new boolean.
18745         (aarch64_needs_frame_chain): New function.
18746         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
18748 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
18750         PR target/84882
18751         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
18752         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
18753         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
18754         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
18755         as true for strict-align.
18756         (aarch64_can_inline_p): Perform checks even when callee has no
18757         attributes to check for strict alignment.
18758         * doc/extend.texi (AArch64 Function Attributes): Document
18759         no-strict-align.
18760         * doc/invoke.texi: (AArch64 Options): Likewise.
18762 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
18764         PR tree-optimization/85853
18765         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
18766         the handling of the root of the node to...
18767         (vect_slp_analyze_node_operations_1): ...this new function,
18768         and run the whole thing with the child nodes' def types
18769         set according to their SLP node's def type.
18771 2018-05-23  Richard Biener  <rguenther@suse.de>
18773         PR middle-end/85874
18774         * tree-data-ref.c (create_runtime_alias_checks): Defer
18775         and ignore overflow warnings.
18777 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
18779         PR tree-optimization/85822
18780         * tree-vrp.c (is_masked_range_test): Fix handling of negative
18781         constants.
18783 2018-05-23  Richard Biener  <rguenther@suse.de>
18785         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
18786         memset constants via native_interpret_expr.
18788 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
18790         PR target/85345
18791         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
18792         attribute.
18793         (cgraph_node::create_alias): Likewise.
18794         (cgraph_node::get_availability): Check ifunc_resolver instead
18795         of looking up ifunc attribute.
18796         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
18797         * varasm.c (do_assemble_alias): Likewise.
18798         (assemble_alias): Likewise.
18799         (default_binds_local_p_3): Likewise.
18800         * cgraph.h (cgraph_node): Add ifunc_resolver.
18801         (cgraph_node::only_called_directly_or_aliased_p): Return false
18802         for IFUNC resolver.
18803         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
18804         attribute.
18805         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
18806         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
18807         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
18808         instead of looking up ifunc attribute.
18810 2018-05-22  Luis Machado  <luis.machado@linaro.org>
18812         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
18814 2018-05-22  Martin Sebor  <msebor@redhat.com>
18816         PR middle-end/85359
18817         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
18818         only when expasion succeeds.
18819         (expand_builtin_strcmp): Same.
18820         (expand_builtin_strncmp): Same.
18822 2018-05-22  Martin Sebor  <msebor@redhat.com>
18824         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
18826 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18827             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18829         * config/aarch64/aarch64-ldpstp.md: Replace uses of
18830         aarch64_mem_pair_operand with memory_operand and delete operand swapping
18831         code.
18832         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18833         Add check for legitimate_address.
18834         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
18835         (aarch64_swap_ldrstr_operands): New.
18836         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
18837         Define prototype.
18839 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18840             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18842         * config/aarch64/aarch64.md: New patterns to generate stp
18843         and ldp.
18844         (store_pair_sw, store_pair_dw): New patterns to generate stp for
18845         single words and double words.
18846         (load_pair_sw, load_pair_dw): Likewise.
18847         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
18848         Delete.
18849         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
18850         Delete.
18851         * config/aarch64/aarch64-ldpstp.md: Modify peephole
18852         for different mode ldpstp and add peephole for merged zero stores.
18853         Likewise for loads.
18854         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18855         Add size check.
18856         (aarch64_gen_store_pair): Rename calls to match new patterns.
18857         (aarch64_gen_load_pair): Rename calls to match new patterns.
18858         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
18859         (load_pair<DREG:mode><DREG2:mode>): ... This.
18860         (store_pair<mode>): Rename to...
18861         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
18862         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
18863         New mode iterators.
18864         (V_INT_EQUIV): Handle SImode.
18865         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
18866         New predicate.
18868 2018-05-22  Martin Sebor  <msebor@redhat.com>
18870         PR c/85623
18871         * calls.c (maybe_warn_nonstring_arg): Use string length to set
18872         or ajust the presumed bound on an operation to avoid unnecessary
18873         warnings.
18875 2018-05-22  Martin Sebor  <msebor@redhat.com>
18877         PR tree-optimization/85826
18878         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
18879         assuming that a DECL necesarily has a constant size.
18881 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
18883         PR middle-end/85862
18884         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
18886 2018-05-22  Richard Biener  <rguenther@suse.de>
18888         PR tree-optimization/85834
18889         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
18890         non-constant and non-zero memset arguments.
18892 2018-05-22  Martin Liska  <mliska@suse.cz>
18894         PR ipa/85607
18895         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
18897 2018-05-22  Richard Biener  <rguenther@suse.de>
18899         PR tree-optimization/85863
18900         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
18901         comparisons when vectype is specified.
18902         (vectorizable_condition): Do not specify vectype for
18903         vect_is_simple_cond when SLP vectorizing.
18905 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
18907         PR target/85657
18908         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
18909         define __ibm128 as long double.
18910         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
18911         as a distinct type when IEEE 128-bit support is enabled.
18912         (init_float128_ieee): Fix up conversions between IFmode and IEEE
18913         128-bit types to use the correct functions.
18914         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
18915         convert between 128-bit floating point types that have different
18916         modes but the same representation, instead of using gen_lowpart to
18917         makean alias.
18918         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
18919         KFmode.
18920         (IFKF_reg): New attributes to give the register constraints for
18921         IFmode and KFmode.
18922         (extend<mode>tf2_internal): New insns to mark an explicit
18923         conversion between 128-bit floating point types that have a
18924         different mode but share the same representation.
18926 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
18928         PR tree-optimization/85814
18929         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
18930         a null return from get_strinfo when unsharing the next
18931         strinfo in the chain.
18933 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
18935         PR gcc/84923
18936         * varasm.c (weak_finish): Clean up weak_decls.
18938 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18940         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
18941         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
18942         UNSPEC_UADALP values.
18943         * config/aarch64/iterators.md (ABAL): New int iterator.
18944         (ABDL2): Likewise.
18945         (ADALP): Likewise.
18946         (sur): Add mappings for the above.
18947         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
18948         New define_insn.
18949         (aarch64_<sur>abal<mode>_4): Likewise.
18950         (aarch64_<sur>adalp<mode>_3): Likewise.
18951         (<sur>sadv16qi): New define_expand.
18953 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
18955         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
18956         (*movdf_internal): Ditto.
18957         (*rcpsf2_sse): Ditto.
18958         (*rsqrtsf2_sse): Ditto.
18959         (*sqrt<mode>2_sse): Ditto.
18961 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
18963         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
18964         eor3q<mode>4.
18965         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
18966         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
18967         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
18968         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
18969         vbcaxq_s64): New.
18970         * config/aarch64/arm_neon.h: Likewise.
18971         * config/aarch64/iterators.md (VQ_I): New.
18973 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
18975         * config.gcc: Add arc/t-multilib-linux to tmake_file for
18976         arc*-*-linux*.
18977         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
18978         MULTILIB_DIRNAMES
18980 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18982         * config/nds32/constraints.md (S): New constraint.
18983         * config/nds32/nds32.md (call_internal): Use constraint S.
18984         (call_value_internal): Likewise.
18985         (sibcall_internal): Likewise.
18986         (sibcall_value_internal): Likewise.
18988 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18989             Chung-Ju Wu  <jasonwucj@gmail.com>
18991         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
18992         into consideration.
18994 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18995             Chung-Ju Wu  <jasonwucj@gmail.com>
18997         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
18998         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
18999         (nds32_rtx_costs_impl): Simplify.
19000         (nds32_address_cost_impl): Simplify.
19001         (nds32_init_rtx_costs): New function.
19002         (nds32_rtx_costs_speed_prefer): Likewise.
19003         (nds32_rtx_costs_size_prefer): Likewise.
19004         (nds32_address_cost_speed_prefer): Likewise.
19005         (nds32_address_cost_speed_fwprop): Likewise.
19006         (nds32_address_cost_size_prefer): Likewise.
19007         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
19008         * config/nds32/nds32.c (nds32_option_override): Use
19009         nds32_init_rtx_costs function.
19011 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
19013         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
19014         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
19015         (TARGET_PIPELINE_N8): Likewise.
19016         (TARGET_PIPELINE_N10): Likewise.
19017         (TARGET_PIPELINE_N13): Likewise.
19018         (TARGET_PIPELINE_GRAYWOLF): Likewise.
19020 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
19022         * config/nds32/nds32-fpu.md: Update copyright year.
19024 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19026         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
19028 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19030         * config/nds32/nds32.c
19031         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
19032         * config/nds32/nds32.opt (minline-asm-r15): New option.
19034 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19036         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
19037         MASK_HW_ABS.
19038         * config/nds32/nds32.md (abssi2): New pattern.
19040 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
19042         * config/i386/i386.md (rex64namesuffix): New mode attribute.
19043         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
19044         Merge insn pattern from sse_cvtsi2ss<round_name> and
19045         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
19046         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
19047         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
19048         using SWI48 mode iterator.
19049         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
19050         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
19051         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
19052         pattern from sse_cvttss2si<round_saeonly_name>
19053         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
19054         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
19055         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
19056         using SWI48 mode iterator.
19057         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
19058         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
19059         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
19060         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
19061         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
19062         using SWI48 mode iterator.
19063         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
19064         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
19065         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
19066         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
19067         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
19068         SWI48 mode iterator.
19069         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
19070         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
19071         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
19072         pattern from sse_cvttsd2si<round_saeonly_name>
19073         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
19075 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19077         * config/nds32/nds32-md-auxiliary.c
19078         (nds32_valid_smw_lwm_base_p): Refine.
19079         (nds32_output_smw_single_word): Refine.
19080         (nds32_output_smw_double_word): New.
19081         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
19083 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19085         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
19086         (nds32_output_stack_pop): Refine.
19087         (nds32_expand_unaligned_load): Refine.
19088         (nds32_expand_unaligned_store): Refine.
19090 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
19091             Chung-Ju Wu  <jasonwucj@gmail.com>
19093         * config/nds32/constants.md: Add TP_REGNUM constant.
19094         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
19095         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
19096         UNSPEC_ADD32.
19097         * config/nds32/nds32-doubleword.md: Consider flag_pic.
19098         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
19099         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
19100         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
19101         and PIC code generation.
19102         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
19103         code generation.
19104         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
19105         optimization.
19106         * config/nds32/nds32.md: Support TLS and PIC.
19107         * config/nds32/nds32.c: Support TLS and PIC.
19108         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
19109         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
19110         predicate.
19112 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19114         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
19115         mode with E_ prefix.
19117 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
19118             Chung-Ju Wu  <jasonwucj@gmail.com>
19120         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
19121         * config/nds32/nds32-md-auxiliary.c
19122         (symbolic_reference_mentioned_p): New.
19123         (nds32_legitimize_ict_address): New.
19124         (nds32_expand_ict_move): New.
19125         (nds32_indirect_call_referenced_p): New.
19126         (nds32_symbol_binds_local_p): Delete.
19127         (nds32_long_call_p): Modify.
19128         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
19129         * config/nds32/nds32-protos.h
19130         (symbolic_reference_mentioned_p): Declare.
19131         (nds32_legitimize_ict_address): Declare.
19132         (nds32_expand_ict_move): Declare.
19133         (nds32_indirect_call_referenced_p): Declare.
19134         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
19135         (nds32_relax_group): Use nds32_ict_const_p as condition.
19136         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
19137         (nds32_asm_file_start): Output ict_model directive in asm code.
19138         (nds32_legitimate_address_p): Consider indirect call.
19139         (nds32_print_operand): Consider indirect call.
19140         (nds32_print_operand_address): Consider indirect call.
19141         (nds32_insert_attributes): Handle "indirect_call" attribute.
19142         (TARGET_LEGITIMATE_ADDRESS_P): Define.
19143         (TARGET_LEGITIMATE_CONSTANT_P): Define.
19144         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
19145         (TARGET_DELEGITIMIZE_ADDRESS): Define.
19146         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
19147         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
19148         (TARGET_ICT_MODEL_SMALL): Define.
19149         (TARGET_ICT_MODEL_LARGE): Define.
19150         * config/nds32/nds32.md (movsi): Consider ict model.
19151         (call, call_value): Consider ict model.
19152         (sibcall, sibcall_value): Consider ict model.
19153         * config/nds32/nds32.opt (mict-model): New option.
19154         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
19155         model.
19157 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
19158             Monk Chiang  <sh.chiang04@gmail.com>
19159             Jim Wilson <jimw@sifive.com>
19161         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
19162         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
19163         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
19164         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
19165         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
19166         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
19167         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
19168         compute save_libcall_adjustment properly.
19169         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
19170         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
19171         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
19172         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
19173         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
19174         (ABI_SPEC): Handle mabi=ilp32e.
19175         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
19176         (RVE): Add RVE mask.
19177         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
19178         <-march>: Add rv32e as an example.
19180 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
19182         PR c++/82899
19183         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
19184         (intra_create_variable_infos): Handle C++ constructors.
19186 2018-05-18  Martin Liska  <mliska@suse.cz>
19188         * passes.def: Remove a redundant pass.
19190 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
19192         PR bootstrap/85838
19193         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
19195 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19197         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
19198         (ARMv4): Update.
19199         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
19200         (ARMv6m): Update.
19201         (armv2, armv2a, armv3, armv3m): Delete architectures.
19202         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
19203         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
19204         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
19205         Delete cpus.
19206         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
19207         (*mulsidi3adddi): Likewise.
19208         (mulsidi3): Likewise.
19209         (*mulsidi3_nov6): Likewise.
19210         (umulsidi3): Likewise.
19211         (umulsidi3_nov6): Likewise.
19212         (umaddsidi4): Likewise.
19213         (*umulsidi3adddi): Likewise.
19214         (smulsi3_highpart): Likewise.
19215         (*smulsi3_highpart_nov6): Likewise.
19216         (umulsi3_highpart): Likewise.
19217         (*umulsi3_highpart_nov6): Likewise.
19218         * config/arm/arm.h (arm_arch3m): Delete.
19219         * config/arm/arm.c (arm_arch3m): Delete.
19220         (arm_option_override_internal): Update armv3-related comment.
19221         (arm_configure_build_target): Delete use of isa_bit_mode32.
19222         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
19223         (arm_rtx_costs_internal): Delete check of arm_arch3m.
19224         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
19225         (mulsa3): Likewise.
19226         (mulusa3): Likewise.
19227         * config/arm/arm-protos.h (arm_arch3m): Delete.
19228         * config/arm/arm-tables.opt: Regenerate.
19229         * config/arm/arm-tune.md: Likewise.
19230         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
19231         deleted architectures.
19233 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19235         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
19236         (armv5t, armv5te): New features.
19237         (ARMv5, ARMv5e): Delete fgroups.
19238         (ARMv5t, ARMv5te): Adjust for above changes.
19239         (ARMv6m): Likewise.
19240         (armv5, armv5e): Delete arches.
19241         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
19242         arm_arch5.
19243         (*call_reg_arm): Likewise.
19244         (*call_value_reg_armv5): Likewise.
19245         (*call_value_reg_arm): Likewise.
19246         (*call_symbol): Likewise.
19247         (*call_value_symbol): Likewise.
19248         (*sibcall_insn): Likewise.
19249         (*sibcall_value_insn): Likewise.
19250         (clzsi2): Likewise.
19251         (prefetch): Likewise.
19252         (define_split and define_peephole2 dependent on arm_arch5):
19253         Likewise.
19254         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
19255         arm_arch5e.
19256         (TARGET_ARM_QBIT): Likewise.
19257         (TARGET_DSP_MULTIPLY): Likewise.
19258         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
19259         (arm_arch5, arm_arch5e): Delete.
19260         (arm_arch5t, arm_arch5te): Declare.
19261         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
19262         (arm_arch5t): Declare.
19263         (arm_option_reconfigure_globals): Update for the above.
19264         (arm_options_perform_arch_sanity_checks): Update comment, replace
19265         use of arm_arch5 with arm_arch5t.
19266         (use_return_insn): Likewise.
19267         (arm_emit_call_insn): Likewise.
19268         (output_return_instruction): Likewise.
19269         (arm_final_prescan_insn): Likewise.
19270         (arm_coproc_builtin_available): Likewise.
19271         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
19272         arm_arch5e with arm_arch5t and arm_arch5te.
19273         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
19274         (arm_arch5t, arm_arch5te): Declare.
19275         * config/arm/arm-tables.opt: Regenerate.
19276         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
19277         * config/arm/t-multilib: Likewise.
19278         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
19279         instead of arm_arch5.
19280         (*call_reg_thumb1): Likewise.
19281         (*call_value_reg_thumb1_v5): Likewise.
19282         (*call_value_reg_thumb1): Likewise.
19283         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
19284         unreachable path.
19285         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
19287 2018-05-18  Martin Liska  <mliska@suse.cz>
19289         PR gcov-profile/84846
19290         * doc/gcov.texi: Document -t option of gcov tool.
19292 2018-05-18  Martin Liska  <mliska@suse.cz>
19294         PR gcov-profile/84846
19295         * gcov.c (print_usage): Add new -t option.
19296         (process_args): Handle the option.
19297         (generate_results): Use stdout as output when requested by
19298         the option.
19300 2018-05-18  Martin Liska  <mliska@suse.cz>
19302         PR gcov-profile/84846
19303         * coverage.c (coverage_init): Write PWD to .gcno file.
19304         * doc/gcov.texi: Document how working directory is printed.
19305         * gcov-dump.c (dump_gcov_file): Print PWD.
19306         * gcov.c (output_intermediate_file): Likewise.
19307         (read_graph_file): Read PWD string.
19308         (output_lines): Print PWD.
19310 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19312         PR middle-end/85817
19313         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
19314         for retval and return false if all args to phi are zero.
19316 2018-05-18  Richard Biener  <rguenther@suse.de>
19318         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
19319         method.
19320         (evrp_dom_walker::before_dom_children): Call it.
19322 2018-05-18  Richard Biener  <rguenther@suse.de>
19324         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
19325         results when processing array refs with variable index.
19327 2018-05-18  Toon Moene  <toon@moene.org>
19329         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
19330         directly after that of -floop-interchange. Indicate that both
19331         options are enabled by default when specifying -O3.
19333 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19335         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
19336         iterator.  Delete separate integer-mode vec_set<mode> expander.
19337         (aarch64_simd_vec_setv2di): Delete.
19338         (vec_setv2di): Delete.
19339         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
19340         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
19341         the "w, r" alternative.
19343 2018-05-18  Martin Liska  <mliska@suse.cz>
19345         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
19346         * tree-pass.h (make_pass_lower_switch_O0): New function.
19347         * tree-switch-conversion.c (node_has_low_bound): Remove.
19348         (node_has_high_bound): Likewise.
19349         (node_is_bounded): Likewise.
19350         (class pass_lower_switch): Make it a template type and create
19351         two instances.
19352         (pass_lower_switch::execute): Add template argument.
19353         (make_pass_lower_switch): New function.
19354         (make_pass_lower_switch_O0): New function.
19355         (do_jump_if_equal): Remove.
19356         (emit_case_nodes): Simplify to just handle all 3 cases and leave
19357         all the hard work to tree optimization passes.
19359 2018-05-18  Martin Liska  <mliska@suse.cz>
19361         * dbgcnt.c (limit_low): Renamed from limit.
19362         (limit_high): New variable.
19363         (dbg_cnt_is_enabled): Check for upper limit.
19364         (dbg_cnt): Adjust dumping.
19365         (dbg_cnt_set_limit_by_index): Add new argument for high
19366         value.
19367         (dbg_cnt_set_limit_by_name): Likewise.
19368         (dbg_cnt_process_single_pair): Parse new format.
19369         (dbg_cnt_process_opt): Use strtok.
19370         (dbg_cnt_list_all_counters): Remove 'value' and add
19371         'limit_high'.
19372         * doc/invoke.texi: Document changes.
19374 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
19376         * doc/sourcebuild.texi (scalar_all_fma): Document.
19377         * tree.def (FMA_EXPR): Delete.
19378         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
19379         * internal-fn.c (ternary_direct): New macro.
19380         (expand_ternary_optab_fn): Likewise.
19381         (direct_ternary_optab_supported_p): Likewise.
19382         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
19383         * builtins.c (fold_builtin_fma): Delete.
19384         (fold_builtin_3): Don't call it.
19385         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
19386         * expr.c (expand_expr_real_2): Likewise.
19387         * fold-const.c (operand_equal_p): Likewise.
19388         (fold_ternary_loc): Likewise.
19389         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
19390         * gimple.c (DEFTREECODE): Likewise.
19391         * gimplify.c (gimplify_expr): Likewise.
19392         * optabs-tree.c (optab_for_tree_code): Likewise.
19393         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19394         * tree-eh.c (operation_could_trap_p): Likewise.
19395         (stmt_could_throw_1_p): Likewise.
19396         * tree-inline.c (estimate_operator_cost): Likewise.
19397         * tree-pretty-print.c (dump_generic_node): Likewise.
19398         (op_code_prio): Likewise.
19399         * tree-ssa-loop-im.c (stmt_cost): Likewise.
19400         * tree-ssa-operands.c (get_expr_operands): Likewise.
19401         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
19402         * fold-const-call.h (fold_fma): Delete.
19403         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
19404         CFN_FNMA and CFN_FNMS.
19405         (fold_fma): Delete.
19406         * genmatch.c (combined_fn): New enum.
19407         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
19408         (commutative_op): New function.
19409         (commutate): Use it.  Handle more than 2 operands.
19410         (dt_operand::gen_gimple_expr): Use commutative_op.
19411         (parser::parse_expr): Allow :c to be used with non-binary
19412         operators if the commutative operand is known.
19413         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
19414         CFN_FMS, CFN_FNMA and CFN_FNMS.
19415         (backprop::process_assign_use): Remove FMA_EXPR handling.
19416         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
19417         (gen_hsa_fma): New function.
19418         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
19419         IFN_FNMA and IFN_FNMS.
19420         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
19421         * gimple-fold.h (follow_all_ssa_edges): Declare.
19422         * gimple-fold.c (follow_all_ssa_edges): New function.
19423         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
19424         gimple_build interface and use follow_all_ssa_edges to fold the result.
19425         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
19426         instead of checking for optabs directly.
19427         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
19428         rather than FMA_EXPRs.
19429         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
19430         call to IFN_FMA instead of an FMA_EXPR.
19432 2018-05-17  Jim Wilson  <jimw@sifive.com>
19434         * expr.c (do_tablejump): When converting index to Pmode, if we have a
19435         sign extended promoted subreg, and the range does not have the sign bit
19436         set, then do a sign extend.
19438         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
19439         test, check for sign extended subreg and/or constant operands, and
19440         do a sign extend in that case.
19442 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
19444         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
19445         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
19446         Add untyped.
19447         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
19448         Change logics_shift_reg to logics_shift_imm.
19449         (thunderx2t99_fp_loadpair_basic): Delete.
19450         (thunderx2t99_fp_storepair_basic): Delete.
19451         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
19452         (thunderx2t99_asimd_polynomial): Delete.
19453         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
19454         and neon_fp_mul_d_scalar_q.
19455         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
19456         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
19457         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
19458         (thunderx2t99_asimd_lut): Add missing tbl types.
19459         (thunderx2t99_asimd_ext): Delete.
19460         (thunderx2t99_asimd_load1_1_mult): Delete.
19461         (thunderx2t99_asimd_load1_2_mult): Delete.
19462         (thunderx2t99_asimd_load1_ldp): New.
19463         (thunderx2t99_asimd_load1): New.
19464         (thunderx2t99_asimd_load2): Add missing *load2* types.
19465         (thunderx2t99_asimd_load3): New.
19466         (thunderx2t99_asimd_load4): New.
19467         (thunderx2t99_asimd_store1_1_mult): Delete.
19468         (thunderx2t99_asimd_store1_2_mult): Delete.
19469         (thunderx2t99_asimd_store2_mult): Delete.
19470         (thunderx2t99_asimd_store2_onelane): Delete.
19471         (thunderx2t99_asimd_store_stp): New.
19472         (thunderx2t99_asimd_store1): New.
19473         (thunderx2t99_asimd_store2): New.
19474         (thunderx2t99_asimd_store3): New.
19475         (thunderx2t99_asimd_store4): New.
19477 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
19479         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
19480         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
19482 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
19483             Segher Boessenkool  <segher@kernel.crashing.org>
19485         PR target/85698
19486         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
19487         operand.
19489 2018-05-17  Richard Biener  <rguenther@suse.de>
19491         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
19492         for pruning loop and prune defs feeding only already visited PHIs.
19494 2018-05-17  Richard Biener  <rguenther@suse.de>
19496         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
19498 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
19499             Richard Biener  <rguenther@suse.de>
19501         PR tree-optimization/85793
19502         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
19503         for VMAT_ELEMENTWISE.
19505 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19507         * internal-fn.h (lookup_internal_fn): Declare
19508         * internal-fn.c (lookup_internal_fn): New function.
19509         * gimple.c (gimple_build_call_from_tree): Handle calls to
19510         internal functions.
19511         * gimple-pretty-print.c (dump_gimple_call): Print "." before
19512         internal function names.
19513         * tree-pretty-print.c (dump_generic_node): Likewise.
19514         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
19516 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19518         * gimple-fold.h (gimple_build): Make the function forms take
19519         combined_fn rather than built_in_function.
19520         (gimple_simplify): Likewise.
19521         * gimple-match-head.c (gimple_simplify): Likewise.
19522         * gimple-fold.c (gimple_build): Likewise.
19523         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
19524         rather than gimple_build_call_internal.
19525         (get_initial_defs_for_reduction): Likewise.
19526         (vect_create_epilog_for_reduction): Likewise.
19527         (vectorizable_live_operation): Likewise.
19529 2018-05-17  Martin Liska  <mliska@suse.cz>
19531         * gimple-ssa-sprintf.c (format_directive): Do not use
19532         space in between 'G_' and '('.
19534 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
19536         PR target/85323
19537         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
19538         even if the mask is not all ones.
19540         PR target/85323
19541         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
19542         vector.
19543         (ix86_gimple_fold_builtin): Likewise.
19545         PR target/85323
19546         * config/i386/i386.c: Include tree-vector-builder.h.
19547         (ix86_vector_shift_count): New function.
19548         (ix86_fold_builtin): Fold shift builtins by scalar count.
19549         (ix86_gimple_fold_builtin): Likewise.
19551         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
19552         _mm512_setzero): New intrinsics.
19554 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
19555             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19557         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
19558         code generation for cases where splatting a value is not useful.
19559         * simplify-rtx.c (simplify_ternary_operation): Simplify
19560         vec_merge across a vec_duplicate and a paradoxical subreg forming
19561         a vector mode to a vec_concat.
19563 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
19565         * config.gcc: Support "goldmont-plus".
19566         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19567         "goldmont-plus".
19568         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19569         PROCESSOR_GOLDMONT_PLUS.
19570         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
19571         (processor_target_table): Add "goldmont-plus".
19572         (PTA_GOLDMONT_PLUS): Define.
19573         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
19574         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
19575         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
19576         (fold_builtin_cpu): Add "goldmont-plus".
19577         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
19578         (ix86_option_override_internal): Add "goldmont-plus".
19579         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
19580         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
19581         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
19582         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
19584 2018-05-17  Richard Biener  <rguenther@suse.de>
19586         PR tree-optimization/85757
19587         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
19588         remove defs that only feed that PHI from further processing.
19590 2018-05-16  Jim Wilson  <jimw@sifive.com>
19592         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
19593         asterisk to name.
19594         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
19596 2018-05-16  Mark Wielaard  <mark@klomp.org>
19598         * dwarf2out.c (count_index_strings): New function.
19599         (output_indirect_strings): Call count_index_strings and generate
19600         header for dwarf_version >= 5.
19602 2018-05-16  Mark Wielaard  <mark@klomp.org>
19604         * dwarf2out.c (dwarf_FORM): New function.
19605         (set_indirect_string): Use dwarf_FORM.
19606         (reset_indirect_string): Likewise.
19607         (size_of_die): Likewise.
19608         (value_format): Likewise.
19609         (output_die): Likewise.
19610         (add_skeleton_AT_string): Likewise.
19611         (output_macinfo_op): Likewise.
19612         (index_string): Likewise.
19613         (output_index_string_offset): Likewise.
19614         (output_index_string): Likewise.
19615         (count_index_strings): Likewise.
19617 2018-05-16  Carl Love  <cel@us.ibm.com>
19619         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19620         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
19622 2018-05-16  Martin Jambor  <mjambor@suse.cz>
19624         * ipa-prop.c (ipa_free_all_edge_args): Remove.
19625         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
19627 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
19629         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
19630         (fnma<mode>4): Likewise.
19631         (fms<mode>4): Likewise.
19632         (fnms<mode>4): Likewise.
19633         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
19634         (aarch64_fnma<mode>4): Likewise.
19635         (aarch64_fms<mode>4): Likewise.
19636         (aarch64_fnms<mode>4): Likewise.
19637         (aarch64_fnmadd<mode>4): Likewise.
19639 2018-05-16  Jason Merrill  <jason@redhat.com>
19641         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
19643 2018-05-16  Richard Biener  <rguenther@suse.de>
19645         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
19646         (dump_stmt_cost): Declare.
19647         (add_stmt_cost): Dump cost we add.
19648         (add_stmt_costs): New function.
19649         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
19650         No longer exported.
19651         (vect_analyze_stmt): Adjust prototype.
19652         (vectorizable_condition): Likewise.
19653         (vectorizable_live_operation): Likewise.
19654         (vectorizable_reduction): Likewise.
19655         (vectorizable_induction): Likewise.
19656         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
19657         cost vector to pass to vectorizable_ and record afterwards.
19658         (vect_model_reduction_cost): Take cost vector argument and adjust.
19659         (vect_model_induction_cost): Likewise.
19660         (vectorizable_reduction): Likewise.
19661         (vectorizable_induction): Likewise.
19662         (vectorizable_live_operation): Likewise.
19663         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
19664         SLP_TREE_NUMBER_OF_VEC_STMTS.
19665         (vect_analyze_slp_cost_1): Remove.
19666         (vect_analyze_slp_cost): Likewise.
19667         (vect_slp_analyze_node_operations): Take visited args and
19668         a target cost vector.  Avoid processing already visited stmt sets.
19669         (vect_slp_analyze_operations): Use a local cost vector to gather
19670         costs and register those of non-discarded instances.
19671         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
19672         (vect_schedule_slp_instance): Remove copying of
19673         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
19674         zero.
19675         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
19676         adding cost.  Record cost entry location.
19677         (vect_prologue_cost_for_slp_op): Function to compute cost of
19678         a constant or invariant generated for SLP vect in the prologue,
19679         split out from vect_analyze_slp_cost_1.
19680         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
19681         (vect_model_promotion_demotion_cost): Likewise.
19682         (vect_model_store_cost): Likewise, make static.
19683         (vect_model_load_cost): Likewise.
19684         (vectorizable_bswap): Add cost vector arg and adjust.
19685         (vectorizable_call): Likewise.
19686         (vectorizable_simd_clone_call): Likewise.
19687         (vectorizable_conversion): Likewise.
19688         (vectorizable_assignment): Likewise.
19689         (vectorizable_shift): Likewise.
19690         (vectorizable_operation): Likewise.
19691         (vectorizable_store): Likewise.
19692         (vectorizable_load): Likewise.
19693         (vectorizable_condition): Likewise.
19694         (vectorizable_comparison): Likewise.
19695         (can_vectorize_live_stmts): Likewise.
19696         (vect_analyze_stmt): Likewise.
19697         (vect_transform_stmt): Adjust calls to vectorizable_*.
19698         * tree-vectorizer.c: Include gimple-pretty-print.h.
19699         (dump_stmt_cost): New function.
19701 2018-05-16  Richard Biener  <rguenther@suse.de>
19703         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
19704         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
19705         * tree-ssa-dse.c: Include tree-ssa-loop.h.
19706         (check_name): New callback.
19707         (dse_classify_store): Track cycles via a visited bitmap of PHI
19708         defs and simplify handling of in-loop and across loop dead stores
19709         and properly fail for loop-variant refs.  Handle byte-tracking with
19710         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
19711         limiting the walk.
19713 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
19715         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
19716         (vect_get_mask_type_for_stmt): Likewise.
19717         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
19718         split out from...
19719         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
19720         to determine the statement's vector type and the vector type that
19721         should be used for calculating nunits.  Deal with cases in which
19722         the type has to be deferred.
19723         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
19724         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
19725         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
19726         (vect_determine_vf_for_stmt): New functions, split out from...
19727         (vect_determine_vectorization_factor): ...here.
19728         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
19729         (vect_get_mask_type_for_stmt): New functions, split out from
19730         vect_determine_vectorization_factor.
19732 2018-05-16  Richard Biener  <rguenther@suse.de>
19734         * tree-cfg.c (verify_gimple_assign_ternary): Properly
19735         verify the [VEC_]COND_EXPR embedded comparison.
19737 2018-05-15  Martin Sebor  <msebor@redhat.com>
19739         PR tree-optimization/85753
19740         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
19741         RECORD_TYPE in addition to ARRAY_TYPE.
19743 2018-05-15  Martin Sebor  <msebor@redhat.com>
19745         PR middle-end/85643
19746         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
19748 2018-05-15  Richard Biener  <rguenther@suse.de>
19750         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
19751         add by_clobber_p one.  Change algorithm to collect all defs
19752         representing uses we need to walk and try reducing them to
19753         a single one before failing.
19754         (dse_dom_walker::dse_optimize_stmt): Adjust.
19756 2018-05-13  Mark Wielaard  <mark@klomp.org>
19758         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
19759         (size_of_loc_descr): Likewise.
19760         (output_loc_operands): Likewise.
19761         (output_loc_operands_raw): Likewise.
19762         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
19763         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
19764         (hash_loc_operands): Likewise.
19765         (compare_loc_operands): Likewise.
19767 2018-05-14  Mark Wielaard  <mark@klomp.org>
19769         * dwarf2out.c (count_index_addrs): New function.
19770         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
19772 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19774         PR tree-optimization/83648
19775         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
19776         return value as malloc candidate.
19778 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19780         PR ipa/85734
19781         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
19782         param as true in call to suggest_attribute.
19784 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
19786         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
19787         -mreadonly-in-sdata.
19789 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19791         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
19792         New pattern.
19793         (aarch64_crypto_aesd_fused): Likewise.
19795 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
19797         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
19798         (movsi_aarch64): Likewise.
19799         (load_pairsi): Likewise.
19800         (load_pairdi): Likewise.
19801         (store_pairsi): Likewise.
19802         (store_pairdi): Likewise.
19803         (load_pairsf): Likewise.
19804         (load_pairdf): Likewise.
19805         (store_pairsf): Likewise.
19806         (store_pairdf): Likewise.
19807         (zero_extend): Likewise.
19808         (trunc): Swap alternatives.
19809         (fcvt_target): Add '?' to prefer w over r.
19811 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
19813         PR target/85756
19814         * config/i386/i386.md: Disallow non-commutative arithmetics in
19815         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
19816         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
19817         in the peephole2 before it.
19819 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
19821         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
19822         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
19823         (ix86_handle_option): Handle -mcldemote.
19824         * config.gcc: New header.
19825         * config/i386/cldemoteintrin.h: New file.
19826         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
19827         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19828         -mcldemote.
19829         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19830         OPTION_MASK_ISA_CLDEMOTE.
19831         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
19832         (ix86_valid_target_attribute_inner_p): Ditto.
19833         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
19834         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
19835         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
19836         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
19837         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
19838         (cldemote): New.
19839         * config/i386/i386.opt: Add -mcldemote.
19840         * config/i386/x86intrin.h: New header.
19841         * doc/invoke.texi: Add -mcldemote.
19843 2018-05-14  Richard Biener  <rguenther@suse.de>
19845         * doc/match-and-simplify.texi: Adjust :s documentation.
19847 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
19849         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
19850         intended memcpy size.
19851         (REORDER_45): Likewise.
19853 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
19855         * sort.cc: New file.
19856         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
19857         * vec.c (qsort_chk): Use gcc_qsort.
19858         * Makefile.in (OBJS-libcommon): Add sort.o.
19859         (build/sort.o): New target.  Use it...
19860         (BUILD_RTL): ... here, and...
19861         (build/gencfn-macros): ... here, and...
19862         (build/genmatch): ... here.
19864 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19865             Chung-Ju Wu  <jasonwucj@gmail.com>
19867         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
19868         * config/nds32/nds32-graywolf.md: New file.
19869         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
19870         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
19871         pipeline.
19872         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
19873         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
19874         * config/nds32/nds32.md (pipeline_model): Add graywolf.
19875         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
19876         * config/nds32/pipelines.md: Include n15 settings.
19878 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19879             Chung-Ju Wu  <jasonwucj@gmail.com>
19881         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
19882         * config/nds32/nds32-n13.md: New file.
19883         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
19884         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
19885         pipeline.
19886         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
19887         * config/nds32/nds32.md (pipeline_model): Add n13.
19888         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
19889         * config/nds32/pipelines.md: Include n13 settings.
19891 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19892             Chung-Ju Wu  <jasonwucj@gmail.com>
19894         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
19895         * config/nds32/nds32-n10.md: New file.
19896         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
19897         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
19898         pipeline.
19899         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
19900         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
19901         * config/nds32/nds32.md (pipeline_model): Add n10.
19902         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
19903         * config/nds32/pipelines.md: Include n10 settings.
19905 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
19906             Kito Cheng  <kito.cheng@gmail.com>
19907             Chung-Ju Wu  <jasonwucj@gmail.com>
19909         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
19910         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19911         Add enum values for DSP extension instructions.
19912         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
19913         New constraints.
19914         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
19915         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
19916         New code iterators.
19917         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
19918         * config/nds32/nds32-dspext.md: New file for DSP implementation.
19919         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
19920         * config/nds32/nds32-intrinsic.md: Likewise.
19921         * config/nds32/nds32_intrinsic.h: Likewise.
19922         * config/nds32/nds32-md-auxiliary.c: Likewise.
19923         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
19924         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
19925         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
19926         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
19927         * config/nds32/nds32-protos.h: New declarations for DSP extension.
19928         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
19929         TYPE_DMAC in switch statement.
19930         * config/nds32/nds32.c: New checking and implementation for DSP
19931         extension instructions.
19932         * config/nds32/nds32.h: Likewise.
19933         * config/nds32/nds32.md: Likewise.
19934         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
19935         * config/nds32/predicates.md: Implement new predicates for DSP
19936         extension.
19938 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
19940         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
19941         Reformat alternatives and attributes so it is easier to identify
19942         which constraints/attributes go with which instruction.
19943         (mov<mode>_hardfloat32, FMOVE64): Likewise.
19944         (mov<mode>_softfloat32, FMOVE64): Likewise.
19945         (mov<mode>_hardfloat64, FMOVE64): Likewise.
19946         (mov<mode>_softfloat64, FMOVE64): Likewise.
19948 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19950         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19951         subsection.
19952         (Basic PowerPC Built-in Functions): The new name of the
19953         subsection previously known as "PowerPC Built-in Functions".
19954         (Basic PowerPC Built-in Functions Available on all Configurations):
19955         New subsubsection.
19956         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
19957         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
19958         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
19959         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
19961 2018-05-11  Martin Jambor  <mjambor@suse.cz>
19963         PR ipa/85655
19964         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
19965         single const.
19967 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19969         PR target/85733
19970         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
19972 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
19974         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
19975         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
19976         (ix86_handle_option): Handle -mwaitpkg.
19977         * config.gcc: New header.
19978         * config/i386/cpuid.h (bit_WAITPKG): New bit.
19979         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
19980         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
19981         function type.
19982         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19983         OPTION_MASK_ISA_WAITPKG.
19984         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
19985         (ix86_option_override_internal): Add PTA_WAITPKG.
19986         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
19987         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
19988         IX86_BUILTIN_TPAUSE.
19989         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
19990         __builtin_ia32_umwait and __builtin_ia32_tpause.
19991         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
19992         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
19993         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
19994         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
19995         UNSPECV_TPAUSE): New.
19996         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
19997         * config/i386/i386.opt: Add -mwaitpkg.
19998         * config/i386/waitpkgintrin.h: New file.
19999         * config/i386/x86intrin.h: New header.
20000         * doc/invoke.texi: Add -mwaitpkg.
20002 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
20004         PR target/85606
20005         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
20006         equivalent.
20007         (cortex-m0): Use armv6s-m isa.
20008         (cortex-m0plus): Likewise.
20009         (cortex-m1): Likewise.
20010         (cortex-m0.small-multiply): Likewise.
20011         (cortex-m0plus.small-multiply): Likewise.
20012         (cortex-m1.small-multiply): Likewise.
20014 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
20015             Jakub Jelinek  <jakub@redhat.com>
20017         PR tree-optimization/85692
20018         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
20019         source permute as well.
20021 2018-05-11  Martin Liska  <mliska@suse.cz>
20023         PR sanitizer/85556
20024         * doc/extend.texi: Document LLVM style format for no_sanitize
20025         attribute.
20027 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
20029         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
20030         mode_supports_vsx_dform_quad to mode_supports_dq_form.
20031         (mode_supports_vsx_dform_quad): Likewise.
20032         (mode_supports_vmx_dform): Move these functions to be next to the
20033         other mode_supports functions.
20034         (mode_supports_dq_form): Likewise.
20035         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
20036         mode_supports_dq_form.
20037         (reg_offset_addressing_ok_p): Likewise.
20038         (offsettable_ok_by_alignment): Likewise.
20039         (rs6000_legitimate_offset_address_p): Likewise.
20040         (legitimate_lo_sum_address_p): Likewise.
20041         (rs6000_legitimize_address): Likewise.
20042         (rs6000_legitimize_reload_address): Likewise.
20043         (rs6000_secondary_reload_inner): Likewise.
20044         (rs6000_preferred_reload_class): Likewise.
20045         (rs6000_output_move_128bit): Likewise.
20047 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
20049         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
20050         Generate SImode target register for null target.
20051         <case IX86_BUILTIN_XGETBV>: Ditto.
20052         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
20053         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
20055 2018-05-10  Carl Love  <cel@us.ibm.com>
20057         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
20058         dcbtt and dcbtstt if operands[2] is 0.
20060 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
20062         PR target/85693
20063         * config/i386/sse.md (usadv64qi): New expander.
20065 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
20067         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
20068         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
20069         -maltivec=be support.
20070         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
20071         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
20072         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
20073         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
20074         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
20075         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
20076         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
20077         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
20078         altivec_vsumsws): Adjust.
20079         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
20080         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
20081         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
20082         support.
20083         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
20084         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
20085         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
20086         (altivec_lve<VI_char>x): Delete expand.
20087         (*altivec_lve<VI_char>x_internal): Rename to...
20088         (altivec_lve<VI_char>x): ... this.
20089         (altivec_lvxl_<mode>): Delete expand.
20090         (*altivec_lvxl_<mode>_internal): Rename to ...
20091         (altivec_lvxl_<mode>): ... this.
20092         (altivec_stvxl_<mode>): Delete expand.
20093         (*altivec_stvxl_<mode>_internal): Rename to ...
20094         (altivec_stvxl_<mode>): ... this.
20095         (altivec_stve<VI_char>x): Delete expand.
20096         (*altivec_stve<VI_char>x_internal): Rename to ...
20097         (altivec_stve<VI_char>x): ... this.
20098         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
20099         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
20100         reduc_plus_scal_<mode>): Adjust.
20101         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
20102         comment.
20103         (rs6000_cpu_cpp_builtins): Adjust.
20104         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
20105         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
20106         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
20107         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
20108         -maltivec=be support.
20109         (rs6000_split_vec_extract_var): Adjust.
20110         (rs6000_split_v4si_init): Adjust.
20111         (swap_selector_for_mode): Delete.
20112         (altivec_expand_lvx_be, altivec_expand_stvx_be,
20113         altivec_expand_stvex_be): Delete.
20114         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
20115         -maltivec=be support.
20116         (rs6000_gimple_fold_builtin): Ditto.
20117         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
20118         Adjust.
20119         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
20120         (TARGET_DIRECT_MOVE_64BIT): Adjust.
20121         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
20122         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
20123         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
20124         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
20125         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
20126         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
20127         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
20128         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
20129         anonymous split): Adjust.
20130         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
20131         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
20133 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
20135         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
20136         when --with-gxx-include-dir is also specified.
20137         * configure: Regenerate.
20139 2018-05-09  Jim Wilson  <jimw@sifive.com>
20141         PR target/84797
20142         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
20143         * config/riscv/t-withmultilib: New.
20144         * config/riscv/withmultilib.h: New.
20145         * doc/install.texi: Document RISC-V --with-multilib-list support.
20147 2018-05-09  Richard Biener  <rguenther@suse.de>
20149         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
20150         vector.
20151         (vect_bb_vectorization_profitable_p): Adjust.  Compute
20152         actual scalar cost using the cost vector and the add_stmt_cost
20153         machinery.
20155 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20157         PR rtl-optimization/85645
20158         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
20159         in the REG_CFA_REGISTER note for LR, don't leave it empty.
20161 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20163         PR rtl-optimization/85645
20164         * shrink-wrap.c (spread_components): Return a boolean saying if
20165         anything was changed.
20166         (try_shrink_wrapping_separate): Iterate spread_components until
20167         nothing changes anymore.
20169 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20171         PR rtl-optimization/85645
20172         * regrename.c (build_def_use): Also kill the chains that include the
20173         destination of a REG_CFA_REGISTER note.
20175 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20177         PR rtl-optimization/85645
20178         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
20179         insn that has a REG_CFA_REGISTER note.
20181 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
20183         * cfgexpand.c (expand_clobber): New function.
20184         (expand_gimple_stmt_1): Use it.
20185         * tree-vect-stmts.c (vect_clobber_variable): New function,
20186         split out from...
20187         (vectorizable_simd_clone_call): ...here.
20188         (vectorizable_store): Emit a clobber either side of an
20189         IFN_STORE_LANES sequence.
20190         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
20192 2018-05-09  Tom de Vries  <tom@codesourcery.com>
20194         PR target/85626
20195         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
20196         (define_insn "trap_if_false"): Add exit after trap.
20198 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
20200         PR rtl-optimization/85638
20201         * bb-reorder.c: Include common/common-target.h.
20202         (create_forwarder_block): New function extracted from...
20203         (fix_up_crossing_landing_pad): ...here.  Rename into...
20204         (dw2_fix_up_crossing_landing_pad): ...this.
20205         (sjlj_fix_up_crossing_landing_pad): New function.
20206         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
20207         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
20208         from both partitions and exit the loop after one iteration.
20210 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20212         Revert:
20213         * doc/extend.texi (PowerPC Built-in Functions): Rename this
20214         subsection.
20215         (Basic PowerPC Built-in Functions): The new name of the
20216         subsection previously known as "PowerPC Built-in Functions".
20217         (Basic PowerPC Built-in Functions Available on all Configurations):
20218         New subsubsection.
20219         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20220         subsubsection.
20221         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20222         subsubsection.
20223         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20224         subsubsection.
20225         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20226         subsubsection.
20228 2018-05-08  Jim Wilson  <jimw@sifive.com>
20230         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
20231         (LD_EMUL_SUFFIX): New.
20232         (LINK_SPEC): Use it.
20234 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20236         * doc/extend.texi (PowerPC Built-in Functions): Rename this
20237         subsection.
20238         (Basic PowerPC Built-in Functions): The new name of the
20239         subsection previously known as "PowerPC Built-in Functions".
20240         (Basic PowerPC Built-in Functions Available on all Configurations):
20241         New subsubsection.
20242         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20243         subsubsection.
20244         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20245         subsubsection.
20246         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20247         subsubsection.
20248         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20249         subsubsection.
20251 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
20253         PR target/85683
20254         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
20255         after cmpelim optimization.
20257 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
20259         * config.gcc: Support "goldmont".
20260         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
20261         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20262         PROCESSOR_GOLDMONT.
20263         * config/i386/i386.c (m_GOLDMONT): Define.
20264         (processor_target_table): Add "goldmont".
20265         (PTA_GOLDMONT): Define.
20266         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
20267         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
20268         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
20269         (fold_builtin_cpu): Add "goldmont".
20270         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
20271         (ix86_option_override_internal): Add "goldmont".
20272         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
20273         (processor_type): Add PROCESSOR_GOLDMONT.
20274         * config/i386/i386.md: Add CPU "glm".
20275         * config/i386/glm.md: New file.
20276         * config/i386/x86-tune.def: Add m_GOLDMONT.
20277         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
20279 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
20281         PR target/85572
20282         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
20283         E_V4DImode.
20284         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
20285         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
20286         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
20288         PR target/85317
20289         * config/i386/i386.c (ix86_fold_builtin): Handle
20290         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
20292         PR target/85480
20293         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
20294         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
20296 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
20298         PR target/85658
20299         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
20300         (check_arch): Likewise.
20301         (check_fpu): Return the result rather than printing it.
20302         (end arch): Fix operator precedence.
20303         (end cpu): Likewise.
20304         (END): Print the result from check_fpu.
20306 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20307             Alan Hayward  <alan.hayward@arm.com>
20308             David Sherwood  <david.sherwood@arm.com>
20310         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
20311         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
20312         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
20313         (*fcmuo<mode>_and): New patterns.
20315 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20317         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
20318         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
20319         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
20320         (cmp_op, sve_imm_con): New code attributes.
20321         (SVE_COND_INT_CMP, imm_con): Delete.
20322         (cmp_op): Remove above unspecs from int attribute.
20323         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
20324         to...
20325         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
20326         comparison-specific unspecs.
20327         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
20328         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
20329         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
20330         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
20331         (*vec_fcm<cmp_op><mode>): Rename to...
20332         (*fcm<cmp_op><mode>): ...this and adjust likewise.
20333         (*vec_fcmuo<mode>): Rename to...
20334         (*fcmuo<mode>): ...this and adjust likewise.
20335         (*pred_fcm<cmp_op><mode>): New pattern.
20336         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
20337         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
20338         functions.
20339         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
20340         and UNORDERED.
20341         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
20342         (aarch64_emit_sve_predicated_cond): New function.
20343         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
20344         (aarch64_emit_unspec_cond_or): Replace with...
20345         (aarch64_emit_sve_or_conds): ...this new function.  Use
20346         aarch64_emit_sve_ptrue_op for the individual comparisons and
20347         aarch64_emit_binop to OR them together.
20348         (aarch64_emit_inverted_unspec_cond): Replace with...
20349         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
20350         aarch64_emit_sve_ptrue_op for the comparison and
20351         aarch64_emit_unop to invert the result.
20352         (aarch64_expand_sve_vec_cmp_float): Update after the above
20353         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
20355 2018-05-07  Nathan Sidwell  <nathan@acm.org>
20357         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
20358         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
20359         (Backwards Compatibility): Likewise.
20361 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20363         PR bootstrap/85681
20364         Revert:
20365         2018-05-07  Luis Machado  <luis.machado@linaro.org>
20367         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20368         <prefetch_dynamic_strides>: New const bool field.
20369         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20370         prefetch_dynamic_strides.
20371         (exynosm1_prefetch_tune): Likewise.
20372         (thunderxt88_prefetch_tune): Likewise.
20373         (thunderx_prefetch_tune): Likewise.
20374         (thunderx2t99_prefetch_tune): Likewise.
20375         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20376         to false.
20377         (aarch64_override_options_internal): Update to set
20378         PARAM_PREFETCH_DYNAMIC_STRIDES.
20379         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20380         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20381         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20382         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20383         prefetch-dynamic-strides setting.
20385         2018-05-07  Luis Machado  <luis.machado@linaro.org>
20387         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20388         <minimum_stride>: New const int field.
20389         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20390         minimum_stride field.
20391         (exynosm1_prefetch_tune): Likewise.
20392         (thunderxt88_prefetch_tune): Likewise.
20393         (thunderx_prefetch_tune): Likewise.
20394         (thunderx2t99_prefetch_tune): Likewise.
20395         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20396         (aarch64_override_options_internal): Update to set
20397         PARAM_PREFETCH_MINIMUM_STRIDE.
20398         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20399         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20400         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20401         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20402         stride is constant and is below the minimum stride threshold.
20404 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20406         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
20407         to 512.
20409 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20411         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20412         <prefetch_dynamic_strides>: New const bool field.
20413         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20414         prefetch_dynamic_strides.
20415         (exynosm1_prefetch_tune): Likewise.
20416         (thunderxt88_prefetch_tune): Likewise.
20417         (thunderx_prefetch_tune): Likewise.
20418         (thunderx2t99_prefetch_tune): Likewise.
20419         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20420         to false.
20421         (aarch64_override_options_internal): Update to set
20422         PARAM_PREFETCH_DYNAMIC_STRIDES.
20423         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20424         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20425         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20426         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20427         prefetch-dynamic-strides setting.
20429 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20431         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20432         <minimum_stride>: New const int field.
20433         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20434         minimum_stride field.
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 minimum_stride to 2048.
20440         (aarch64_override_options_internal): Update to set
20441         PARAM_PREFETCH_MINIMUM_STRIDE.
20442         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20443         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20444         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20445         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20446         stride is constant and is below the minimum stride threshold.
20448 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
20450         PR c++/85659
20451         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
20452         the type is addressable.  Don't force op into register if it has
20453         BLKmode.
20455 2018-05-05  Roland McGrath  <mcgrathr@google.com>
20457         PR other/77609
20458         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
20459         any section for which we don't know a specific type it should have,
20460         regardless of name.  Previously this was done only for the exact
20461         names ".init_array", ".fini_array", and ".preinit_array".
20462         (default_elf_asm_named_section): Add comment about
20463         relationship with default_section_type_flags and SECTION_NOTYPE.
20464         (get_section): Don't consider it a type conflict if one side has
20465         SECTION_NOTYPE and the other doesn't, as long as neither has the
20466         SECTION_BSS et al used in the default_section_type_flags logic.
20468 2018-05-05  Tom de Vries  <tom@codesourcery.com>
20470         PR target/85653
20471         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
20472         (workaround_barsyncs): New function.
20473         (nvptx_reorg): Use workaround_barsyncs.
20474         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
20475         (define_expand "nvptx_membar_cta"): New define_expand.
20476         (define_insn "*nvptx_membar_cta"): New insn.
20478 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
20480         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
20481         To improve optimization opportunities.
20482         * builtin-types.def: The new needed builtin types for the above.
20484 2018-05-04  Richard Biener  <rguenther@suse.de>
20486         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
20487         * gimple-ssa-store-merging.c
20488         (imm_store_chain_info::output_merged_store): Remove redundant create,
20489         release split_store vector contents on failure.
20490         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
20491         scalar stmt vector on cache hit.
20493 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
20495         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
20496         Xilinx FP support.
20497         * config.gcc (powerpc-xilinx-eabi*): Remove.
20498         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
20499         support.
20500         (fusion_addis_mem_combo_load): Ditto.
20501         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
20502         FP support.
20503         (rs6000_cpu_cpp_builtins): Ditto.
20504         * config/rs6000/rs6000-linux.c
20505         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
20506         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
20507         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
20508         support.
20509         (rs6000_setup_reg_addr_masks): Ditto.
20510         (rs6000_init_hard_regno_mode_ok): Ditto.
20511         (rs6000_option_override_internal): Ditto.
20512         (legitimate_lo_sum_address_p): Ditto.
20513         (rs6000_legitimize_address): Ditto.
20514         (rs6000_legitimize_reload_address): Ditto.
20515         (rs6000_legitimate_address_p): Ditto.
20516         (abi_v4_pass_in_fpr): Ditto.
20517         (setup_incoming_varargs): Ditto.
20518         (rs6000_gimplify_va_arg): Ditto.
20519         (rs6000_split_multireg_move): Ditto.
20520         (rs6000_savres_strategy): Ditto.
20521         (rs6000_emit_prologue_components): Ditto.
20522         (rs6000_emit_epilogue_components): Ditto.
20523         (rs6000_emit_prologue): Ditto.
20524         (rs6000_emit_epilogue): Ditto.
20525         (rs6000_elf_file_end): Ditto.
20526         (rs6000_function_value): Ditto.
20527         (rs6000_libcall_value): Ditto.
20528         * config/rs6000/rs6000.h: Ditto.
20529         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
20530         (TARGET_MINMAX): ... this.  New.
20531         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
20532         * config/rs6000/rs6000.md: Remove Xilinx FP support.
20533         (*movsi_internal1_single): Delete.
20534         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
20535         mfpu=, mxilinx-fpu): Delete.
20536         * config/rs6000/singlefp.h: Delete.
20537         * config/rs6000/sysv4.h: Remove Xilinx FP support.
20538         * config/rs6000/t-rs6000: Ditto.
20539         * config/rs6000/t-xilinx: Delete.
20540         * config/rs6000/titan.md: Adjust for fp_type removal.
20541         * config/rs6000/vsx.md: Remove Xilinx FP support.
20542         (VStype_simple): Delete.
20543         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
20544         * config/rs6000/xfpu.h: Delete.
20545         * config/rs6000/xfpu.md: Delete.
20546         * config/rs6000/xilinx.h: Delete.
20547         * config/rs6000/xilinx.opt: Delete.
20548         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
20549         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
20551 2018-05-04  Tom de Vries  <tom@codesourcery.com>
20553         PR libgomp/85639
20554         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
20555         if ignore == 0.
20557 2018-05-04  Richard Biener  <rguenther@suse.de>
20559         PR middle-end/85627
20560         * tree-complex.c (update_complex_assignment): We are always in SSA form.
20561         (expand_complex_div_wide): Likewise.
20562         (expand_complex_operations_1): Likewise.
20563         (expand_complex_libcall): Preserve EH info of the original stmt.
20564         (tree_lower_complex): Handle removed blocks.
20565         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
20566         on complex multiplication and division libcall builtins.
20568 2018-05-04  Richard Biener  <rguenther@suse.de>
20570         PR middle-end/85574
20571         * fold-const.c (negate_expr_p): Restrict negation of operand
20572         zero of a division to when we know that can happen without
20573         overflow.
20574         (fold_negate_expr_1): Likewise.
20576 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
20578         PR libstdc++/85466
20579         * real.h (real_nextafter): Declare.
20580         * real.c (real_nextafter): New function.
20581         * fold-const-call.c (fold_const_nextafter): New function.
20582         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
20583         CASE_CFN_NEXTTOWARD.
20584         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
20585         even when arg1_mode is different from arg0_mode.
20587 2018-05-03  Nathan Sidwell  <nathan@acm.org>
20589         * doc/extend.texi (Deprecated Features): Remove
20590         -ffriend-injection.
20591         (Backwards Compatibility): Likewise.
20592         * doc/invoke.texi (C++ Language Options): Likewise.
20593         (C++ Dialect Options): Likewise.
20595 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
20597         PR target/85530
20598         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
20599         _mm512_mask_mullox_epi64): New intrinsics.
20601 2018-05-03  Tom de Vries  <tom@codesourcery.com>
20603         PR testsuite/85106
20604         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20605         dump files): Add offload-tree.
20607 2018-05-03  Richard Biener  <rguenther@suse.de>
20609         PR tree-optimization/85615
20610         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
20611         to loops not nested in BBs loop father to avoid creating multi-entry
20612         loops.
20614 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20616         PR tree-optimization/70291
20617         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
20618         arguments.  Change return type to tree.  Emit libcall as a new
20619         statement rather than replacing existing one when inplace_p is true.
20620         (expand_complex_multiplication_components): New function.
20621         (expand_complex_multiplication): Expand floating-point complex
20622         multiplication using the above.
20623         (expand_complex_division): Rename inner_type parameter to type.
20624         Update expand_complex_libcall call-site.
20625         (expand_complex_operations_1): Update expand_complex_multiplication
20626         and expand_complex_division call-sites.
20628 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20630         PR target/85582
20631         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20632         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20633         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
20634         the highest significant bit of the shift count mask is clear.  In
20635         check whether and[sq]i3 is needed verify that all significant bits
20636         of the shift count other than the highest are set.
20638 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20640         PR libgomp/82428
20641         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
20642         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
20643         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
20644         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
20645         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
20646         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
20647         __builtin_goacc_parlevel_size.
20649 2018-05-02  Richard Biener  <rguenther@suse.de>
20651         PR tree-optimization/85597
20652         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
20653         do not use split vect_get_vec_defs call but call vect_get_slp_defs
20654         directly.
20656 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20658         PR testsuite/85106
20659         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20660         dump files): Add ltrans-tree.
20662 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20664         PR testsuite/85106
20665         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20666         dump files): Add wpa-ipa.
20668 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
20670         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
20671         powerpc*-*-linux*paired* target.
20672         * config/rs6000/750cl.h: Delete.
20673         * config/rs6000/paired.h: Delete.
20674         * config/rs6000/paired.md: Delete.
20675         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
20676         float support.
20677         * config/rs6000/rs6000-builtin.def: Remove paired float support.
20678         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20679         comment.  Remove paired float support.
20680         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
20681         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
20682         VECTOR_PAIRED.
20683         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
20684         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
20685         declarations.
20686         * config/rs6000/rs6000.c: Remove paired float support.
20687         (paired_expand_vector_init, paired_expand_vector_move,
20688         paired_emit_vector_compare, paired_emit_vector_cond_expr,
20689         (paired_expand_lv_builtin, paired_expand_stv_builtin,
20690         paired_expand_builtin, paired_expand_predicate_builtin,
20691         paired_init_builtins): Delete.
20692         * config/rs6000/rs6000.h: Remove paired float support.
20693         * config/rs6000/rs6000.md: Remove paired float support.
20694         (move_from_CR_ov_bit): Delete.
20695         * config/rs6000/rs6000.opt (mpaired): Delete.
20696         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
20697         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
20699 2018-05-02  Richard Biener  <rguenther@suse.de>
20701         PR middle-end/85567
20702         * gimplify.c (gimplify_save_expr): When in SSA form allow
20703         SAVE_EXPRs to compute to SSA vars.
20705 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20707         PR target/85582
20708         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20709         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20710         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
20711         clobber operands[2], instead use a new pseudo.  Formatting fixes.
20713 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
20715         PR tree-optimization/85586
20716         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
20717         exit early for statements in the same group if the accesses are
20718         not strided.
20720 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20722         PR lto/85451
20723         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
20724         error message.
20726 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
20728         PR tree-optimization/85143
20729         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
20731 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20733         PR lto/85451
20734         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
20735         not found" error message.
20737 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20739         PR other/83786
20740         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
20741         * vec.c (test_ordered_remove_if): New function.
20742         (vec_c_tests): Call test_ordered_remove_if.
20743         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
20744         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
20745         * tree-vect-patterns.c (vect_pattern_recog_1): Use
20746         VEC_ORDERED_REMOVE_IF.
20748 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20750         PR tree-optimization/82665
20751         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
20752         pointer subtraction where arguments come from a memchr call.
20754 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
20756         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
20757         --push-state --as-needed and --pop-state instead of --as-needed and
20758         --no-as-needed if ld supports it.
20759         * configure: Regenerated.
20761         PR web/85578
20762         * doc/install.texi2html: Replace _002d with - and _002a with * in
20763         generated html files using sed.
20765 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20767         PR c++/85523
20768         * gcc-rich-location.c (blank_line_before_p): New function.
20769         (use_new_line): New function.
20770         (gcc_rich_location::add_fixit_insert_formatted): New function.
20771         * gcc-rich-location.h
20772         (gcc_rich_location::add_fixit_insert_formatted): New function.
20774 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20776         * selftest.c (assert_streq): Rename "expected" and "actual" to
20777         "val1" and "val2".  Extend NULL-handling to cover both inputs
20778         symmetrically, while still requiring both to be non-NULL for a pass.
20779         * selftest.h (assert_streq): Rename "expected" and "actual" to
20780         "val1" and "val2".
20781         (ASSERT_EQ): Likewise.
20782         (ASSERT_EQ_AT): Likewise.
20783         (ASSERT_KNOWN_EQ): Likewise.
20784         (ASSERT_KNOWN_EQ_AT): Likewise.
20785         (ASSERT_NE): Likewise.
20786         (ASSERT_MAYBE_NE): Likewise.
20787         (ASSERT_MAYBE_NE_AT): Likewise.
20788         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
20789         the assertion to pass.
20790         (ASSERT_STREQ_AT): Likewise.
20792 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20794         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
20795         interaction with -pie.
20797 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20799         * selftest.h: Fix alphabetization of per-source-file selftest
20800         declarations.
20802 2018-04-30  Jason Merrill  <jason@redhat.com>
20804         PR c++/61982 - dead stores to destroyed objects.
20805         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
20806         of clobber.
20808 2018-04-30  Jason Merrill  <jason@redhat.com>
20810         * tree.c (build_clobber): New.
20811         * tree.h: Declare it.
20812         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
20814 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20816         * diagnostic-show-locus.c (layout::layout): Update for
20817         location_get_source_line returning a char_span.
20818         (struct char_span): Move to input.h.
20819         (struct correction): Update for fields in char_span becoming
20820         private.
20821         (struct source_line): Update for location_get_source_line
20822         returning a char_span.
20823         (layout::print_line): Likewise.
20824         * edit-context.c (edited_file::print_content): Likewise.
20825         (edited_file::print_diff_hunk): Likewise.
20826         (edited_file::print_run_of_changed_lines): Likewise.
20827         (edited_file::get_num_lines): Likewise.
20828         (edited_line::edited_line): Likewise.
20829         * final.c (asm_show_source): Likewise.
20830         * input.c (location_get_source_line): Convert return type
20831         from const char * to char_span, losing the final "line_len"
20832         param.
20833         (dump_location_info): Update for the above.
20834         (get_substring_ranges_for_loc): Likewise.  Use a char_span
20835         when handling the literal within the line.
20836         (test_reading_source_line): Update for location_get_source_line
20837         returning a char_span.
20838         * input.h (class char_span): Move here from
20839         diagnostic-show-locus.c, converting from a struct to a class.
20840         Make data members private.
20841         (char_span::operator bool): New.
20842         (char_span::length): New.
20843         (char_span::get_buffer): New.
20844         (char_span::operator[]): New.
20845         (char_span::subspan): Make const.
20846         (char_span::xstrdup): New.
20847         (location_get_source_line): Convert return type from const char *
20848         to char_span, losing the final "line_size" param.
20850 2018-04-30  Jan Hubicka  <jh@suse.cz>
20852         * lto-wrapper.c (ltrans_priorities): New static var.
20853         (cmp_priority): New.
20854         (run_gcc): Read priorities and if doing parallel build order
20855         the Makefile by them.
20857 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20859         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
20861 2018-04-30  Richard Biener  <rguenther@suse.de>
20863         * tree-cfg.c (verify_address): Remove base argument, add
20864         flag whether to check TREE_ADDRESSABLE and do that.
20865         (verify_expr): Remove.
20866         (verify_types_in_gimple_reference): Add pieces from verify_expr.
20867         (verify_gimple_assign_single): Likewise.
20868         (verify_gimple_switch): Likewise.
20869         (verify_expr_location_1): Dereference tp once.  Add (disabled)
20870         piece from verify_expr.
20871         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
20873 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20875         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
20877 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20879         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
20880         (small_data_pattern): Likewise.
20881         (arc_rewrite_small_data): Likewise.
20882         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
20883         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
20884         (get_symbol_alignment): New function.
20885         (legitimate_small_data_address_p): Likewise.
20886         (legitimate_scaled_address): Update, call
20887         legitimate_small_data_address_p.
20888         (output_sdata): New static variable.
20889         (arc_print_operand): Update how we handle small data operands.
20890         (arc_print_operand_address): Likewise.
20891         (arc_legitimate_address_p): Update, use
20892         legitimate_small_data_address_p.
20893         (arc_rewrite_small_data_p): Remove.
20894         (arc_rewrite_small_data_1): Likewise.
20895         (arc_rewrite_small_data): Likewise.
20896         (small_data_pattern): Likewise.
20897         (compact_sda_memory_operand): Update to use
20898         legitimate_small_data_address_p and get_symbol_alignment.
20899         (prepare_move_operands): Don't rewite sdata pattern.
20900         (prepare_extend_operands): Remove.
20901         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
20902         pattern.
20903         (zero_extendqisi2): Likewise.
20904         (zero_extendhisi2): Likewise.
20905         (extendqihi2): Likewise.
20906         (extendqisi2): Likewise.
20907         (extendhisi2): Likewise.
20908         (addsi3): Likewise.
20909         (subsi3): Likewise.
20910         (andsi3): Likewise.
20911         * config/arc/constraints.md (Usd): Change it to memory constraint.
20913 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20915         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
20916         as source of std instructions.
20917         * config/arc/arc.md (movsi_insn): Update pattern predicate to
20918         allow 6-bit constants as source for store instructions.
20919         (movdi_insn): Update instruction pattern to allow 6-bit constants
20920         as source for store instructions.
20922 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20924         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
20926 2018-04-30  Nathan Sidwell  <nathan@acm.org>
20927             Sandra Loosemore <sandra@codesourcery.com>
20929         * dumpfile.c (dump_open): Allow '-' for stdout.
20930         * doc/invoke.texi (Developer Options): Document dump filename
20931         determination early.  Document stdin/stdout selection.
20933 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
20935         Microblaze Target: PIC data text relative
20937         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
20938         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
20939         Add declaration.
20940         * config/microblaze/microblaze.h (microblaze_constant_address_p):
20941         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
20942         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
20943         New addressing mode for data-text relative position indepenedent code.
20944         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
20945         'ADDRESS_SYMBOLIC_TXT_REL'.
20946         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
20947         (microblaze_legitimate_pic_operand): Exclude function calls from
20948         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
20949         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
20950         addresses cases.
20951         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20952         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20953         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
20954         for 'address + offset'.
20955         (microblaze_expand_prologue): Add new function prologue call for
20956         'r20' assignation.
20957         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
20958         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
20959         table in case of TARGET_PIC_DATA_TEXT_REL.
20960         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
20961         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
20962         Add new macros 'UNSPEC_TEXT',
20963         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
20964         + exclude function calls from 'UNSPEC_PLT' in case of data text
20965         relative mode.
20966         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20967         new target hook for generating address diff vector tables in case of
20968         flag_pic.
20969         * doc/tm.texi : Regenerate.
20970         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
20971         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
20972         of addr diff vector generation.
20973         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20974         target hook definition.
20975         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
20976         Add default function for generate_pic_addr_diff_vec -> flag_pic.
20977         * doc/invoke.texi (Add new pic option): Add new microblaze pic
20978         option for data text relative.
20980 2018-04-30  Richard Biener  <rguenther@suse.de>
20982         * tree-chrec.h (evolution_function_is_constant_p): Remove
20983         redundant check.
20984         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
20986 2018-04-30  Richard Biener  <rguenther@suse.de>
20988         PR bootstrap/85571
20989         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
20991 2018-04-30  Richard Biener  <rguenther@suse.de>
20993         PR tree-optimization/28364
20994         PR tree-optimization/85275
20995         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
20996         copying first exit test.
20998 2018-04-28  Mark Wielaard  <mark@klomp.org>
21000         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
21001         dwarf_version >= 5.
21002         (dwarf_AT): Handle DW_AT_addr_base.
21003         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
21005 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
21007         PR target/84431
21008         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
21009         (*ashl<dwi>3_doubleword_mask_1): Ditto.
21010         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
21011         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
21013 2018-04-28  Richard Biener  <rguenther@suse.de>
21015         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
21016         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
21017         to reflect use.  Only add interesting stmts.
21019 2018-04-27  Martin Jambor  <mjambor@suse.cz>
21021         PR ipa/85549
21022         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
21023         the jump function allows for passing through aggregate values.
21025 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
21027         * input.h (in_system_header_at): Convert from macro to inline
21028         function.
21029         (from_macro_expansion_at): Likewise.
21030         (from_macro_definition_at): Likewise.
21032 2018-04-27  Jeff Law  <law@redhat.com>
21034         * config.gcc: Mark tile* targets as deprecated/obsolete.
21036 2018-04-27  Richard Biener  <rguenther@suse.de>
21038         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
21039         fix for ILP32.
21041 2018-04-27  Richard Biener  <rguenther@suse.de>
21043         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
21045 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
21047         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
21048         with Yd constraint. Set "preferred_for_speed" attribute from
21049         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
21050         with Yd constraint.
21051         (*movdi_internal): Ditto.
21052         (movti_interunit splitters): Remove
21053         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
21054         (movdi_interunit splitters): Ditto.
21055         * config/i386/constraints.md (Ye): Remove.
21056         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
21058 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21060         PR target/85512
21061         * config/aarch64/constraints.md (Usg): Limit to 31.
21062         (Usj): Limit to 63.
21064 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
21066         PR tree-optimization/85529
21067         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
21068         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
21069         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
21070         zero extension or masking of the MSB bit.
21071         (optimize_range_tests): Add FIRST_BB argument, pass it through
21072         to optimize_range_tests_var_bound.
21073         (maybe_optimize_range_tests, reassociate_bb): Adjust
21074         optimize_range_tests callers.
21076 2018-04-26  Richard Biener  <rguenther@suse.de>
21077             Jakub Jelinek  <jakub@redhat.com>
21079         * cgraph.h (symbol_table): Just declare debug method here.
21080         * symtab.c (symbol_table::debug): Define.
21082 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
21084         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
21086 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
21088         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
21089         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
21090         (*movdi_internal): Substitute Yi and Yj constraint with x
21091         and Ym and Yn constraint with y constraint.  Update "isa"
21092         attribute and set "preferred_for_speed" attribute from
21093         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
21094         (*movsi_internal): Ditto.
21095         (*movdf_internal): Ditto.
21096         (*movsf_internal): Ditto.
21097         (*zero_extendsidi2): Ditto.
21098         * config/i386/sse.md (vec_set<mode>_0): Ditto.
21099         (sse2_loadld): Ditto.
21100         (*vec_extract<ssevecmodelower>_0): Ditto.
21101         (*vec_extractv4si_0_zext_sse4): Ditto.
21102         (vec_concatv2di): Ditto.
21103         (*vec_dup<mode>): Ditto.
21104         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
21105         * config/i386/constraints.md (Yi): Remove.
21106         (Yj): Remove.
21107         (Ym): Remove.
21108         (Yn): Remove.
21110 2018-04-26  Nathan Sidwell  <nathan@acm.org>
21112         * dumpfile.c (dump_open): New.
21113         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
21114         (dump_finish): Detect stdio/stderr by value not name.
21116 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
21118         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
21120 2018-04-26  Tom de Vries  <tom@codesourcery.com>
21122         PR target/84952
21123         * config/nvptx/nvptx.c (verify_neutering_jumps)
21124         (verify_neutering_labels): New function
21125         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
21127 2018-04-26  Tom de Vries  <tom@codesourcery.com>
21129         PR target/84025
21130         * config/nvptx/nvptx.c (needs_neutering_p): New function.
21131         (nvptx_single): Use needs_neutering_p to skip over insns that do not
21132         need neutering.
21134 2018-04-26  Richard Biener <rguenther@suse.de>
21135             Tom de Vries  <tom@codesourcery.com>
21137         PR lto/85422
21138         * lto-streamer-out.c (output_function): Fixup loops if required to match
21139         discovery done in the reader.
21141 2018-04-26  Richard Biener  <rguenther@suse.de>
21143         PR tree-optimization/85116
21144         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
21145         have a loop exit from the single latch predecessor.  Remove
21146         case of header with just condition.
21147         (ch_base::copy_headers): Exclude infinite loops from any
21148         processing.
21149         (pass_ch::execute): Record exits.
21151 2018-04-26  Richard Biener  <rguenther@suse.de>
21153         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
21154         prologue cost vector and pass it to vect_get_load_cost.
21155         (vect_get_peeling_costs_all_drs): Likewise.
21156         (vect_peeling_hash_get_lowest_cost): Likewise.
21157         (vect_enhance_data_refs_alignment): Likewise.
21159 2018-04-26  Richard Biener  <rguenther@suse.de>
21161         PR middle-end/85450
21162         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
21163         checking of integer<->pointer conversions.
21164         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
21165         sign-/zero-extending pointer types.
21166         (expand_omp_for_static_chunk): Likewise.
21168 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
21169             Jean Lee  <xiaoyur347@gmail.com>
21171         * config/mips/mips.c (mips_asan_shadow_offset): New function.
21172         (TARGET_ASAN_SHADOW_OFFSET): Define.
21173         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
21174         true for -fsanitize=address.
21176 2018-04-25  Mark Wielaard  <mark@klomp.org>
21178         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
21179         shorter ones.
21181 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
21183         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
21184         than "alu", remove explicit "memory" and "imm_disp" attributes.
21185         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
21187         PR middle-end/85414
21188         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
21189         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
21190         gen_lowpart_no_emit.
21192 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
21194         PR target/85473
21195         * config/i386/i386.c (ix86_expand_builtin): Change memory
21196         operand to XI, extend p0 to Pmode.
21197         * config/i386/i386.md: Change unspec volatile and operand
21198         1 mode to XI, change operand 0 mode to P.
21200 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21202         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
21203         GET_MODE_MASK before any checking.
21204         (nds32_can_use_bset_p): Likewise.
21205         (nds32_can_use_btgl_p): Likewise.
21207 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21209         * config/nds32/nds32-doubleword.md: New define_split pattern for
21210         illegal register number.
21212 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21214         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
21216 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21218         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
21220 2018-04-25  Richard Biener  <rguenther@suse.de>
21222         * lto-streamer.h (LTO_major_version): Bump to 8.
21224 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
21226         * BASE-VER: Set to 9.0.0.
21228 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
21230         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
21231         in __abskf2 and __powikf2.
21233 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21235         PR target/85512
21236         * config/aarch64/constraints.md (Usg, Usj): New constraints.
21237         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
21238         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
21239         Use the above on operand 2.  Reindent.
21240         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
21242 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
21244         PR target/85485
21245         * common/config/i386/i386-common.c (ix86_handle_option): Don't
21246         handle OPT_mcet.
21247         * config/i386/i386.opt (mcet): Removed.
21248         * doc/install.texi: Remove -mcet documentation.
21249         * doc/invoke.texi: Likewise.
21251 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
21253         PR target/85485
21254         * doc/install.texi: Remove -mcet from bootstrap-cet.
21256 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21258         PR target/85511
21259         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
21260         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
21261         if TARGET_64BIT.
21263         PR target/85503
21264         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
21265         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
21266         containing a CONST_VECTOR.
21268 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
21270         * doc/install.texi: Update newlib dependency for nvptx.
21272 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21274         PR target/85508
21275         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
21276         instead of INTVAL when shifting x left.
21278 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
21280         PR tree-optimization/85478
21281         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
21282         vect_grouped_store_supported for single element vectors.
21284 2018-04-24  Richard Biener  <rguenther@suse.de>
21286         PR target/85491
21287         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
21288         load cost increase to the case of non-constant step.
21290 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21292         PR target/84828
21293         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
21294         destination if any_malformed_asm.
21296 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
21298         PR middle-end/85496
21299         * expr.c (store_field): In the bitfield case, if the value comes from
21300         a function call and is returned in registers by means of a PARALLEL,
21301         do not change the mode of the temporary unless BLKmode and VOIDmode.
21303 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
21305         PR rtl-optimization/85423
21306         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
21307         dependencies to debug insns when the previous insn is non-debug.
21309 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21311         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
21312         enums into a single definition.
21313         (fls): Fix predicates and printing.
21314         (seti): Likewise.
21316 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21318         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
21319         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
21320         and short u6 immediate.
21321         (check_if_valid_sleep_operand): Remove.
21322         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
21324 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21326         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
21327         flag_always_save_lp condition.
21328         * config/nds32/nds32.opt (malways-save-lp): New option.
21330 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21332         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
21333         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
21334         * config/nds32/nds32.h
21335         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
21336         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
21338 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21340         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
21341         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
21343 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21344             Chung-Ju Wu  <jasonwucj@gmail.com>
21346         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
21347         Declare.
21348         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
21349         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
21351 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21353         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
21355 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21357         * config/nds32/nds32-protos.h (nds32_data_alignment,
21358         nds32_local_alignment): Declare.
21359         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
21360         nds32_local_alignment): New functions.
21361         (TARGET_CONSTANT_ALIGNMENT): Define.
21362         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
21364 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21366         * config/nds32/nds32.c
21367         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
21368         (TARGET_MODES_TIEABLE_P): Likewise.
21370 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21372         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
21373         level Ofast and Og.
21375 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
21376             Chung-Ju Wu  <jasonwucj@gmail.com>
21378         * config/nds32/constants.md (unspec_volatile_element): Add enum values
21379         for unaligned access.
21380         * config/nds32/nds32-intrinsic.c: Implementation of expanding
21381         unaligned access.
21382         * config/nds32/nds32-intrinsic.md: Likewise.
21383         * config/nds32/nds32_intrinsic.h: Likewise.
21384         * config/nds32/nds32.h (nds32_builtins): Likewise.
21385         * config/nds32/nds32.opt (munaligned-access): New option.
21386         * config/nds32/nds32.c (nds32_asm_file_start): Display
21387         flag_unaligned_access status.
21389 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
21391         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
21392         -mno-relax is present.
21393         * config/riscv/linux.h (LINK_SPEC): Ditto.
21395 2018-04-20  Martin Sebor  <msebor@redhat.com>
21397         PR c/85365
21398         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
21399         for null pointers.
21400         (gimple_fold_builtin_stxcpy_chk): Same.
21401         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
21403 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
21405         PR target/85456
21406         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
21407         __powikf2 when long double is IEEE 128-bit.
21409 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
21411         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
21412         step to make sure stack always aligned.
21414 2018-04-20  Carl Love  <cel@us.ibm.com>
21416         PR target/83402
21417         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
21418         size check for arg0.
21420 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
21421             Tom de Vries  <tom@codesourcery.com>
21423         PR target/85445
21424         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
21425         Emit insns for calls too.
21426         (nvptx_find_par): Always look for worker-level predecessor insn.
21427         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
21428         calls.
21429         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
21430         (nvptx_process_pars): Propagate frames for calls.
21432 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
21434         PR target/85469
21435         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
21436         Removed.
21437         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
21438         (ix86_handle_option): Don't handle OPT_mibt.
21439         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
21440         __SHSTK__.
21441         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
21442         has_ibt and ibt.
21443         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
21444         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
21445         (ix86_target_macros): Define __CET__ with flag_cf_protection
21446         for -fcf-protection.
21447         * config/i386/i386.c (isa2_opts): Remove -mibt.
21448         * config/i386/i386.h (TARGET_IBT): Removed.
21449         (TARGET_IBT_P): Likewise.
21450         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
21451         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
21452         * config/i386/i386.opt (mcet): Update help message.
21453         (mshstk): Likewise.
21454         (mibt): Removed.
21455         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
21456         -mcet as an alias for -mshstk.
21458 2018-04-20  Richard Biener <rguenther@suse.de>
21460         PR middle-end/85475
21461         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
21462         complexity by forcing a single use of the multiply operand.
21464 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21466         ipa/85449
21467         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
21468         recursion dependency to only apply to non-clones.
21470 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21472         ipa/85447
21473         * ipa-cp.c (create_specialized_node): Check that clones of
21474         self-recursive edges exist during IPA-CP.
21476 2018-04-19  Toon Moene  <toon@moene.org>
21478         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
21479         by -O3.
21481 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21483         PR tree-optimization/85467
21484         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
21485         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
21486         VECTOR_CST element to type.
21488 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21490         PR target/85397
21491         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
21492         * config/i386/i386.md (builtin_setjmp_setup): Removed.
21493         (builtin_longjmp): Likewise.
21494         (save_stack_nonlocal): New pattern.
21495         (restore_stack_nonlocal): Likewise.
21497 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21499         PR target/85404
21500         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21501         Replace ASM_OUTPUT_LABEL with fprintf.
21503 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21505         PR target/85417
21506         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21507         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
21508         * config/i386/i386-c.c (ix86_target_macros_internal): Also
21509         define __IBT__ and __SHSTK__ for -fcf-protection.
21510         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
21511         TARGET_IBT.
21512         (ix86_trampoline_init): Likewise.
21513         (x86_output_mi_thunk): Likewise.
21514         (ix86_notrack_prefixed_insn_p): Likewise.
21515         (ix86_option_override_internal): Don't disallow -fcf-protection.
21516         * config/i386/i386.md (rdssp<mode>): Also enable for
21517         -fcf-protection.
21518         (incssp<mode>): Likewise.
21519         (nop_endbr): Likewise.
21520         * config/i386/i386.opt (mcet): Change help message to built-in
21521         functions only.
21522         (mibt): Likewise.
21523         (mshstk): Likewise.
21524         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
21525         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
21526         enable CET built-in functions.
21528 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
21530         * common/config/i386/i386-common.c
21531         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
21532         OPTION_MASK_ISA_MOVDIRI_UNSET,
21533         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
21534         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
21535         * config.gcc (movdirintrin.h): New header.
21536         * config/i386/cpuid.h (bit_MOVDIRI,
21537         bit_MOVDIR64B): New bits.
21538         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
21539         and -mmvodir64b.
21540         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
21541         (VOID, PVOID, PCVOID)): New function types.
21542         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
21543         __builtin_ia32_directstoreu_u64,
21544         __builtin_ia32_movdir64b): New builtins.
21545         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
21546         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
21547         and -mmovdiri.
21548         (ix86_valid_target_attribute_inner_p): Ditto.
21549         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
21550         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
21551         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
21552         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
21553         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
21554         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
21555         (movdiri<mode>, movdir64b_<mode>): New.
21556         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
21557         * config/i386/immintrin.h: Include movdirintrin.h.
21558         * config/i386/movdirintrin.h: New file.
21559         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
21561 2018-04-19  Richard Biener  <rguenther@suse.de>
21563         PR middle-end/85455
21564         * cfg.c (clear_bb_flags): When loop state says we have
21565         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
21567 2018-04-19  Richard Biener  <rguenther@suse.de>
21569         PR tree-optimization/84737
21570         * tree-vect-data-refs.c (vect_copy_ref_info): New function
21571         copying restrict info.
21572         (vect_setup_realignment): Use it.
21573         * tree-vectorizer.h (vect_copy_ref_info): Declare.
21574         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
21575         the first DR to all generated stores.
21576         (vectorizable_load): Likewise for loads.
21578 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21580         PR tree-optimization/85446
21581         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
21582         the integral and pointer types to have the same precision.
21584         * doc/install.texi: Document --disable-cet being the default and
21585         --enable-cet=auto.
21587 2018-04-18  Martin Liska  <mliska@suse.cz>
21589         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
21590         style.
21592 2018-04-18  Martin Liska  <mliska@suse.cz>
21594         Revert
21595         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
21597         PR ipa/83983
21598         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21599         arguments if they are comparable.
21601 2018-04-18  Martin Liska  <mliska@suse.cz>
21603         Revert
21604         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21606         PR lto/84805
21607         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21608         incomplete types.
21610 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
21612         PR target/85388
21613         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
21614         ENDBR after calling __morestack.
21616 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
21618         PR jit/85384
21619         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
21620         by using gcc_base_ver to generate a gcc_driver_version, and use
21621         it when generating GCC_DRIVER_NAME.
21622         * configure: Regenerate.
21624 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
21626         PR target/81084
21627         * config.gcc: Obsolete powerpc*-*-*spe*.
21629 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21631         PR debug/84637
21632         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
21633         (stabstr_D): Change type of unum from unsigned int to
21634         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
21635         type.
21637 2018-04-17  Jim Wilson  <jimw@sifive.com>
21639         PR 84856
21640         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
21641         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
21642         Set arg_pointer_offset after using pretend_args_size.
21644 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21646         PR rtl-optimization/85431
21647         * dse.c (record_store): Ignore zero width stores.
21649         PR sanitizer/85230
21650         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
21651         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
21652         __builtin_stack_restore rather than after it.
21653         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
21654         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
21655         argument instead of virtual_dynamic_stack_rtx.
21657 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21659         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
21660         New prototype.
21661         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21662         Add note to error message to explain internal mapping of overloaded
21663         built-in function name to non-overloaded built-in function name.
21664         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
21665         function.
21667 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
21669         PR target/85424
21670         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
21671         where the inputs overlap with the output.
21673 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21675         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
21676         (=v, v) alternative and explicit "memory" attribute.
21677         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
21678         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21679         attributes.
21680         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
21681         "sselog1" type instead of "sselog".
21682         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
21683         "sselog".  Remove explicit "memory" attribute.
21684         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
21685         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21686         attributes.
21687         (vec_extract_hi_v32hi): Merge all alternatives into one, use
21688         "sselog1" type instead of "sselog".  Remove explicit "memory"
21689         attribute.
21690         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
21691         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21692         attribute.
21693         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
21694         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21695         attributes.
21696         (vec_extract_hi_v64qi): Merge all alternatives into one, use
21697         "sselog1" type instead of "sselog".  Remove explicit "memory"
21698         attribute.
21699         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
21700         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21701         attribute.
21703         PR target/85430
21704         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
21706         PR middle-end/85414
21707         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
21708         on a SUBREG.
21710 2018-04-17  Martin Jambor  <mjambor@suse.cz>
21712         PR ipa/85421
21713         * ipa-cp.c (create_specialized_node): Call
21714         expand_all_artificial_thunks if necessary.
21716 2018-04-17  Martin Liska  <mliska@suse.cz>
21718         PR lto/85405
21719         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
21720         in message, remote space in between '_G' and '('.
21722 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21724         PR target/85281
21725         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
21726         avx512f_vmcmp<mode>3<round_saeonly_name>,
21727         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
21728         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
21729         avx512f_rndscale<mode><round_saeonly_name>,
21730         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
21731         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
21732         Use %<iptr>2 instead of %2 for -masm=intel.
21733         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
21734         avx512f_vcvttss2usi<round_saeonly_name>,
21735         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
21736         -masm=intel.
21737         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
21738         avx512f_vcvttsd2usi<round_saeonly_name>,
21739         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
21740         Use %q1 instead of %1 for -masm=intel.
21741         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
21742         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
21743         of %3 for -masm=intel.
21744         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
21745         -masm=intel.
21746         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
21747         -masm=intel.
21748         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
21749         -masm=intel.
21750         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
21751         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
21752         %g1.
21753         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
21754         -masm=intel.
21755         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
21756         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
21757         %g1 and one with %0 and %1.
21758         (avx512er_vmrcp28<mode><round_saeonly_name>,
21759         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
21760         %1 for -masm=intel.
21761         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
21762         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
21763         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
21764         of %0 and %{%4%} for -masm=intel.
21765         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
21766         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
21767         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
21768         order of %0 and %{%5%}%{z%} for -masm=intel.
21770 2018-04-17  Jan Hubicka  <jh@suse.cz>
21772         PR lto/85405
21773         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
21775 2018-04-17  Martin Liska  <mliska@suse.cz>
21777         PR ipa/85329
21778         * multiple_target.c (create_dispatcher_calls): Set apostrophes
21779         for target_clone error message.  Make default implementation
21780         clone to be a local declaration.
21781         (separate_attrs): Add new argument and check for an empty
21782         string.
21783         (expand_target_clones): Handle it.
21784         (ipa_target_clone): Make redirection just for target_clones
21785         functions.
21787 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
21788             Tom de Vries  <tom@codesourcery.com>
21790         PR middle-end/84955
21791         * omp-expand.c (expand_oacc_for): Add dummy false branch for
21792         tiled basic blocks without omp continue statements.
21794 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
21796         PR target/83660
21797         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
21798         vec_extract expression as having side effects to make sure it gets
21799         a cleanup point.
21801 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
21803         PR target/85403
21804         * config/i386/i386.c (get_builtin_code_for_version): Check
21805         error_mark_node.
21807 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
21809         PR target/84331
21810         * config.gcc: Support "skylake".
21811         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21812         PROCESSOR_SKYLAKE.
21813         * config/i386/i386.c (m_SKYLAKE): Define.
21814         (processor_target_table): Add "skylake".
21815         (ix86_option_override_internal): Add "skylake".
21816         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
21817         PROCESSOR_CANNONLAKE.
21818         (get_builtin_code_for_version): Fix priority for
21819         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
21820         PROCESSOR_SKYLAKE-AVX512.
21821         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
21822         (processor_type): Add PROCESSOR_SKYLAKE.
21824 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
21825             Jason Merrill  <jason@redhat.com>
21827         PR c++/85112
21828         * convert.c (convert_to_integer_1): Use direct recursion for
21829         enumeral types and types with a precision less than the number
21830         of bits in their mode.
21832 2018-04-16  Julia Koval  <julia.koval@intel.com>
21834         PR target/84413
21835         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
21836         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
21838 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
21840         PR target/85293
21841         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
21842         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
21843         and -mno-direct-move.
21845 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
21847         PR target/83402
21848         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
21849         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
21850         Ensure negative shifts result in {0}.
21852 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
21854         PR rtl-optimization/79916
21855         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
21856         regs (if any) to define how to gnerate SD moves when LRA is in
21857         progress.
21859 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
21861         PR rtl-optimization/85393
21862         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
21863         * except.c (expand_dw2_landing_pad_for_region): Make static.
21864         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
21865         a label and unconditional jump to old_bb, rather than
21866         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
21867         basic block.
21869         PR rtl-optimization/85376
21870         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
21871         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
21872         instead of a specific value.
21874 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21875             Bin Cheng  <bin.cheng@arm.com>
21877         PR tree-optimization/82965
21878         PR tree-optimization/83991
21879         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
21880         by_profile_only parameter.
21881         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
21882         information if the loop was predicted to iterate too many times.
21883         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
21885 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21887         PR lto/71991
21888         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
21889         always inline.
21891 2018-04-13  Martin Liska  <mliska@suse.cz>
21892             Jakub Jelinek  <jakub@redhat.com>
21894         PR middle-end/81657
21895         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
21896         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
21897         * builtins.c (expand_builtin_memory_copy_args): Use
21898         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
21899         handle dest_addr == pc_rtx.
21901 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
21903         PR target/85291
21904         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
21905         asked to not generate direct moves.
21906         (fix_trunc<mode>si2_stfiwx): Similar.
21907         (fix_trunc<mode>si2_internal): Similar.
21909 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21911         PR debug/83157
21912         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
21913         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
21914         lookup if dest in some wider mode is known to be const0_rtx and
21915         if so, record permanent equivalence for it to be ZERO_EXTEND of
21916         the narrower mode destination.
21918 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21920         * lto-streamer-out.c (output_function): Revert 259346.
21921         * omp-expand.c (expand_oacc_for): Likewise.
21923 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
21925         PR rtl-optimization/85354
21926         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
21927         * sel-sched.c (sel_global_init): ... here.
21929 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
21931         PR target/85238
21932         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
21933         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
21934         mode for PE-COFF targets.
21935         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
21936         (i386_pe_asm_lto_end): Likewise.
21937         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
21938         (TARGET_ASM_LTO_END): Likewise.
21939         * config/i386/winnt.c (saved_debug_info_level): New static variable.
21940         (i386_pe_asm_lto_start): New function.
21941         (i386_pe_asm_lto_end): Likewise.
21943 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21944             Richard Biener  <rguenther@suse.de>
21946         PR middle-end/84955
21947         * lto-streamer-out.c (output_function): Fix CFG loop state before
21948         streaming out.
21949         * omp-expand.c (expand_oacc_for): Handle calls to internal
21950         functions like regular functions.
21952 2018-04-12  Richard Biener  <rguenther@suse.de>
21954         PR lto/85371
21955         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
21956         for the early LTO debug to properly generate references to it
21957         during DIE emission.  Do not re-use that for the skeleton for
21958         split-dwarf.
21959         (dwarf2out_early_finish): Likewise.
21961 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21963         PR target/85328
21964         * config/i386/sse.md
21965         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
21966         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
21967         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
21968         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
21969         and output is a reg, avoid creating invalid lowpart subreg, but
21970         instead split into a 512-bit move.  Don't split if not AVX512VL,
21971         input is xmm16+ reg and output is a mem.
21972         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
21973         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
21974         xmm16+ reg and output is a mem.
21976 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21978         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
21979         also for flag_dwarf2_cfi_asm.
21981 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21983         PR rtl-optimization/85342
21984         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
21985         a bool scalar var inside of the loop instead.  Don't try to update
21986         recog_data.operand after failed apply_change_group.
21988 2018-04-12  Tom de Vries  <tom@codesourcery.com>
21990         PR target/85296
21991         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
21992         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
21993         array with flexible array member as array without given dimension.
21994         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
21995         argument for undefined param to true.
21997 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
21999         PR target/85321
22000         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
22001         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
22002         from PowerPC section.
22003         * config/rs6000/sysv4.opt (mcall-): Improve help text.
22004         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
22005         help text that is too long.
22006         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
22007         help text that is too long.
22008         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
22009         help text that is too long.
22011 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
22013         * config/alpha/alpha.md (stack_probe_internal): Rename
22014         from "probe_stack".  Update all callers.
22016 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
22018         PR rtl-optimization/84566
22019         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
22020         sched_macro_fuse_insns.
22022 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
22024         PR target/84301
22025         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
22026         (compute_block_dependences): ... from here.
22028 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22030         PR tree-optimization/85331
22031         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
22032         from int to HOST_WIDE_INT.
22034 2018-04-11  Martin Jambor  <mjambor@suse.cz>
22036         PR ipa/84149
22037         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
22038         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
22039         not the same as the source val.
22040         (cgraph_edge_brings_value_p): New parameter.
22041         (gather_edges_for_value): Pass destination value to
22042         cgraph_edge_brings_value_p.
22043         (perhaps_add_new_callers): Likewise.
22044         (get_info_about_necessary_edges): Likewise and exclude values brought
22045         only by self-recursive edges.
22046         (create_specialized_node): Redirect only clones of self-calling edges.
22047         (+self_recursive_pass_through_p): New function.
22048         (find_more_scalar_values_for_callers_subset): Use it.
22049         (find_aggregate_values_for_callers_subset): Likewise.
22050         (known_aggs_to_agg_replacement_list): Removed.
22051         (decide_whether_version_node): Re-calculate known constants for all
22052         remaining context clones.
22054 2018-04-11  Richard Biener  <rguenther@suse.de>
22056         PR lto/85339
22057         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
22058         from early DWARF output.
22059         (dwarf2out_early_finish): Output line info unconditionally into
22060         early DWARF and add reference to it.
22062 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22064         PR target/85281
22065         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
22066         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
22067         other than V2DFmode using iptr mode attribute.
22068         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
22070 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
22072         PR rtl-optimization/84659
22073         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
22075 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22077         PR debug/85302
22078         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
22079         SIZEP is NULL.
22080         (output_loc_list): Pass address of a dummy size variable even in the
22081         locview handling loop.
22082         (index_location_lists): Add comment on why skip_loc_list_entry can't
22083         call size_of_locs.
22085 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22087         PR target/85261
22088         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
22089         into register.
22091 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
22093         PR target/85321
22094         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
22095         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
22096         and -mstring-compare-inline-limit.
22098 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22100         PR target/85287
22101         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
22102         for stack clash protection in a register whenever we need it to be in
22103         a register.
22105 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22107         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
22108         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
22110 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22112         PR target/85321
22113         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
22114         the help text.
22115         (mlong-double-): Ditto.
22116         * config/rs6000/sysv4.opt (msdata=): Ditto.
22117         (mtls-size=): Ditto.
22119 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22121         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
22122         erroneous entries for
22123         "vector int vec_ldl (int, long int *)", and
22124         "vector unsigned int vec_ldl (int, unsigned long int *)".
22125         Add comments and entries for
22126         "vector bool char vec_ldl (int, bool char *)",
22127         "vector bool short vec_ldl (int, bool short *)",
22128         "vector bool int vec_ldl (int, bool int *)",
22129         "vector bool long long vec_ldl (int, bool long long *)",
22130         "vector pixel vec_ldl (int, pixel *)",
22131         "vector long long vec_ldl (int, long long *)",
22132         "vector unsigned long long vec_ldl (int, unsigned long long *)".
22133         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
22134         type tree bool_long_long_type_node and correct definition of
22135         bool_V2DI_type_node to make reference to this new type tree.
22136         (rs6000_mangle_type): Replace erroneous reference to
22137         bool_long_type_node with bool_long_long_type_node.
22138         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
22139         comments to emphasize sign distinctions for char and int types and
22140         replace RS6000_BTI_bool_long constant with
22141         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
22142         use of RS6000_BTI_pixel.
22143         (bool_long_type_node): Remove this macro definition.
22144         (bool_long_long_type_node): New macro definition
22146 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
22148         PR rtl-optimization/85300
22149         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
22150         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
22151         simplify_unary_operation fails.
22153 2018-04-10  Martin Liska  <mliska@suse.cz>
22155         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
22156         cgraph_edge and ipa_ref.
22158 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
22160         PR target/85177
22161         PR target/85255
22162         * config/i386/sse.md
22163         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
22164         computation of the VEC_MERGE selector from mask.
22165         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
22166         Fix decoding of the VEC_MERGE selector into mask.
22168 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
22170         PR tree-optimization/85286
22171         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
22173 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
22175         * final.c (final_1): Set insn_last_address as well as
22176         insn_current_address.
22178 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22180         PR target/85173
22181         * explow.c (emit_stack_probe): Call validize_mem on memory location
22182         before passing it to gen_probe_stack.  Create address operand and
22183         legitimize it for the probe_stack_address case.
22185 2018-04-09  Jan Hubicka  <jh@suse.cz>
22187         PR lto/85078
22188         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
22189         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
22190         * tree.c (free_lang_data_in_type): Fix handling of binfos;
22191         walk basetypes.
22192         (free_lang_data): Rebuild type inheritance graph.
22194 2018-04-09  Martin Sebor  <msebor@redhat.com>
22196         * invoke.texi (-finline-small-functions): Mention other optimization
22197         options.
22198         (-findirect-inlining, -fpartial-inlining): Same.
22199         (-finline-functions-called-once): Same.
22200         (-freorder-blocks-and-partition): Same.
22202 2018-04-09  Jan Hubicka  <jh@suse.cz>
22204         PR rtl/84058
22205         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
22206         jumps; choose last target that matches the criteria (i.e.
22207         no partition changes for non-crossing jumps).
22208         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
22209         support for redirecting crossing jumps to non-crossing.
22211 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
22213         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
22214         also for naked functions.
22216 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
22218         * config/arc/arc.md (add_shift): New pattern.
22219         (add_shift2): Likewise.
22220         (sub_shift): Likewise.
22221         (sub_shift_cmp0_noout): Likewise.
22222         (compare_si_ashiftsi): Likewise.
22223         (xbfu_cmp0_noout): New combine pattern.
22224         (xbfu_cmp0"): Likewise.
22225         (movsi_set_cc_insn): Place the predicable variant first.
22226         (commutative_binary_cmp0_noout): Remove clobber.
22227         (commutative_binary_cmp0): New pattern.
22228         (noncommutative_binary_cmp0): Likewise.
22229         (noncommutative_binary_cmp0_noout): Likewise.
22230         (noncommutative_binary_comparison_result_used): Removed.
22231         (rsub_cmp0): New pattern.
22232         (rsub_cmp0_noout): Likewise.
22233         (extzvsi): Changed, keep only meaningful variants.
22234         (SQH, SEZ): New iterators.
22235         (SQH_postfix): New mode attribute.
22236         (SEZ_prefix): New code attribute.
22237         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
22238         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
22239         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
22240         of numerical value.
22241         (noncommutative_operator): Check the availability of barrel
22242         shifter option.
22244 2018-04-09  Richard Biener  <rguenther@suse.de>
22246         PR tree-optimization/85284
22247         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22248         Only use the niter constraining form of simple_iv when the exit
22249         is always executed.
22251 2018-04-09  Tom de Vries  <tom@codesourcery.com>
22253         PR target/84041
22254         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
22255         (define_expand "*memory_barrier"): New define_expand.
22256         (define_insn "memory_barrier"): New insn.
22258 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22260         PR rtl-optimization/80463
22261         PR rtl-optimization/83972
22262         PR rtl-optimization/83480
22264         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
22265         correct producer for the insn.
22266         (tidy_control_flow): Fixup seqnos in case of debug insns.
22268 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22270         PR rtl-optimization/83913
22272         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
22273         different sched-times when merging exprs.
22275 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22277         PR rtl-optimization/83962
22279         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
22280         tidy_fallthru_edge and tidy_control_flow.
22282 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22284         PR rtl-optimization/83530
22286         * sel-sched.c (force_next_insn): New global variable.
22287         (remove_insn_for_debug): When force_next_insn is true, also leave only
22288         next insn in the ready list.
22289         (sel_sched_region): When the region wasn't scheduled, make another pass
22290         over it with force_next_insn set to 1.
22292 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
22294         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
22295         into tm_file.
22296         * config/nds32/constants.md (unspec_volatile_element): Add enum values
22297         for interrupt control.
22298         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22299         functions for interrupt control.
22300         * config/nds32/nds32-intrinsic.md: Likewise.
22301         * config/nds32/nds32_intrinsic.h: Likewise.
22302         * config/nds32/nds32.h (nds32_builtins): Likewise.
22304 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
22306         * config/nds32/nds32.c (nds32_init_machine_status,
22307         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
22308         strict_aligned_p field.
22309         (nds32_expand_to_rtl_hook): New function.
22310         (TARGET_EXPAND_TO_RTL_HOOK): Define.
22311         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
22313 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22314             Chung-Ju Wu  <jasonwucj@gmail.com>
22316         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
22317         * config/nds32/nds32-n7.md: New file.
22318         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
22319         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
22320         pipeline.
22321         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
22322         * config/nds32/nds32.md (pipeline_model): Add n7.
22323         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
22324         * config/nds32/pipelines.md: Include n7 settings.
22326 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22327             Chung-Ju Wu  <jasonwucj@gmail.com>
22329         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
22330         * config/nds32/nds32-e8.md: New file.
22331         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
22332         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
22333         pipeline.
22334         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
22335         * config/nds32/nds32.md (pipeline_model): Add e8.
22336         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
22337         * config/nds32/pipelines.md: Include e8 settings.
22339 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22340             Chung-Ju Wu  <jasonwucj@gmail.com>
22342         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
22343         * config/nds32/nds32-n8.md: New file.
22344         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
22345         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
22346         pipeline.
22347         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
22348         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
22349         * config/nds32/nds32.md (pipeline_model): Add n8.
22350         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
22351         * config/nds32/pipelines.md: Include n8 settings.
22353 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22354             Chung-Ju Wu  <jasonwucj@gmail.com>
22356         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
22357         * config/nds32/nds32-n9-2r1w.md: New file.
22358         * config/nds32/nds32-n9-3r2w.md: New file.
22359         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
22360         nds32_register_ports): New or modify for cpu n9.
22361         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
22362         pipeline.
22363         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
22364         * config/nds32/nds32-utils.c: New file.
22365         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
22366         TARGET_MUL_SLOW): Define.
22367         * config/nds32/nds32.md (pipeline_model): New attribute.
22368         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
22369         New options that support cpu n9.
22370         * config/nds32/pipelines.md: Include n9 settings.
22371         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
22373 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
22375         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
22376         information if necessary.
22377         (output_cond_branch_compare_zero): Likewise.
22378         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
22379         (nds32_target_alignment): Refine for alignment.
22380         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
22381         (FUNCTION_BOUNDARY): Modify.
22382         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
22383         align case.
22384         * config/nds32/nds32.opt (malways-align, malign-functions): New.
22386 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
22388         * config/nds32/constants.md (unspec_volatile_element): Add values for
22389         TLB operation and data prefetch.
22390         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22391         functions for TLB operation and data prefetch.
22392         * config/nds32/nds32-intrinsic.md: Likewise.
22393         * config/nds32/nds32_intrinsic.h: Likewise.
22394         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
22395         (nds32_print_operand): Likewise.
22396         * config/nds32/nds32.h (nds32_builtins): Likewise.
22398 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22399         Andrew Pinski <pinsika@gcc.gnu.org>
22401         PR middle-end/82976
22402         * match.pd: Use constant_boolean_node of correct type instead of
22403         boolean_true_node or boolean_false_node for simplifying
22404         pointer comparisons to zero.
22406 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
22408         PR tree-optimization/80021
22409         * tree.c (verify_type_variant): Make error call in verify_variant_match
22410         translatable and remove final full stop.
22412 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22414         * config/nds32/constants.md (unspec_volatile_element): Add
22415         UNSPEC_VOLATILE_EH_RETURN.
22416         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
22417         nds32_output_stack_pop): Support dwarf exception handling process.
22418         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
22419         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
22420         exception handling process.
22421         (nds32_compute_stack_frame): Likewise.
22422         (nds32_return_addr_rtx): Likewise.
22423         (nds32_initial_elimination_offset): Likewise.
22424         (nds32_expand_prologue): Likewise.
22425         (nds32_expand_epilogue): Likewise.
22426         (nds32_dynamic_chain_address): New function.
22427         * config/nds32/nds32.h (machine_function): Add fields for dwarf
22428         exception handling.
22429         (DYNAMIC_CHAIN_ADDRESS): Define.
22430         (EH_RETURN_DATA_REGNO): Define.
22431         (EH_RETURN_STACKADJ_RTX): Define.
22432         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
22433         patterns for dwarf exception handling.
22435 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22437         * config/nds32/nds32.h: Clean up obsolete macros.
22439 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22441         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22442         Add enum values for particular instructions.
22443         * config/nds32/nds32-intrinsic.c: Implementation of expanding
22444         particular intrinsic functions.
22445         * config/nds32/nds32-intrinsic.md: Likewise.
22446         * config/nds32/nds32_intrinsic.h: Likewise.
22447         * config/nds32/nds32.h (nds32_builtins): Likewise.
22448         * config/nds32/nds32.md (type): Add pbsad and pbsada.
22449         (btst, ave): New patterns for particular instructions.
22451 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22453         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22454         Add enum values for atomic load/store and memory sync.
22455         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
22456         and memory sync.
22457         * config/nds32/nds32-intrinsic.md: Likewise.
22458         * config/nds32/nds32_intrinsic.h: Likewise.
22459         * config/nds32/nds32.h (nds32_builtins): Likewise.
22461 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
22463         PR tree-optimization/85257
22464         * fold-const.c (native_encode_vector): If not all elts could fit
22465         and off is -1, return 0 rather than offset.
22466         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
22467         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
22468         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
22469         adjust buffer in native_interpret_expr call.
22471 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22473         * config/nds32/constants.md (unspec_volatile_element): Add cache
22474         control enum values.
22475         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
22476         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
22477         * config/nds32/nds32.c (nds32_cctl_names): New.
22478         (nds32_print_operand): Handle cache control register names.
22479         * config/nds32/nds32.h (nds32_builtins): New enum values.
22480         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
22481         macros.
22482         * config/nds32/nds32.md (type): Add mmu.
22483         * config/nds32/pipelines.md (simple_insn): Add mmu.
22485 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22487         * config/nds32/nds32.md (type): Remove call.
22488         * config/nds32/pipelines.md (simple_insn): Likewise.
22490 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22492         * config/nds32/constants.md (unspec_volatile_element): Add
22493         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
22494         UNSPEC_VOLATILE_FMFCFG.
22495         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
22496         description for fmfcfg and fmfcsr.
22497         (bdesc_1arg): Add fmtcsr.
22498         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
22499         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
22500         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
22501         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
22502         unspec_fmfcfg): New patterns.
22503         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
22504         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
22505         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
22506         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
22507         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
22508         __nds32__fmfcfg): Define.
22510 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22512         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
22513         intrinsic register names.
22514         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
22515         intrinsic register enum values and macros.
22517 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22519         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
22520         for load/store addressing form.
22521         (nds32_print_operand_address): Likewise.
22523 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
22525         PR target/85196
22526         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
22527         based on LABEL_REF.  Remove useless assertion.
22528         (pic_address_needs_scratch): Fix formatting.
22529         (sparc_legitimize_pic_address): Minor tweaks.
22530         (sparc_delegitimize_address): Adjust assertion accordingly.
22531         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
22532         into symbolic_operand.
22533         (movsi_high_pic_label_ref): Likewise.
22534         (movsi_lo_sum_pic_label_ref): Likewise.
22535         (movdi_pic_label_ref): Likewise.
22536         (movdi_high_pic_label_ref): Likewise.
22537         (movdi_lo_sum_pic_label_ref): Likewise.
22539 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
22541         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
22542         custom LIB_SPEC setup.
22544 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
22545             Kito Cheng  <kito.cheng@gmail.com>
22547         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
22548         * config/riscv/freebsd.h: New.
22550 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22552         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
22553         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
22554         file.
22556 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22557             Kito Cheng  <kito.cheng@gmail.com>
22559         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
22560         nds32_output_call, nds32_symbol_binds_local_p): New functions.
22561         * config/nds32/nds32-protos.h (nds32_output_call,
22562         nds32_output_return): Declare.
22563         * config/nds32/nds32.md: Refine all the call and return patterns.
22565 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22567         PR debug/85252
22568         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
22569         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
22571         PR rtl-optimization/84872
22572         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
22573         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
22574         EDGE_CROSSING edge.
22576 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
22578         * expr.c (copy_blkmode_to_reg): Revert 254862.
22579         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
22581 2018-04-06  Richard Biener  <rguenther@suse.de>
22583         PR middle-end/85244
22584         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
22585         after seeing a component reference with an adjacent field.  Treat
22586         refs to arrays at struct end of external decls similar to
22587         refs to unconstrained commons.
22589 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22591         PR sanitizer/85213
22592         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
22593         look through SAVE_EXPRs with non-side-effects argument.  Adjust
22594         recursive calls.
22595         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
22596         save_p here.
22598 2018-04-06  Richard Biener  <rguenther@suse.de>
22600         PR middle-end/85180
22601         * alias.c (find_base_term): New wrapper around find_base_term
22602         unwinding CSELIB_VAL_PTR changes.
22603         (find_base_term): Do not restore CSELIB_VAL_PTR during the
22604         recursion.
22606 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22608         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
22609         instructions.
22610         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
22611         constant definitions.
22612         ("nop"): lr 0,0 -> nopr r0
22613         ("nop_lr0", "nop_lr1"): New insn definitions.
22615 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22617         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
22618         NDS32_V3PUSH_AVAILABLE_P macro.
22620 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
22621             Chung-Ju Wu  <jasonwucj@gmail.com>
22623         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
22624         (nds32*-*-*): Add float and fpu_config into supported_defaults.
22625         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
22626         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
22627         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
22628         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
22629         * config/nds32/constraints.md: New constraints and checking for hard
22630         float configuration.
22631         * config/nds32/iterators.md: New mode iterator and attribute for hard
22632         float configuration.
22633         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
22634         patterns.
22635         * config/nds32/nds32-fpu.md: New file.
22636         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
22637         deal with hard float code generation.
22638         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
22639         ARCH_V3S.
22640         (abi_type, float_reg_number): New enum type.
22641         * config/nds32/nds32-predicates.c: New predicates for hard float.
22642         * config/nds32/nds32-protos.h: Declare functions for hard float.
22643         * config/nds32/nds32.c: Implementation for hard float configuration.
22644         * config/nds32/nds32.h: Definitions for hard float configuration.
22645         * config/nds32/nds32.md: Include hard float machine description and
22646         modify patterns for hard float configuration.
22647         * config/nds32/nds32.opt: New options for hard float configuration.
22648         * config/nds32/predicates.md: New predicates for hard float
22649         configuration.
22651 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22653         * common/config/nds32/nds32-common.c
22654         (nds32_option_optimization_table): Enable -mreleax-hint by default.
22656 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
22658         PR middle-end/85195
22659         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
22660         CONSTRUCTOR_ELT (ctor, ...)->value.
22662 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
22664         PR target/85193
22665         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
22667 2018-04-05  Tom de Vries  <tom@codesourcery.com>
22669         PR target/85204
22670         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
22671         cond jump.
22673 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22674             Kito Cheng  <kito.cheng@gmail.com>
22676         * config/nds32/constraints.md (U33): Fine-tune checking condition.
22677         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
22678         * config/nds32/nds32.h (nds32_16bit_address_type): Add
22679         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
22681 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22682             Kito Cheng  <kito.cheng@gmail.com>
22684         * config/nds32/constraints.md (Ufe): New memory constraint.
22685         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
22686         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
22687         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
22688         operands.
22689         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
22690         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
22692 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22694         * config/nds32/nds32.md: Use optimize_size in the condition for
22695         alu-shift instructions.
22697 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22699         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
22701 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22703         * config/nds32/nds32.md (negsi2): Refine pattern.
22705 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22706             Chung-Ju Wu  <jasonwucj@gmail.com>
22708         * config/nds32/iterators.md (shift_rotate): New code iterator.
22709         (shift): New code attribute.
22710         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
22711         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
22712         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
22713         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
22714         bit-wise operations.
22715         (andsi3, *andsi3): Ditto.
22716         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
22717         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
22718         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
22719         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
22720         nds32_ior_operand, nds32_xor_operand): New predicates.
22722 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22724         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
22725         (addsi3, subsi3): ... this.
22727 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22729         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
22731 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22733         * config/nds32/nds32.md: Adjust indention.
22735 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22737         * config/nds32/nds32.md (feature): New attribute.
22739 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22741         * config/nds32/nds32.md (subtype): New attribute.
22743 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22745         PR target/85203
22746         * config/arm/arm-builtins.c (arm_expand_builtin): Change
22747         expansion to perform a bitwise AND of the argument followed by a
22748         boolean negation of the result.
22750 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
22752         PR rtl-optimization/84878
22753         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
22754         the basic block.  Assert the use reference is not artificial and that
22755         it has an associated insn.
22757 2018-04-04  Michael Matz  <matz@suse.de>
22759         * builtins.c (compute_objsize): Pass correct operand
22760         to array_at_struct_end_p.
22762 2018-04-04  Richard Biener  <rguenther@suse.de>
22764         PR lto/85176
22765         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
22766         from contexts for DINFO_LEVEL_TERSE and below.
22768 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22770         * config/nds32/nds32-doubleword.md (move_<mode>): Require
22771         resiter_operand condition.
22772         * config/nds32/nds32.md (*move<mode>): Ditto.
22774 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22775             Monk Chiang  <sh.chiang04@gmail.com>
22777         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
22779 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22781         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
22783 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22784             Kito Cheng  <kito.cheng@gmail.com>
22786         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
22787         nds32_cond_code_str, output_cond_branch,
22788         output_cond_branch_compare_zero, nds32_expand_cbranch,
22789         nds32_expand_cstore, nds32_expand_movcc,
22790         nds32_output_cbranchsi4_equality_zero,
22791         nds32_output_cbranchsi4_equality_reg,
22792         nds32_output_cbranchsi4_equality_reg_or_const_int,
22793         nds32_output_cbranchsi4_greater_less_zero: New functions.
22794         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
22795         nds32_expand_cstore, nds32_expand_movcc,
22796         nds32_output_cbranchsi4_equality_zero,
22797         nds32_output_cbranchsi4_equality_reg,
22798         nds32_output_cbranchsi4_equality_reg_or_const_int,
22799         nds32_output_cbranchsi4_greater_less_zero): Declare.
22800         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
22801         nds32_rimm11s_operand): New predicates.
22802         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
22803         * config/nds32/nds32.md: Rewrite all the branch and conditional move
22804         patterns.
22806 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22808         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
22809         * config/nds32/nds32.md: Ditto.
22810         * config/nds32/pipelines.md: Ditto.
22812 2018-04-04  Richard Biener  <rguenther@suse.de>
22814         PR tree-optimization/85168
22815         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
22816         propagating abnormals.
22818 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22820         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
22822 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22823             Kito Cheng  <kito.cheng@gmail.com>
22825         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
22826         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
22827         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
22828         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
22829         * config/nds32/nds32.md (sibcall_internal): New.
22830         (sibcall_register): Remove.
22831         (sibcall_immediate): Remove.
22832         (sibcall_value_internal): New.
22833         (sibcall_value_register): Remove.
22834         (sibcall_value_immediate): Remove.
22835         * config/nds32/predicates.md (nds32_general_register_operand): New.
22836         (nds32_call_address_operand): New.
22838 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22840         PR rtl-optimization/85167
22841         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
22842         bb_defs if *split_p, instead preinitialize it to NULL.
22844         PR tree-optimization/85156
22845         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
22846         evaluating the argument multiple times.
22848 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
22850         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
22851         than vector.
22852         (_mm_cvtpd_ps): Likewise.
22853         (_mm_cvttpd_epi32): Likewise.
22854         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
22855         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
22856         vector, pixel, and bool following altivec.h include.
22858 2018-04-03  Martin Sebor  <msebor@redhat.com>
22860         * doc/extend.texi (Common Function Attributes): Clarify.
22861         (const attribute): Likewise.
22862         (pure attribute): Likewise.
22864 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22866         PR target/85169
22867         * config/i386/i386.c (ix86_expand_vector_set): Use
22868         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
22870 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
22872         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
22873         instructions when changing rounding bits to preserve precision bits
22874         in the x87 control word.
22876 2018-04-03  Martin Liska  <mliska@suse.cz>
22878         PR tree-optimization/82491
22879         * rtl.h (strip_offset_and_add): Replace += suboffset with
22880         poly_uint64 () + suboffset.
22882 2018-03-29  Martin Liska  <mliska@suse.cz>
22883             Martin Jambor  <mjambor@suse.cz>
22885         PR ipa/84947
22886         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
22887         param_type is not an integral or pointer type.
22889 2018-04-03  Richard Biener  <rguenther@suse.de>
22891         * sese.h (recompute_all_dominators): Remove.
22893 2018-04-02  Martin Sebor  <msebor@redhat.com>
22895         * doc/invoke.texi (-Wrestrict): Fix typos.
22897 2018-04-02  Jim Wilson  <jimw@sifive.com>
22899         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
22900         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
22901         (<optab>di3, <optab>si3_extend): Likewise.
22902         (<optab>si3_mask, <optab>si3_mask_1): New.
22903         (<optab>di3_mask, <optab>di3_mask_1): New.
22904         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
22905         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
22906         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
22908 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
22910         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
22911         example.
22913 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
22915         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
22916         (nds32_canonicalize_comparison): New function.
22918 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22919             Kito Cheng  <kito.cheng@gmail.com>
22920             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22922         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
22923         * config/nds32/constants.md (unspec_volatile_element): Add
22924         UNSPEC_VOLATILE_RELAX_GROUP.
22925         * config/nds32/nds32-relax-opt.c: New file.
22926         * config/nds32/nds32-predicates.c
22927         (nds32_symbol_load_store_p): New function.
22928         * config/nds32/nds32-protos.h
22929         (nds32_symbol_load_store_p): Declare function.
22930         (make_pass_nds32_relax_opt): Declare new rtl pass function.
22931         * config/nds32/nds32.c
22932         (nds32_register_pass): New function to register pass.
22933         (nds32_register_passes): New function to register passes.
22934         * config/nds32/nds32.md (relax_group): New pattern.
22935         * config/nds32/nds32.opt (mrelax-hint): New option.
22936         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
22938 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
22940         * config/nds32/t-nds32: Modify files dependency.
22942 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22944         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
22945         (PROFILE_HOOK): Define its implementation.
22947 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22949         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
22950         type and 32-bit size.
22952 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
22954         PR middle-end/85090
22955         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
22956         (V_128_256): New mode iterator.
22957         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
22958         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
22959         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
22960         of V.
22961         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
22962         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
22964 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
22966         PR target/83315
22967         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
22968         NaN inputs correctly.
22970 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
22972         PR target/80546
22973         * config/rs6000/vsx.md (??r): New mode attribute.
22974         (*vsx_mov<mode>_64bit): Use it.
22975         (*vsx_mov<mode>_32bit): Likewise.
22977 2018-03-30  Martin Sebor  <msebor@redhat.com>
22979         PR tree-optimization/84818
22980         * builtins.c (check_access): Use warning_n.
22982 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22984         PR target/83822
22985         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
22986         condition.
22987         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
22988         condition.
22990 2018-03-30  Julia Koval  <julia.koval@intel.com>
22992         PR target/84413
22993         * x86-tune.def (movx, partial_reg_dependency): Enable for
22994         m_SKYLAKE_AVX512.
22996 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
22998         PR inline-asm/84985
22999         * lra-constraints.c (process_alt_operands): Move setting
23000         this_alternative_matches below.
23002 2018-03-29  Martin Liska  <mliska@suse.cz>
23004         PR lto/84995.
23005         * doc/invoke.texi: Document how LTO works with debug info.
23006         Describe auto-load support of binutils.  Mention 'x86-64'
23007         as valid option value of -march option.
23009 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
23011         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
23013         PR c/85094
23014         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
23015         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
23016         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
23017         checking.
23019 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
23021         PR target/84912
23022         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
23023         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
23024         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
23025         for RS6000_BTM_POWERPC64.
23026         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
23027         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
23028         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
23029         definition.
23030         (DIVDE): Use it.
23031         (DIVDEU): Likewise.
23033 2018-03-28  Carl Love  <cel@us.ibm.com>
23035         Revert
23036         2017-09-27  Carl Love  <cel@us.ibm.com>
23038         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
23039         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
23040         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
23041         fctiw instruction.
23043 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23045         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
23046         instead of __vector bool.
23047         (_mm_max_pu8): Likewise.
23048         (_mm_min_pi16): Likewise.
23050 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
23052         PR target/84912
23053         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
23054         (DIVWEUO): Likewise.
23055         (DIVDEO): Likewise.
23056         (DIVDEUO): Likewise.
23057         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
23058         DIVWEUO and DIVDEUO.
23059         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
23060         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
23061         (div_extend): Likewise.
23062         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
23063         builtin function.
23064         (__builtin_divweuo): Likewise.
23065         (__builtin_divdeo): Likewise.
23066         (__builtin_divdeuo): Likewise.
23068 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
23070         PR target/85095
23071         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
23072         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
23074         PR tree-optimization/82004
23075         * gimple-match-head.c (optimize_pow_to_exp): New function.
23076         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
23077         Don't fold to exp if optimize_pow_to_exp is false.
23079 2018-03-28  Martin Liska  <mliska@suse.cz>
23081         PR other/84819
23082         * calls.c (initialize_argument_information): Fix trailing space.
23083         * common.opt: Fix typo and provide better explanation for
23084         -fsanitize-coverage option.
23085         * config/i386/i386.opt: Fix typo.
23087 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
23088             Martin Liska  <mliska@suse.cz>
23090         PR sanitizer/85081
23091         * gimplify.c (asan_poison_variable): Don't do the check for
23092         gimplify_omp_ctxp here.
23093         (gimplify_decl_expr): Do it here.
23094         (gimplify_target_expr): Likewise.
23096 2018-03-28  Martin Liska  <mliska@suse.cz>
23098         PR target/84988
23099         * config/i386/i386.c (ix86_function_arg_advance): Do not call
23100         chkp_type_bounds_count if MPX is not enabled.
23102 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
23104         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
23106 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
23108         PR target/84914
23109         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
23110         function to create the function decl for complex long double
23111         multiply and divide for -mabi=ieeelongdouble.
23112         (init_float128_ieee): Call it.
23114 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
23116         PR target/85044
23117         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
23118         -fcf-protection=branch -mibt.
23119         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
23121 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
23123         PR target/81863
23124         * config/arm/arm.c (arm_valid_symbolic_address): Handle
23125         arm_word_relocations.
23127 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
23129         PR target/85056
23130         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
23131         extern array declarations.
23133 2018-03-27  Richard Biener  <rguenther@suse.de>
23135         PR middle-end/84067
23136         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
23137         explicit single_use checks.
23139 2018-03-27  Richard Biener  <rguenther@suse.de>
23141         PR tree-optimization/85082
23142         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
23143         Valueize the VUSE.
23145 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
23147         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
23148         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
23149         Turn on fasynchronous-unwind-tables and funwind-tables.
23151 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
23153         PR target/85073
23154         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
23155         (*bmi_blsr_<mode>_ccz): Ditto.
23157 2018-03-26  Tom de Vries  <tom@codesourcery.com>
23159         PR tree-optimization/85063
23160         * omp-general.c (offloading_function_p): New function.  Factor out
23161         of ...
23162         * omp-offload.c (pass_omp_target_link::gate): ... here.
23163         * omp-general.h (offloading_function_p): Declare.
23164         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
23165         with attribute omp declare target for offloading functions.
23167 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
23169         PR tree-optimization/84005
23170         * tree-data-ref.h (get_base_for_alignment): Declare.
23171         * tree-data-ref.c (get_base_for_alignment_1): New function.
23172         (get_base_for_alignment): Likewise.
23173         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
23174         get_base_for_alignment to find a suitable base object, instead
23175         of always using drb->base_address.
23177 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
23179         PR inline-asm/85022
23180         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
23181         known size by default.
23183 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
23185         PR inline-asm/85030
23186         * lra-constraints.c (process_alt_operands): Don't match BLKmode
23187         and non BLKmode operands.
23189 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23191         PR target/85026
23192         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
23193         Clean up attributes.
23195 2018-03-23  Richard Biener  <rguenther@suse.de>
23197         PR debug/85020
23198         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
23199         we are going to emit early debug for LTO.
23201 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
23203         PR inline-asm/85034
23204         * function.c (match_asm_constraints_1): Don't optimize if input
23205         doesn't satisfy general_operand predicate for output's mode.
23207         PR inline-asm/85022
23208         * alias.c (write_dependence_p): Don't require for x_canonicalized
23209         non-VOIDmode if x has VOIDmode.
23211         PR sanitizer/85029
23212         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
23213         just don't try to optimize it rather than assert it never happens.
23215 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23217         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
23218         macro expansions for definition of ST_INTERNAL_<mode> and
23219         LD_INTERNAL_<mode> builtins.
23220         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
23221         Remove prototype.
23222         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
23223         function.
23224         (altivec_expand_st_builtin): Likewise.
23225         (altivec_expand_builtin): Remove calls to deleted functions.
23226         (rs6000_address_for_altivec): Delete this function.
23227         * config/rs6000/vector.md: Remove expands for
23228         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
23230 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
23232         PR target/84826
23233         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
23234         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
23235         re-computing once computed.
23236         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
23237         (arm_init_machine_status): Initialize
23238         machine->static_chain_stack_bytes.
23240 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23242         PR target/84760
23243         * doc/extend.texi: Add four new prototypes for vec_ld.
23244         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
23245         definitions for more logical presentation.
23246         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
23247         entries for V1TI variants of __builtin_altivec_ld builtin.
23248         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
23249         handling of V1TI variant of LVX icode pattern.
23250         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
23251         (rs6000_gimple_fold_builtin): Likewise.
23252         (altivec_init_builtins): Add code to define
23253         __builtin_altivec_lvx_v1ti function.
23255 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
23257         PR inline-asm/84941
23258         * function.c (match_asm_constraints_1): Don't do the optimization
23259         if input isn't a REG, SUBREG, MEM or constant.
23261 2018-03-22  Tom de Vries  <tom@codesourcery.com>
23263         PR tree-optimization/84956
23264         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
23265         bb_has_abnormal_pred.
23267 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
23269         PR sanitizer/85018
23270         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
23271         DECL_INITIAL (decl) to decl at the end.
23272         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
23273         adjust the comment.
23275 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
23277         * doc/extend.texi (__builtin_tgmath): Document when complex
23278         integer types are treated as _Complex _Float64.
23280 2018-03-21  Tom de Vries  <tom@codesourcery.com>
23282         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
23284 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23286         PR tree-optimization/84960
23287         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
23288         if it is ENTRY block, move them into single succ of ENTRY in that case.
23290 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
23292         PR tree-optimization/84811
23293         * poly-int.h (poly_span_traits): Remove the T3 parameter and
23294         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
23295         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
23296         (known_subrange_p): Update accordingly.  Cast each value involved
23297         in the size comparison, rather than casting the result of the
23298         subtraction.
23300 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23302         PR tree-optimization/84982
23303         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
23304         by flipping the least significant bit rather than all bits from
23305         bitpos to bitpos + bitsize - 1.
23307 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23309         * doc/extend.texi (Deprecated Features): Remove mention of
23310         long-deleted deprecations.
23312 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23314         PR jit/84288
23315         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
23316         * configure: Regenerate.
23318 2018-03-21  Tom de Vries  <tom@codesourcery.com>
23320         PR tree-optimization/83126
23321         * tree-parloops.c (num_phis): New function.
23322         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
23324 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23326         * doc/extend.texi (Deprecated Features): Update deprecated flags,
23327         mention anon-struct/union members and trailing attributes.
23329 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
23331         PR tree-optimization/84969
23332         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
23333         builtin memset partitions if they set different rhs values.
23335 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23337         PR rtl-optimization/84989
23338         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
23339         VEC_DUPLICATE with scalar result mode.
23341 2018-03-21  Martin Liska  <mliska@suse.cz>
23343         PR ipa/84963
23344         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
23345         not intended return statement.
23347 2018-03-21  Martin Liska  <mliska@suse.cz>
23349         PR target/84988
23350         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
23351         (chkp_find_bound_slots_1): Limit number of iterations.
23353 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
23355         PR target/84838
23356         * Minor grammar fixes for x86 options.
23358 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23360         PR debug/84875
23361         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
23362         holding REG_CFA_RESTORE notes, instead turn them into a USE.
23364 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
23366         PR target/83789
23367         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
23368         (altivec_lvx_<mode>_1op): Likewise.
23369         (altivec_stvx_<mode>_2op): Likewise.
23370         (altivec_stvx_<mode>_1op): Likewise.
23371         (altivec_lvx_<VM2:mode>): New define_expand.
23372         (altivec_stvx_<VM2:mode>): Likewise.
23373         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
23374         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23375         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
23376         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23377         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
23378         (rs6000_gen_lvx): Likewise.
23379         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
23380         (altivec_expand_stv_builtin): Likewise.
23381         (altivec_expand_builtin): Likewise.
23382         * config/rs6000/vector.md: Likewise.
23384 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23386         PR target/82518
23387         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
23388         BYTES_BIG_ENDIAN.
23390 2018-03-20  Richard Biener  <rguenther@suse.de>
23392         PR target/84986
23393         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
23394         sign-conversions as zero, fall back to standard scalar_stmt
23395         cost for the rest.
23397 2018-03-20  Martin Liska  <mliska@suse.cz>
23399         PR ipa/84825
23400         * predict.c (rebuild_frequencies): Handle case when we have
23401         PROFILE_ABSENT, but flag_guess_branch_prob is false.
23403 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23405         PR target/84990
23406         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
23407         flag_section_anchors.
23408         * varasm.c (use_blocks_for_decl_p): Remove hack for
23409         dw2_force_const_mem.
23411         PR target/84845
23412         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
23413         to ...
23414         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
23415         be created, use lowpart_subreg of operands[0] rather than operands[0]
23416         itself.
23417         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
23418         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
23419         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
23420         and n constraint instead of aarch64_shift_imm_di and Usd.
23421         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
23422         (*aarch64_<optab>_reg_minus<mode>3): ... this.
23424 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
23426         PR target/82989
23427         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
23428         to favor GPR over NEON registers.
23429         (<shift>di3_neon): Likewise.
23431 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23433         PR target/84952
23434         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
23435         (nvptx_process_pars): Emit bar.sync asap and alap.
23437 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23439         PR target/84954
23440         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
23441         seen_label if seen_label is already set.
23443 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23445         PR target/84945
23446         * config/i386/i386.c (fold_builtin_cpu): For features above 31
23447         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
23448         Use 1U instead of 1.  Formatting fixes.
23450         PR c/84953
23451         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
23452         instead of TREE_TYPE (s1) for the return value.
23454 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
23456         PR tree-optimization/84946
23457         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
23458         bitsize + bitsize in poly_uint64 rather than poly_int64.
23460         PR sanitizer/78651
23461         * dwarf2asm.c: Include fold-const.c.
23462         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
23463         of decl rather than decl itself.
23465         PR rtl-optimization/84643
23466         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
23468 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
23470         PR sanitizer/78651
23471         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
23472         calling assemble_variable.
23474 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
23476         PR target/81647
23477         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
23478         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
23480 2018-03-19  Jim Wilson  <jimw@sifive.com>
23482         PR bootstrap/84856
23483         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
23484         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
23485         (riscv_first_stack_step): Likewise.
23486         (riscv_option_override): Use STACK_BOUNDARY instead of
23487         MIN_STACK_BOUNDARY.
23488         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
23489         MIN_STACK_BOUNDARY.
23490         (BIGGEST_ALIGNMENT): Set to 128.
23491         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
23492         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
23493         STACK_BOUNDARY.
23495 2018-03-19  Richard Biener  <rguenther@suse.de>
23497         PR tree-optimization/84933
23498         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
23499         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
23501 2018-03-19  Richard Biener  <rguenther@suse.de>
23503         PR tree-optimization/84859
23504         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
23505         (cond_if_else_store_replacement): Perform sinking operation on
23506         single-store BBs regardless of MAX_STORES_TO_SINK setting.
23507         Generalize what a BB with a single eligible store is.
23509 2018-03-19  Richard Biener  <rguenther@suse.de>
23511         PR tree-optimization/84929
23512         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
23513         chrec_is_positive against non-chrec arg.
23515 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
23517         PR target/84711
23518         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
23520 2018-03-18  Martin Liska  <mliska@suse.cz>
23522         PR rtl-optimization/84635
23523         * regrename.c (build_def_use): Use matches_mode only when
23524         matches >= 0.
23526 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
23528         PR tree-optimization/84913
23529         * tree-vect-loop.c (vectorizable_reduction): Don't try to
23530         vectorize chains of COND_EXPRs.
23532 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23534         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
23536 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23538         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
23540 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23542         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
23544 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23545             Kito Cheng  <kito.cheng@gmail.com>
23547         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
23548         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
23549         (nds32_adjust_reg_alloc_order): New function.
23550         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
23552 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
23554         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
23555         nds32_print_operand, nds32_print_operand_address): Use
23556         HOST_WIDE_INT_PRINT_DEC instead.
23558 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23560         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
23562 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
23564         PR target/84902
23565         * config/i386/i386.c (initial_ix86_tune_features,
23566         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
23567         unsigned long long.
23568         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
23569         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
23570         rather than 1u << ix86_tune.  Formatting fix.
23571         (ix86_option_override_internal): Change ix86_arch_mask from
23572         unsigned int to unsigned HOST_WIDE_INT, initialize to
23573         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
23574         (ix86_function_specific_restore): Likewise.
23576 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23578         PR target/84899
23579         * postreload.c (reload_combine_recognize_pattern): Perform
23580         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
23581         truncate_int_for_mode the result for the destination's mode.
23583         PR c/84909
23584         * hsa-gen.c (mem_type_for_type): Fix comment typo.
23585         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
23586         Likewise.
23587         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23588         Likewise.
23590 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
23592         PR target/84876
23593         * lra-assigns.c (lra_split_hard_reg_for): Don't use
23594         regno_allocno_class_array and sorted_pseudos.
23595         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
23596         insns where regno is used.
23598 2018-03-16  Martin Liska  <mliska@suse.cz>
23600         PR ipa/84833
23601         * multiple_target.c (create_dispatcher_calls): Redirect
23602         reference in the symbol table.
23604 2018-03-16  Martin Liska  <mliska@suse.cz>
23606         PR ipa/84722
23607         * multiple_target.c (create_dispatcher_calls): Redirect also
23608         an alias.
23610 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23612         PR c++/79937
23613         PR c++/82410
23614         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
23615         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
23616         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
23618 2018-03-16  Julia Koval  <julia.koval@intel.com>
23620         * doc/invoke.texi (Skylake Server): Add CLWB.
23621         Cannonlake): Remove CLWB.
23623 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23625         PR tree-optimization/84841
23626         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
23627         1 << 3.
23628         (FLOAT_ONE_CONST_TYPE): Define.
23629         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
23630         (sort_by_operand_rank): Put entries with higher constant_type last
23631         rather than first to match comments.
23633 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
23635         * config/nios2/nios2.md (movsi_internal): Fix thinko in
23636         split predicate.
23638 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23640         PR c++/79085
23641         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
23642         check and use address of target always.
23644 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
23646         PR target/84574
23647         * config/i386/i386.c (indirect_thunk_needed): Update comments.
23648         (indirect_thunk_bnd_needed): Likewise.
23649         (indirect_thunks_used): Likewise.
23650         (indirect_thunks_bnd_used): Likewise.
23651         (indirect_return_needed): New.
23652         (indirect_return_bnd_needed): Likewise.
23653         (output_indirect_thunk_function): Add a bool argument for
23654         function return.
23655         (output_indirect_thunk_function): Don't generate alias for
23656         function return thunk.
23657         (ix86_code_end): Call output_indirect_thunk_function to generate
23658         function return thunks.
23659         (ix86_output_function_return): Set indirect_return_bnd_needed
23660         and indirect_return_needed instead of indirect_thunk_bnd_needed
23661         and indirect_thunk_needed.
23663 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
23665         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
23666         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
23667         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
23669 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
23670             Paul Hua <paul.hua.gm@gmail.com>
23672         PR c/84852
23673         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
23675 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
23677         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
23678         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
23679         resp. SFmode cases.
23681 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
23683         PR target/84711
23684         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
23685         instead of GET_MODE_SIZE when comparing Units.
23687 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
23689         PR target/68256
23690         * varasm.c (hash_section): Return an unchangeble hash value
23691         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
23692         Return !aarch64_can_use_per_function_literal_pools_p ().
23694 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23696         PR target/84860
23697         * optabs.c (emit_conditional_move): Pass address of cmode's copy
23698         rather than address of cmode as last argument to prepare_cmp_insn.
23700 2018-03-15  Julia Koval  <julia.koval@intel.com>
23702         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
23703         F_AVX512VNNI, F_AVX512BITALG): New.
23705 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
23707         PR target/83451
23708         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
23709         insn for floating-point loads and stores.
23711 2018-03-14  Carl Love  <cel@us.ibm.com>
23713         * config/rs6000/rs6000-c.c: Add macro definitions for
23714         ALTIVEC_BUILTIN_VEC_PERMXOR.
23715         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
23716         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
23717         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
23718         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
23719         UNSPEC_VPERMXOR.
23720         * config/doc/extend.texi: Add prototypes for vec_permxor.
23722 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
23724         PR c/84852
23725         * diagnostic-show-locus.c (class layout_point): Convert m_line
23726         from int to linenum_type.
23727         (line_span::comparator): Use linenum "compare" function when
23728         comparing line numbers.
23729         (test_line_span): New function.
23730         (layout_range::contains_point): Convert param "row" from int to
23731         linenum_type.
23732         (layout_range::intersects_line_p): Likewise.
23733         (layout::will_show_line_p): Likewise.
23734         (layout::print_source_line): Likewise.
23735         (layout::should_print_annotation_line_p): Likewise.
23736         (layout::print_annotation_line): Likewise.
23737         (layout::print_leading_fixits): Likewise.
23738         (layout::annotation_line_showed_range_p): Likewise.
23739         (struct line_corrections): Likewise for field m_row.
23740         (line_corrections::line_corrections): Likewise for param "row".
23741         (layout::print_trailing_fixits): Likewise.
23742         (layout::get_state_at_point): Likewise.
23743         (layout::get_x_bound_for_row): Likewise.
23744         (layout::print_line): Likewise.
23745         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
23746         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
23747         * input.c (selftest::test_linenum_comparisons): New function.
23748         (selftest::input_c_tests): Call it.
23749         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
23750         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
23751         * selftest.h (ASSERT_GT): New macro.
23752         (ASSERT_GT_AT): New macro.
23753         (ASSERT_LT): New macro.
23754         (ASSERT_LT_AT): New macro.
23756 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
23758         PR rtl-optimization/84780
23759         * combine.c (distribute_links): Don't make a link based on pc_rtx.
23761 2018-03-14  Martin Liska  <mliska@suse.cz>
23763         * tree.c (record_node_allocation_statistics): Use
23764         get_stats_node_kind.
23765         (get_stats_node_kind): New function extracted from
23766         record_node_allocation_statistics.
23767         (free_node): Use get_stats_node_kind.
23769 2018-03-14  Richard Biener  <rguenther@suse.de>
23771         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
23772         that the value-set of ANTIC_IN doesn't grow.
23774         Revert
23775         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23776         member.
23777         (BB_VISITED_WITH_VISITED_SUCCS): New define.
23778         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23780 2018-03-14  Julia Koval  <julia.koval@intel.com>
23782         * config.gcc (icelake-client, icelake-server): New.
23783         (icelake): Remove.
23784         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
23785         (initial_ix86_arch_features): Ditto.
23786         (PTA_SKYLAKE): Add SGX.
23787         (PTA_ICELAKE): Remove.
23788         (PTA_ICELAKE_CLIENT): New.
23789         (PTA_ICELAKE_SERVER): New.
23790         (ix86_option_override_internal): Split up icelake on icelake client and
23791         icelake server.
23792         (get_builtin_code_for_version): Ditto.
23793         (fold_builtin_cpu): Ditto.
23794         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
23795         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
23796         * config/i386/i386.h (processor_type): Ditto.
23797         * doc/invoke.texi: Ditto.
23799 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
23801         PR sanitizer/83392
23802         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
23803         INTEGER_CST offset, add it together with bitpos / 8 and
23804         sign extend based on POINTER_SIZE.
23806         PR target/84844
23807         Revert
23808         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
23810         PR target/78090
23811         * config/i386/constraints.md (Yc): New register constraint.
23812         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
23813         Use Yc constraint for alternative 2 of operand 0.  Remove
23814         preferred_for_speed attribute.
23816 2018-03-14  Richard Biener  <rguenther@suse.de>
23818         PR tree-optimization/84830
23819         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
23820         with the old one to avoid oscillations.
23822 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
23824         PR target/83712
23825         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
23826         pseudos.
23827         (assign_by_spills): Return a flag of reload assignment failure.
23828         Do not process the reload assignment failures.  Do not spill other
23829         reload pseudos if they has the same reg class.  Update n if
23830         necessary.
23831         (lra_assign): Add a return arg.  Set up from the result of
23832         assign_by_spills call.
23833         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23834         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23835         usage_insns if it is not NULL.
23836         (spill_hard_reg_in_range): New function.
23837         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23838         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23839         function prototypes.
23840         (lra_assign): Change prototype.
23841         * lra.c (lra): Add code to deal with fails by splitting hard reg
23842         live ranges.
23844 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
23846         * config/riscv/riscv.opt (mrelax): New option.
23847         * config/riscv/riscv.c (riscv_file_start): Emit ".option
23848         "norelax" when riscv_mrelax is disabled.
23849         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
23851 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23853         PR target/84743
23854         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
23855         reassociation for int modes.
23857 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23859         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
23860         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
23861         for big-endian.
23862         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
23863         * config/aarch64/aarch64-sve.md
23864         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
23865         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
23866         (*extend<mode><Vwide>2): Rename to...
23867         (aarch64_sve_extend<mode><Vwide>2): ...this.
23868         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
23869         renaming the old pattern to...
23870         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
23871         unsigned packs.
23872         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
23873         define_expand, renaming the old pattern to...
23874         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
23875         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
23876         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
23877         account when deciding which SVE instruction the optab should use.
23878         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
23880 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23882         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
23883         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
23884         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
23885         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
23886         (tlsdesc_small_<mode>): Turn a define_expand and use
23887         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
23888         (tlsdesc_small_advsimd_<mode>): ...this.
23889         (tlsdesc_small_sve_<mode>): New pattern.
23891 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23893         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
23894         (UNSPEC_UMUL_HIGHPART): New constants.
23895         (MUL_HIGHPART): New int iteraor.
23896         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
23897         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
23898         define_expand.
23899         (*<su>mul<mode>3_highpart): New define_insn.
23901 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
23903         PR lto/84805
23904         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23905         incomplete types.
23907 2018-03-13  Martin Liska  <mliska@suse.cz>
23909         PR ipa/84658.
23910         * (sem_item_optimizer::sem_item_optimizer): Initialize new
23911         vector.
23912         (sem_item_optimizer::~sem_item_optimizer): Release it.
23913         (sem_item_optimizer::merge_classes): Register variable aliases.
23914         (sem_item_optimizer::fixup_pt_set): New function.
23915         (sem_item_optimizer::fixup_points_to_sets): Likewise.
23916         * ipa-icf.h: Declare new variables and functions.
23918 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
23920         PR middle-end/84834
23921         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
23922         integer_pow2p@2 and test integer_pow2p in condition.
23923         (A < 0 ? C : 0): Similarly for @1.
23925         PR middle-end/84831
23926         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
23927         characters starting at p contain '\0' character, don't look beyond
23928         that.
23930         PR target/84827
23931         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
23932         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
23934         PR target/84828
23935         * reg-stack.c (change_stack): Change update_end var from int to
23936         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
23937         also call set_block_for_insn on the newly added insns and rescan.
23939         PR target/84786
23940         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
23941         on the last operand.
23943         PR c++/84704
23944         * tree.c (stabilize_reference_1): Return save_expr (e) for
23945         STATEMENT_LIST even if it doesn't have side-effects.
23947 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
23949         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
23951 2018-03-12  Renlin Li  <renlin.li@arm.com>
23953         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
23954         aarch64_output_scalar_simd_mov_immediate.
23956 2018-03-12  Martin Sebor  <msebor@redhat.com>
23958         PR tree-optimization/83456
23959         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
23960         for perfectly overlapping calls to memcpy.
23961         (gimple_fold_builtin_memory_chk): Same.
23962         (gimple_fold_builtin_strcpy): Handle no-warning.
23963         (gimple_fold_builtin_stxcpy_chk): Same.
23964         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
23966 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23968         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
23969         parameter.  Use it for SFmode.
23970         (rs6000_function_arg_advance_1): Adjust.
23971         (rs6000_function_arg): Adjust.
23972         (rs6000_gimplify_va_arg): Pass false for that new parameter.
23974 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23976         PR rtl-optimization/84169
23977         PR rtl-optimization/84780
23978         * combine.c (can_combine_p): Check for a 2-insn combination whether
23979         the destination register is used between the two insns, too.
23981 2018-03-12  Richard Biener  <rguenther@suse.de>
23983         PR tree-optimization/84803
23984         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
23985         for refs DR analysis didn't process.
23987 2018-03-12  Richard Biener  <rguenther@suse.de>
23989         PR tree-optimization/84777
23990         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
23991         force-vectorize loops ignore whether we are optimizing for size.
23993 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
23995         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
23996         (TARGET_MD_ASM_ADJUST): Define.
23998 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
23999             Kito Cheng  <kito.cheng@gmail.com>
24000             Chung-Ju Wu  <jasonwucj@gmail.com>
24002         * config/nds32/nds32.c (nds32_compute_stack_frame,
24003         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
24004         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
24005         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
24006         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
24007         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
24008         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
24009         * config/nds32/nds32.md (prologue, epilogue): Use macro
24010         NDS32_V3PUSH_AVAILABLE_P to do checking.
24012 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
24014         PR debug/58150
24015         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
24016         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
24017         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
24018         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
24019         addition of most attributes on !orig_type_die or the attribute not
24020         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
24022 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
24023             Chung-Ju Wu  <jasonwucj@gmail.com>
24025         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
24026         __NDS32_VH__ macro.
24027         * config/nds32/nds32.opt (mvh): New option.
24029 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
24030             Chung-Ju Wu  <jasonwucj@gmail.com>
24032         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
24033         function.
24034         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
24035         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
24036         definition.
24038 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
24039             Chung-Ju Wu  <jasonwucj@gmail.com>
24041         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
24042         function.
24043         * config/nds32/nds32-multiple.md (strlensi): New pattern.
24044         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
24046 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
24047             Kito Cheng  <kito.cheng@gmail.com>
24048             Chung-Ju Wu  <jasonwucj@gmail.com>
24050         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
24051         UNSPEC_FFMISM and UNSPEC_FLMISM.
24052         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
24053         for ffb, ffmism and flmism.
24054         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
24055         (unspec_ffmism): Ditto.
24056         (unspec_flmism): Ditto.
24057         (nds32_expand_builtin_impl): Check if string extension is available.
24058         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
24059         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
24061 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
24063         Reverting patch:
24064         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
24066         PR target/83712
24067         * lra-assigns.c (assign_by_spills): Return a flag of reload
24068         assignment failure.  Do not process the reload assignment
24069         failures.  Do not spill other reload pseudos if they has the same
24070         reg class.
24071         (lra_assign): Add a return arg.  Set up from the result of
24072         assign_by_spills call.
24073         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24074         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
24075         usage_insns if it is not NULL.
24076         (spill_hard_reg_in_range): New function.
24077         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24078         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24079         function prototypes.
24080         (lra_assign): Change prototype.
24081         * lra.c (lra): Add code to deal with fails by splitting hard reg
24082         live ranges.
24084 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
24086         PR target/84807
24087         * config/i386/i386.opt: Replace Enforcment with Enforcement.
24089 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
24091         PR debug/84620
24092         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
24093         (dw_val_node): Add val_symbolic_view.
24094         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
24095         (symview_upper_bound): New.
24096         (new_line_info_table): Initialize symviews_since_reset.
24097         (dwarf2out_source_line): Count symviews_since_reset and set
24098         symview_upper_bound.
24099         (dw_val_equal_p): Handle symview.
24100         (add_AT_symview): New.
24101         (print_dw_val): Handle symview.
24102         (attr_checksum, attr_checksum_ordered): Likewise.
24103         (same_dw_val_p, size_of_die): Likewise.
24104         (value_format, output_die): Likewise.
24105         (add_high_low_attributes): Use add_AT_symview for entry_view.
24106         (dwarf2out_finish): Reset symview_upper_bound, clear
24107         zero_view_p.
24109 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
24111         PR target/83969
24112         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
24113         Add strict argument and use it.
24114         (rs6000_split_multireg_move): Update for new strict argument.
24115         (mem_operand_gpr): Disallow all non-offsettable addresses.
24116         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
24118 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
24120         PR target/84772
24121         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
24122         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
24123         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
24125         PR c++/84767
24126         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
24127         decl, use remap_type if we want to use the type.
24129 2018-03-09  Martin Sebor  <msebor@redhat.com>
24131         PR tree-optimization/84526
24132         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
24133         Remove dead code.
24134         (builtin_access::generic_overlap): Be prepared to handle non-array
24135         base objects.
24137 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
24139         PR rtl-optimization/84682
24140         * lra-constraints.c (process_address_1): Check is_address flag
24141         for address constraints.
24142         (process_alt_operands): Likewise.
24143         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
24144         preprocess_constraints.
24145         * recog.h (preprocess_constraints): Add oploc parameter.
24146         Adjust callers.
24147         * recog.c (preprocess_constraints): Test address_operand for
24148         CT_ADDRESS constraints.
24150 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
24152         PR target/83712
24153         * lra-assigns.c (assign_by_spills): Return a flag of reload
24154         assignment failure.  Do not process the reload assignment
24155         failures.  Do not spill other reload pseudos if they has the same
24156         reg class.
24157         (lra_assign): Add a return arg.  Set up from the result of
24158         assign_by_spills call.
24159         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24160         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
24161         usage_insns if it is not NULL.
24162         (spill_hard_reg_in_range): New function.
24163         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24164         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24165         function prototypes.
24166         (lra_assign): Change prototype.
24167         * lra.c (lra): Add code to deal with fails by splitting hard reg
24168         live ranges.
24170 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24172         PR target/83193
24173         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
24174         Accept complain bool parameter.  Only emit errors if it is true.
24175         (arm_parse_cpu_option_name): Likewise.
24176         (arm_target_thumb_only): Adjust callers of the above.
24177         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
24178         prototype to take a default true bool parameter.
24179         (arm_parse_arch_option_name): Likewise.
24181 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
24182             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
24184         PR jit/64089
24185         PR jit/84288
24186         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
24187         * configure: Regenerate.
24188         * configure.ac ("linker --version-script option"): New.
24189         ("linker soname option"): New.
24191 2018-03-09  Richard Biener  <rguenther@suse.de>
24193         PR tree-optimization/84775
24194         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
24195         immediate uses of predicate stmts and mark them modified.
24197         Revert
24198         PR tree-optimization/84178
24199         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24200         to caller.
24201         (version_loop_for_if_conversion): Delay update_ssa call.
24202         (tree_if_conversion): Delay update_ssa until after predicate
24203         insertion.
24205 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
24207         PR target/84763
24208         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
24209         when the function accesses prior frames.
24211 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
24213         PR debug/84456
24214         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
24215         gen_llsym, otherwise call maybe_gen_llsym.
24217         PR inline-asm/84742
24218         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
24219         has ',' character inside of it.
24221 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24223         PR target/84748
24224         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
24225         as clobbering CC_REGNUM.
24227 2018-03-08  Richard Biener  <rguenther@suse.de>
24229         PR middle-end/84552
24230         * tree-scalar-evolution.c: Include tree-into-ssa.h.
24231         (follow_copies_to_constant): Do not follow SSA names registered
24232         for update.
24234 2018-03-08  Richard Biener  <rguenther@suse.de>
24236         PR tree-optimization/84178
24237         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24238         to caller.
24239         (version_loop_for_if_conversion): Delay update_ssa call.
24240         (tree_if_conversion): Delay update_ssa until after predicate
24241         insertion.
24243 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
24245         PR tree-optimization/84178
24246         * tree-if-conv.c (release_bb_predicate): Remove the
24247         the assertion that the stmts have NULL use_ops.
24248         Discard the statements, asserting that they haven't
24249         yet been added to a BB.
24251 2018-03-08  Richard Biener  <rguenther@suse.de>
24253         PR tree-optimization/84746
24254         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
24255         (phi_translate): Pass in destination ANTIC_OUT set.
24256         (phi_translate_1): Likewise.  For a simplified result lookup
24257         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
24258         (phi_translate_set): Adjust.
24259         (do_pre_regular_insertion): Likewise.
24260         (do_pre_partial_partial_insertion): Likewise.
24262 2018-03-08  Martin Liska  <mliska@suse.cz>
24264         PR gcov-profile/84735
24265         * doc/gcov.texi: Document usage of profile files.
24266         * gcov-io.h: Document changes in the format.
24268 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
24270         PR debug/84404
24271         PR debug/84408
24272         * dwarf2out.c (struct dw_line_info_table): Update comments for
24273         view == -1.
24274         (FORCE_RESET_NEXT_VIEW): New.
24275         (FORCE_RESETTING_VIEW_P): New.
24276         (RESETTING_VIEW_P): Check for -1 too.
24277         (ZERO_VIEW_P): Likewise.
24278         (new_line_info_table): Force-reset next view.
24279         (dwarf2out_begin_function): Likewise.
24280         (dwarf2out_source_line): Simplify zero_view_p initialization.
24281         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
24282         view directly.  Omit view when omitting .loc at line 0.
24284 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
24286         PR tree-optimization/84740
24287         * tree-switch-conversion.c (process_switch): Call build_constructors
24288         only if info.phi_count is non-zero.
24290         PR tree-optimization/84739
24291         * tree-tailcall.c (find_tail_calls): Check call arguments against
24292         DECL_ARGUMENTS (current_function_decl) rather than
24293         DECL_ARGUMENTS (func) when checking for tail recursion.
24295 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24297         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
24298         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
24299         Volker Reichelt's entry and add entries for people that perform
24300         GCC fuzzy testing and report numerous bugs.
24302 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
24304         PR target/82411
24305         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
24306         readonly data in sdata, if that is disabled.
24307         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
24308         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
24309         -mreadonly-in-sdata option.
24311 2018-03-07  Martin Sebor  <msebor@redhat.com>
24313         PR tree-optimization/84468
24314         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
24315         basic block when looking for nul assignment.
24317 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
24319         PR target/84277
24320         * except.h (output_function_exception_table): Adjust prototype.
24321         * except.c (output_function_exception_table): Remove FNNAME parameter
24322         and add SECTION parameter.  Ouput one part of the table at a time.
24323         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
24324         the first part of the exception table and emit unwind directives.
24325         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
24326         (i386_pe_seh_cold_init): Likewise.
24327         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
24328         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
24329         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
24330         (ix86_output_call_insn): Emit a nop in one more case for SEH.
24331         * config/i386/winnt.c: Include except.h.
24332         (struct seh_frame_state): Add reg_offset, after_prologue and
24333         in_cold_section fields.
24334         (i386_pe_seh_end_prologue): Set seh->after_prologue.
24335         (i386_pe_seh_cold_init): New function.
24336         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
24337         to seh->in_cold_section.
24338         (seh_emit_push): Record the offset of the push.
24339         (seh_emit_save): Record the offet of the save.
24340         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
24341         Test seh->after_prologue to disregard the epilogue.
24342         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
24343         (i386_pe_end_cold_function): New function.
24345 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24347         PR fortran/84565
24348         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
24349         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
24351         PR c++/84704
24352         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
24353         on tmp_var.
24354         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
24355         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
24357         PR middle-end/84723
24358         * multiple_target.c: Include tree-inline.h and intl.h.
24359         (expand_target_clones): Diagnose and fail if node->definition and
24360         !tree_versionable_function_p (node->decl).
24362 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
24364         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
24365         sprint_ul.
24366         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
24367         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
24368         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
24370 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
24372         PR target/84710
24373         * combine.c (try_combine): Use reg_or_subregno instead of handling
24374         just paradoxical SUBREGs and REGs.
24376 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
24378         * config/arc/arc.c (arc_finalize_pic): Remove function.
24379         (arc_must_save_register): We use single base PIC register, remove
24380         checks to save/restore the PIC register.
24381         (arc_expand_prologue): Likewise.
24382         * config/arc/arc-protos.h (arc_set_default_type_attributes):
24383         Remove.
24384         (arc_verify_short): Likewise.
24385         (arc_attr_type): Likewise.
24386         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
24387         (walk_stores): Likewise.
24388         (arc_address_cost): Make it static.
24389         (arc_verify_short): Likewise.
24390         (branch_dest): Likewise.
24391         (arc_attr_type): Likewise.
24392         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
24393         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
24394         (arc_final_prescan_insn): Remove inserting the nops due to
24395         hardware hazards.  It is done in reorg step.
24396         (insn_length_variant_t): Remove.
24397         (insn_length_parameters_t): Likewise.
24398         (arc_insn_length_parameters): Likewise.
24399         (arc_get_insn_variants): Likewise.
24400         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
24402 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
24404         PR inline-asm/84683
24405         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
24406         assertion failure.
24408         PR tree-optimization/84687
24409         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
24410         on new_node->decl.
24411         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
24413 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24415         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
24416         Rename to ppc_speculation_barrier.
24417         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
24418         __builtin_ppc_speculation_barrier.
24420 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
24422         PR target/84700
24423         * combine.c (combine_simplify_rtx): Don't try to simplify if
24424         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
24425         are equal to x.
24427 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
24429         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
24430         to 32 bytes when compiling for POWER9.
24432 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
24434         PR target/84564
24435         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
24436         regparm >= 3 with no arg reg available also for calls with
24437         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
24439         PR target/84524
24440         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
24441         orig,vex.
24442         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
24444 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
24446         PR target/84264
24447         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
24449 2018-03-05  Richard Biener  <rguenther@suse.de>
24451         PR tree-optimization/84486
24452         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
24453         When inserting a __builtin_assume_aligned call set the LHS
24454         SSA name alignment info accordingly.
24456 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
24458         PR tree-optimization/84114
24459         * config/aarch64/aarch64.c (aarch64_reassociation_width)
24460         Avoid reassociation of FLOAT_MODE addition.
24462 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
24464         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
24465         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
24466         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
24467         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
24468         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
24469         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
24470         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
24471         and -mwbnoinvd.
24472         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
24473         __builtin_ia32_wbinvd): New builtins.
24474         (SPECIAL_ARGS2): New.
24475         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
24476         (SPECIAL_ARGS2): New.
24477         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
24478         (ix86_valid_target_attribute_inner_p): Ditto.
24479         (ix86_init_mmx_sse_builtins): Add special_args2.
24480         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
24481         TARGET_WBNOINVD_P): New.
24482         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
24483         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
24484         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
24485         * config/i386/immintrin.h (_wbinvd): New intrinsic.
24486         * config/i386/pconfigintrin.h: New file.
24487         * config/i386/wbnoinvdintrin.h: Ditto.
24488         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
24489         wbnoinvdintrin.h.
24490         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
24492 2018-03-05  Richard Biener  <rguenther@suse.de>
24494         PR tree-optimization/84670
24495         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
24496         member.
24497         (BB_VISITED_WITH_VISITED_SUCCS): New define.
24498         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
24499         (compute_antic_aux): Only assert the number of values in ANTIC_IN
24500         doesn't grow if all successors (recursively) were visited at least
24501         once.
24503 2018-03-05  Richard Biener  <rguenther@suse.de>
24505         PR tree-optimization/84650
24506         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
24507         if executed in the loop pipeline.
24509 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
24511         * doc/configfiles.texi (Configuration Files): Move info about
24512         conditionalizing $target-protos.h to...
24513         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
24514         differs from $target-protos.h.
24516 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
24517             Chung-Ju Wu  <jasonwucj@gmail.com>
24519         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
24520         * config/nds32/nds32-multiple.md (setmemsi): Define.
24521         * config/nds32/nds32-memory-manipulation.c
24522         (nds32_gen_dup_4_byte_to_word_value): New.
24523         (emit_setmem_word_loop): New.
24524         (emit_setmem_byte_loop): New.
24525         (nds32_expand_setmem_loop): New.
24526         (nds32_expand_setmem_loop_v3m): New.
24527         (nds32_expand_setmem_unroll): New.
24528         (nds32_expand_setmem): New.
24530 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24531             Chung-Ju Wu  <jasonwucj@gmail.com>
24533         * config/nds32/nds32-memory-manipulation.c
24534         (nds32_emit_load_store): New.
24535         (nds32_emit_post_inc_load_store): New.
24536         (nds32_emit_mem_move): New.
24537         (nds32_emit_mem_move_block): New.
24538         (nds32_expand_movmemsi_loop_unknown_size): New.
24539         (nds32_expand_movmemsi_loop_known_size): New.
24540         (nds32_expand_movmemsi_loop): New.
24541         (nds32_expand_movmemsi_unroll): New.
24542         (nds32_expand_movmemqi): Rename ...
24543         (nds32_expand_movmemsi): ... to this.
24544         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
24545         (movmemsi): ... to this.
24546         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
24547         (nds32_expand_movmemsi): ... to this.
24549 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24550             Monk Chiang  <sh.chiang04@gmail.com>
24551             Chung-Ju Wu  <jasonwucj@gmail.com>
24553         * config/nds32/nds32-protos.h
24554         (nds32_expand_load_multiple): New arguments.
24555         (nds32_expand_store_multiple): Ditto.
24556         (nds32_valid_multiple_load_store): Rename ...
24557         (nds32_valid_multiple_load_store_p): ... to this.
24558         * config/nds32/nds32-memory-manipulation.c
24559         (nds32_expand_load_multiple): Refine implementation.
24560         (nds32_expand_store_multiple): Ditto.
24561         * config/nds32/nds32-multiple.md
24562         (load_multiple): Update nds32_expand_load_multiple interface.
24563         (store_multiple): Update nds32_expand_store_multiple interface.
24564         * config/nds32/nds32-predicates.c
24565         (nds32_valid_multiple_load_store): Rename ...
24566         (nds32_valid_multiple_load_store_p): ... to this and refine
24567         implementation.
24568         * config/nds32/predicates.md
24569         (nds32_load_multiple_and_update_address_operation): New predicate.
24570         (nds32_store_multiple_and_update_address_operation): New predicate.
24572 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24573             Chung-Ju Wu  <jasonwucj@gmail.com>
24575         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
24576         (combo): New attribute.
24577         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
24579 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
24581         * config/nds32/nds32.opt: Change -mcmodel= default value.
24583 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
24584             Monk Chiang  <sh.chiang04@gmail.com>
24585             Chung-Ju Wu  <jasonwucj@gmail.com>
24587         * config/nds32/constants.md (unspec_element): New enum.
24588         * config/nds32/constraints.md (Umw): New constraint.
24589         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
24590         * config/nds32/nds32-intrinsic.md: Likewise.
24591         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
24592         (nds32_valid_smw_lwm_base_p): New.
24593         (nds32_output_smw_single_word): New.
24594         (nds32_output_lmw_single_word): New.
24595         (nds32_expand_unaligned_load): New.
24596         (nds32_expand_unaligned_store): New.
24597         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
24598         (nds32_output_smw_single_word): Declare.
24599         (nds32_output_lmw_single_word): Declare.
24600         (nds32_expand_unaligned_load): Declare.
24601         (nds32_expand_unaligned_store): Declare.
24602         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
24603         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
24604         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
24605         NDS32_BUILTIN_UASTORE_DW.
24606         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
24607         predicate.
24609 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
24610             Kito Cheng  <kito.cheng@gmail.com>
24611             Chung-Ju Wu  <jasonwucj@gmail.com>
24613         * config/nds32/nds32-intrinsic.c
24614         (nds32_expand_builtin_null_ftype_reg): Delete.
24615         (nds32_expand_builtin_reg_ftype_imm): Ditto.
24616         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
24617         (nds32_read_argument): New.
24618         (nds32_legitimize_target): Ditto.
24619         (nds32_legitimize_argument): Ditto.
24620         (nds32_check_constant_argument): Ditto.
24621         (nds32_expand_unop_builtin): Ditto.
24622         (nds32_expand_unopimm_builtin): Ditto.
24623         (nds32_expand_binop_builtin): Ditto.
24624         (nds32_builtin_decl_impl): Ditto.
24625         (builtin_description): Ditto.
24626         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
24627         (nds32_init_builtins_impl): Ditto.
24628         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
24629         (nds32_builtin_decl): New.
24630         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
24631         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
24633 2018-03-02  Jeff Law  <law@redhat.com>
24635         * reorg.c (stop_search_p): Handle DEBUG_INSN.
24636         (redundant_insn, fill_simple_delay_slots): Likewise.
24637         (fill_slots_from_thread): Likewise.
24638         * resource.c (mark_referenced_resources): Likewise.
24639         (mark_set_resources, find_dead_or_set_registers): Likewise.
24641 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24643         * substring-locations.h (format_warning_va): Formatting fix for
24644         ATTRIBUTE_GCC_DIAG.
24645         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
24646         argument.
24647         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
24648         * substring-locations.c: Include intl.h.
24649         (format_warning_va): Turned into small wrapper around
24650         format_warning_n_va, renamed to ...
24651         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
24652         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
24653         use ngettext.
24654         (format_warning_at_substring_n): New function.
24655         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
24656         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
24657         format_warning_at_substring with just a shorter name instead of
24658         const function pointer.
24659         (fmtwarn_n): New function.
24660         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
24661         appropriate, get rid of all the fmtstr temporaries, move conditionals
24662         with G_() wrapped string literals directly into fmtwarn arguments,
24663         cast dir.len to (int), formatting fixes.
24665 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
24667         * doc/invoke.texi: Remove "Cilk Plus" references.
24669 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24670             Richard Biener  <rguenther@suse.de>
24672         PR ipa/84628
24673         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
24674         for error or warning attributes if CALL_FROM_THUNK_P is set.
24675         Formatting fixes.
24677 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24679         PR target/56540
24680         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
24681         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
24683         PR target/56540
24684         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
24685         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
24687         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
24688         instead of -1U in last predictors element's probability member.
24690 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
24692         PR ipa/83983
24693         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
24694         arguments if they are comparable.
24696 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24698         PR tree-optimization/84634
24699         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
24700         masks and masked_loop_p with a single loop_masks, making sure it's
24701         null for bb vectorization.
24703 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24705         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
24706         (vect_analyze_data_ref_access): Use loop->safe_len rather than
24707         loop->force_vectorize to check whether there is no alias.
24709 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24711         PR target/84614
24712         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
24713         prototypes.
24714         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
24715         comments.
24716         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
24717         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
24718         instead of a loop around prev_real_insn.
24719         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
24720         prev_real_insn.
24722         PR inline-asm/84625
24723         * config/i386/i386.c (ix86_print_operand): Use conditional
24724         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
24725         zero vector.
24727 2018-03-02  Richard Biener  <rguenther@suse.de>
24729         PR tree-optimization/84427
24730         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
24731         (bitmap_set_subtract_values): Rewrite to handle multiple
24732         exprs per value.
24733         (clean): Likewise.
24734         (prune_clobbered_mems): Likewise.
24735         (phi_translate): Take edge instead of pred/phiblock.
24736         (phi_translate_1): Likewise.
24737         (phi_translate_set): Likewise.  Insert all translated
24738         exprs for a value into the set, keeping possibly multiple
24739         expressions per value.
24740         (compute_antic_aux): Adjust for phi_translate changes.
24741         When intersecting union the expressions and prune those
24742         not in the final value set, keeping possibly multiple
24743         expressions per value.  Do not use value-insertion
24744         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
24745         all expressions.  Add verification that the value-sets
24746         only shrink during iteration.
24747         (compute_partial_antic_aux): Adjust for the phi_translate changes.
24748         (do_pre_regular_insertion): Likewise.
24749         (do_pre_partial_partial_insertion): Likewise.
24751 2018-03-02  Richard Biener  <rguenther@suse.de>
24753         PR target/82005
24754         * config/darwin.c (saved_debug_info_level): New static global.
24755         (darwin_asm_lto_start): Disable debug info generation for LTO out.
24756         (darwin_asm_lto_end): Restore debug info generation settings.
24758 2018-03-01  Martin Liska  <mliska@suse.cz>
24760         PR sanitizer/82484
24761         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
24762         volatile arguments.
24764 2018-03-01  Richard Biener  <rguenther@suse.de>
24766         PR debug/84645
24767         * dwarf2out.c (gen_variable_die): Properly handle late VLA
24768         type annotation with LTO when debug was disabled at compile-time.
24770 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
24772         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
24773         XINT with INTVAL.
24774         (mips_final_postscan_insn): Likewise.
24776 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
24778         PR rtl-optimization/84528
24779         * alias.c (init_alias_target): Add commentary.
24780         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
24781         a unique base value if the frame pointer is not eliminated
24782         to the stack pointer.
24784 2018-03-01  Tom de Vries  <tom@codesourcery.com>
24786         PR rtl-optimization/83327
24787         * lra-int.h (hard_regs_spilled_into): Declare.
24788         * lra.c (hard_regs_spilled_into): Define.
24789         (init_reg_info): Init hard_regs_spilled_into.
24790         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
24791         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
24792         (process_bb_lives): Handle hard_regs_spilled_into.
24793         (lra_create_live_ranges_1): Before doing liveness propagation, clear
24794         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
24796 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
24798         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
24799         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
24800         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
24801         * config/rs6000/aix72.h: New file.
24803 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
24805         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
24806         instead of warning_at with conditional singular and plural messages
24807         where possible.
24809         PR target/52991
24810         * stor-layout.c (update_alignment_for_field): For
24811         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
24812         && !DECL_PACKED (field), do the alignment update, just use
24813         only desired_align instead of MAX (type_align, desired_align)
24814         as the alignment.
24815         (place_field): Don't do known_align < desired_align handling
24816         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
24817         is non-NULL, instead do it after rli->prev_field handling and
24818         only if not within a bitfield word.  For DECL_PACKED (field)
24819         use type_align of BITS_PER_UNIT.
24821 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
24823         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
24824         superfluous parentheses and trailing spaces.
24826 2018-02-28  Richard Biener  <rguenther@suse.de>
24828         PR tree-optimization/84584
24829         * graphite-scop-detection.c (scop_detection::add_scop): Discard
24830         SCoPs with fake exit edge.
24832 2018-02-28  Martin Liska  <mliska@suse.cz>
24834         PR testsuite/84597
24835         * timevar.c (timer::print): Fix format to properly print 100%
24836         values.
24838 2018-02-28  Richard Biener  <rguenther@suse.de>
24840         PR middle-end/84607
24841         * genmatch.c (capture_info::walk_match): Do not mark
24842         captured expressions without operands as expr_p given
24843         they act more like predicates and should be subject to
24844         "lost tail" side-effect preserving.
24846 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
24848         PR rtl-optimization/81611
24849         * auto-inc-dec.c (attempt_change): Move dead note from
24850         mem_insn if it's the next use of regno
24851         (find_address): Take address use of reg holding
24852         non-incremented value.  Add parm to limit search to the named
24853         reg only.
24854         (merge_in_block): Attempt to use a mem insn that is the next
24855         use of the original regno.
24857 2018-02-27  Martin Sebor  <msebor@redhat.com>
24859         PR c++/83871
24860         * doc/invoke.texi (-Wmissing-attributes): New option.
24861         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
24863 2018-02-27  Martin Sebor  <msebor@redhat.com>
24865         PR translation/84207
24866         * diagnostic-core.h (warning_n, error_n, inform_n): Change
24867         n argument to unsigned HOST_WIDE_INT.
24868         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
24869         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
24870         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
24871         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
24873 2018-02-27  Richard Biener  <rguenther@suse.de>
24875         PR tree-optimization/84512
24876         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
24877         Do not use the estimate returned from record_stmt_cost for
24878         the scalar iteration cost but sum properly using add_stmt_cost.
24880 2018-02-27  Richard Biener  <rguenther@suse.de>
24882         PR tree-optimization/84466
24883         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24884         Adjust last change to less strictly validate use operands.
24886 2018-02-27  Martin Liska  <mliska@suse.cz>
24888         PR gcov-profile/84548
24889         * gcov.c (process_file): Allow partial overlap and consider it
24890         also as group functions.
24891         (output_lines): Properly calculate range of lines for a group.
24893 2018-02-27  Martin Liska  <mliska@suse.cz>
24895         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
24896         'ggc' suffixes.  Change first column width.
24897         (timer::print): Fix formatting of the column.
24899 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
24901         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
24902         preserve inline entry blocks for the sake of debug inline
24903         entry point markers alone.
24904         (remove_unused_locals): Suggest in comments a better place to
24905         force the preservation of inline entry blocks that are
24906         otherwise unused, but do not preserve them.
24908 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24910         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
24912 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24914         PR target/84039
24915         * config/i386/constraints.md (Bs): Replace
24916         ix86_indirect_branch_register with
24917         TARGET_INDIRECT_BRANCH_REGISTER.
24918         (Bw): Likewise.
24919         * config/i386/i386.md (indirect_jump): Likewise.
24920         (tablejump): Likewise.
24921         (*sibcall_memory): Likewise.
24922         (*sibcall_value_memory): Likewise.
24923         Peepholes of indirect call and jump via memory: Likewise.
24924         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
24925         (*sibcall_value_GOT_32): Likewise.
24926         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24927         (GOT_memory_operand): Likewise.
24928         (call_insn_operand): Likewise.
24929         (sibcall_insn_operand): Likewise.
24930         (GOT32_symbol_operand): Likewise.
24931         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
24933 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24935         PR rtl-optimization/83496
24936         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
24937         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
24938         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
24939         redundant insn, if any.
24940         (relax_delay_slots): Likewise.
24941         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
24943 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
24945         PR tree-optimization/83965
24946         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
24947         that grouped statements are part of a reduction chain.  Return
24948         true if the statement is not marked as a reduction itself but
24949         is part of a group.
24950         (vect_recog_dot_prod_pattern): Don't check whether the statement
24951         is part of a group here.
24952         (vect_recog_sad_pattern): Likewise.
24953         (vect_recog_widen_sum_pattern): Likewise.
24955 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24957         PR debug/84545
24958         * final.c (rest_of_clean_state): Also look for calls inside sequences.
24960 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24962         PR target/84530
24963         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
24964         the bool argument.
24965         (ix86_output_indirect_function_return): New prototype.
24966         (ix86_split_simple_return_pop_internal): Likewise.
24967         * config/i386/i386.c (indirect_return_via_cx): New.
24968         (indirect_return_via_cx_bnd): Likewise.
24969         (indirect_thunk_name): Handle return va CX_REG.
24970         (output_indirect_thunk_function): Create alias for
24971         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
24972         (ix86_output_indirect_jmp): Remove the bool argument.
24973         (ix86_output_indirect_function_return): New function.
24974         (ix86_split_simple_return_pop_internal): Likewise.
24975         * config/i386/i386.md (*indirect_jump): Don't pass false
24976         to ix86_output_indirect_jmp.
24977         (*tablejump_1): Likewise.
24978         (simple_return_pop_internal): Change it to define_insn_and_split.
24979         Call ix86_split_simple_return_pop_internal to split it for
24980         -mfunction-return=.
24981         (simple_return_indirect_internal): Call
24982         ix86_output_indirect_function_return instead of
24983         ix86_output_indirect_jmp.
24985 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
24987         PR bootstrap/84405
24988         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
24989         memset and value initialization afterwards.
24991 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
24993         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
24995 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
24997         PR target/84521
24998         * common/config/aarch64/aarch64-common.c
24999         (aarch_option_optimization_table[]): Switch
25000         off fomit-frame-pointer
25002 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
25003             Chung-Ju Wu  <jasonwucj@gmail.com>
25005         * config/nds32/nds32-multiple.md (load_multiple): Disallow
25006         volatile memory.
25007         (store_multiple): Ditto.
25009 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
25011         * config.gcc: Add --with-cpu support for nds32 target.
25012         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
25013         * config/nds32/nds32.opt: Add -mcpu= option.
25015 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
25017         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
25018         isel=yes): Warn for these deprecated options.
25020 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
25022         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
25023         ISA_2_5_MASKS_EMBEDDED.
25025 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
25027         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
25028         p->max as pointers rather than using iterative_hash_expr.
25030 2018-02-23  Carl Love  <cel@us.ibm.com>
25032         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
25033         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
25034         BU_P8V_OVERLOAD_2.
25035         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
25036         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
25037         P8V_BUILTIN_VEC_VUNSIGNED2.
25039 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
25041         PR target/81572
25042         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
25043         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
25044         LRA_UNKNOWN_ALT.
25045         * lra-constraints.c (curr_insn_transform): Set up
25046         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
25047         LRA_UNKNOWN_ALT.
25048         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
25049         * lra-eliminations.c (spill_pseudos): Ditto.
25050         (process_insn_for_elimination): Ditto.
25051         * lra-lives.c (reg_early_clobber_p): Use the new macros.
25052         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
25053         LRA_NON_CLOBBERED_ALT.
25055 2018-02-22  Martin Sebor  <msebor@redhat.com>
25057         PR tree-optimization/84480
25058         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
25059         to maybe_diag_stxncpy_trunc.  Call it.
25060         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
25061         from gimple_fold_builtin_strcpy.  Print inlining stack.
25062         (handle_builtin_stxncpy): Print inlining stack.
25063         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
25065 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
25067         PR target/84176
25068         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
25069         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
25070         and -fcheck-pointer-bounds are used together.
25071         (indirect_thunk_prefix): New enum.
25072         (indirect_thunk_need_prefix): New function.
25073         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
25074         "_nt" instead of "_bnd" for NOTRACK prefix.
25075         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
25076         (output_indirect_thunk_function): Likewise.
25077         (): Likewise.
25078         (ix86_code_end): Update output_indirect_thunk_function calls.
25079         (ix86_output_indirect_branch_via_reg): Replace
25080         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
25081         (ix86_output_indirect_branch_via_push): Likewise.
25082         (ix86_output_function_return): Likewise.
25083         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
25084         incompatible with -fcf-protection=branch and
25085         -fcheck-pointer-bounds.
25087 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
25089         PR target/83335
25090         * config/aarch64/aarch64.c (aarch64_print_address_internal):
25091         Change gcc_assert call to output_operand_lossage.
25093 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
25095         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
25097 2018-02-22  DJ Delorie  <dj@redhat.com>
25098             Sebastian Perta  <sebastian.perta@renesas.com>
25099             Oleg Endo  <olegendo@gcc.gnu.org>
25101         * config/rx/rx.c (rx_rtx_costs): New function.
25102         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
25104 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25106         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
25108 2018-02-22  Martin Liska  <mliska@suse.cz>
25110         PR driver/83193
25111         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
25112         Add "native" as a possible value.
25114 2018-02-22  Martin Liska  <mliska@suse.cz>
25116         PR driver/83193
25117         * config/i386/i386.c (ix86_option_override_internal):
25118         Add "native" as a possible value for -march and -mtune.
25120 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
25122         PR target/84502
25123         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
25124         to all type variants.
25126         PR tree-optimization/84503
25127         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
25128         width as info->bitpos + info->bitsize - start.
25129         (merged_store_group::merge_overlapping): Simplify width computation.
25130         (check_no_overlap): New function.
25131         (imm_store_chain_info::try_coalesce_bswap): Compute expected
25132         start + width and last_order of the group, fail if check_no_overlap
25133         fails.
25134         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
25135         to group if check_no_overlap fails.
25137 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
25139         * config/rs6000/altivec.md: Delete contraint arguments to
25140         define_expand, define_split, and define_peephole2, and in
25141         define_insn_and_split if always unused.
25142         * config/rs6000/darwin.md: Ditto.
25143         * config/rs6000/dfp.md: Ditto.
25144         * config/rs6000/rs6000.md: Ditto.
25145         * config/rs6000/sync.md: Ditto.
25146         * config/rs6000/vector.md: Ditto.
25147         * config/rs6000/vsx.md: Ditto.
25149 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
25151         * config/rs6000/altivec.md: Write output control strings as braced
25152         blocks instead of double-quoted strings.
25153         * config/rs6000/darwin.md: Ditto.
25154         * config/rs6000/rs6000.md: Ditto.
25155         * config/rs6000/vector.md: Ditto.
25156         * config/rs6000/vsx.md: Ditto.
25158 2018-02-21  Jason Merrill  <jason@redhat.com>
25160         PR c++/84314 - ICE with templates and fastcall attribute.
25161         * attribs.c (build_type_attribute_qual_variant): Remove assert.
25163 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
25165         * ipa-cp.c (determine_versionability): Fix comment typos.
25167 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
25169         PR c/84229
25170         * ipa-cp.c (determine_versionability): Do not version functions caling
25171         va_arg_pack.
25173 2018-02-21  Martin Liska  <mliska@suse.cz>
25175         PR driver/83193
25176         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
25177         Add "native" as a possible value.
25178         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
25179         the macro when native cpu detection is available.
25181 2018-02-21  Martin Liska  <mliska@suse.cz>
25183         PR driver/83193
25184         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
25185         Add "native" as a possible value.
25186         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
25187         when native cpu detection is available.
25189 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
25190             Martin Sebor  <msebor@redhat.com>
25192         PR tree-optimization/84478
25193         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
25194         false.
25195         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
25196         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
25197         support which is conservatively correct, for 2 only stay conservative
25198         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
25199         argument to the 2 argument get_range_strlen, adjust 6 arg
25200         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
25201         false.
25202         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
25203         (gimple_fold_builtin_strlen): Pass true as last argument to
25204         get_range_strlen.
25206 2018-02-20  Martin Sebor  <msebor@redhat.com>
25208         PR middle-end/84095
25209         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
25210         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
25211         (builtin_memref::builtin_memref): Factor out parts into
25212         set_base_and_offset and call it.
25214 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
25216         PR middle-end/84406
25217         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
25218         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
25219         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
25220         search at the associated MODE_INT.
25222 2018-02-20  Jeff Law  <law@redhat.com>
25224         PR middle-end/82123
25225         PR tree-optimization/81592
25226         PR middle-end/79257
25227         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
25228         for range data rather than using global data.
25229         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
25230         range data rather than using global data.
25231         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
25232         pass it to children as needed.
25233         (struct directive::fmtresult): Similarly.
25234         (struct directive::set_width): Similarly.
25235         (struct directive::set_precision): Similarly.
25236         (format_integer, format_directive, parse_directive): Similarly.
25237         (format_none): Accept unnamed vr_values parameter.
25238         (format_percent, format_floating, format_character): Similarly.
25239         (format_string, format_plain): Similarly.
25240         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
25241         the EVRP range analyzer for range data rather than using global data.
25242         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
25243         gimple-ssa-evrp-analyze.h
25244         (class sprintf_dom_walker): Add after_dom_children member function.
25245         Add evrp_range_analyzer member.
25246         (sprintf_dom_walker::before_dom_children): Call into the EVRP
25247         range analyzer as needed.
25248         (sprintf_dom_walker::after_dom_children): New member function.
25249         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
25250         if not optimizing.
25251         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
25252         (evrp_range_analyzer::pop_to_marker): Likewise.
25254 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
25256         PR tree-optimization/84419
25257         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
25258         with the required type if its current type is compatible but
25259         different.
25261 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
25263         PR middle-end/82004
25264         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
25265         after vectorization.
25267 2018-02-20  Martin Liska  <mliska@suse.cz>
25269         PR driver/83193
25270         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
25271         possible values if we don't have a hint.
25273 2018-02-20  Martin Liska  <mliska@suse.cz>
25275         PR c/84310
25276         PR target/79747
25277         * final.c (shorten_branches): Build align_tab array with one
25278         more element.
25279         * opts.c (finish_options): Add alignment option limit check.
25280         (MAX_CODE_ALIGN): Likewise.
25281         (MAX_CODE_ALIGN_VALUE): Likewise.
25282         * doc/invoke.texi: Document maximum allowed option value for
25283         all -falign-* options.
25285 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
25287         PR target/84146
25288         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
25289         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
25290         * var-tracking.c (emit_note_insn_var_location): Remove all references
25291         to NOTE_INSN_CALL_ARG_LOCATION.
25292         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
25293         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
25294         Use copy_rtx_if_shared.
25295         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
25296         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
25297         (dwarf2out_var_location): Remove handling of
25298         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
25299         on call_insn.
25300         * final.c (final_scan_insn): Remove all references to
25301         NOTE_INSN_CALL_ARG_LOCATION.
25302         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
25303         before dumping final insns.
25304         * except.c (emit_note_eh_region_end): Remove all references to
25305         NOTE_INSN_CALL_ARG_LOCATION.
25306         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
25307         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
25308         * config/arc/arc.c (hwloop_optimize): Likewise.
25309         * config/arm/arm.c (create_fix_barrier): Likewise.
25310         * config/s390/s390.c (s390_chunkify_start): Likewise.
25311         * config/sh/sh.c (find_barrier): Likewise.
25312         * config/i386/i386.c (rest_of_insert_endbranch,
25313         ix86_seh_fixup_eh_fallthru): Likewise.
25314         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
25315         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
25316         * config/frv/frv.c (frv_function_prologue): Likewise.
25317         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
25318         reg note.
25319         (note_outside_basic_block_p): Remove all references to
25320         NOTE_INSN_CALL_ARG_LOCATION.
25321         * gengtype.c (adjust_field_rtx_def): Likewise.
25322         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
25323         Likewise.
25324         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
25325         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
25327         PR c++/84444
25328         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
25329         is ADDR_EXPR.
25331         PR tree-optimization/84452
25332         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
25333         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
25334         is NULL.
25336 2018-02-19  Martin Liska  <mliska@suse.cz>
25338         PR sanitizer/82183
25339         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
25341 2018-02-19  Martin Liska  <mliska@suse.cz>
25342             Richard Sandiford  <richard.sandiford@linaro.org>
25344         PR tree-optimization/82491
25345         * gimple-fold.c (get_base_constructor): Make earlier bail out
25346         to prevent ubsan.
25348 2018-02-19  Carl Love  <cel@us.ibm.com>
25350         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
25351         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
25352         BU_P8V_OVERLOAD_1.
25353         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
25354         P8V_BUILTIN_VEC_NEG.
25356 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
25358         * config/rl78/rl78.md (movdf): New define expand.
25360 2018-02-19  Martin Liska  <mliska@suse.cz>
25362         PR other/80589
25363         * doc/invoke.texi: Fix typo.
25364         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
25366 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
25368         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
25369         handle rs6000_single_float and rs6000_double_float specially for
25370         e500 family CPUs.
25372 2018-02-16  Jeff Law  <law@redhat.com>
25374         * config/rx/rx.c (add_pop_cfi_notes): New function.;
25375         (pop_regs): Use it.
25377 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
25379         PR ipa/84425
25380         * ipa-inline.c (inline_small_functions): Fix a typo.
25382 2018-02-16  Nathan Sidwell  <nathan@acm.org>
25384         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
25386 2018-02-16  Carl Love  <cel@us.ibm.com>
25388         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
25389         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
25390         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
25391         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
25392         expansion to P8V_BUILTIN_VEC_FLOAT2.
25394 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
25396         PR rtl-optimization/70023
25397         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
25398         src_regno into account.
25400 2018-02-16  Carl Love  <cel@us.ibm.com>
25402         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
25403         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
25404         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
25405         * config/rs6000/rs6000.c: Remove case statements for
25406         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
25407         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
25408         and P9V_BUILTIN_VEC_VINSERT4B.
25409         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
25410         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
25411         * config/rs6000/vsx.md:
25412         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
25413         vec_insert4b.
25415 2018-02-16  Carl Love  <cel@us.ibm.com>
25417         * config/rs6000/altivec.h: Add builtin names vec_extract4b
25418         vec_insert4b.
25419         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
25420         definitions.
25421         * config/rs6000/rs6000-c.c: Add the definitions for
25422         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
25423         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
25424         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
25425         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
25426         definition for insert4b and define insn *insert3b_internal.
25427         * doc/extend.texi: Add documentation for vec_extract4b.
25429 2018-02-16  Nathan Sidwell  <nathan@acm.org>
25431         * doc/extend.texi (Backwards Compatibility): Mention friend
25432         injection.  Note for-scope is deprecated.
25433         * doc/invoke.texi (-ffriend-injection): Deprecate.
25435 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
25437         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
25438         that moved to I2, also allow destinations that are a paradoxical
25439         subreg (instead of a normal reg).
25441 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
25443         PR target/83831
25444         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
25445         to QImode.
25447 2018-02-16  Richard Biener  <rguenther@suse.de>
25449         PR tree-optimization/84037
25450         PR tree-optimization/84016
25451         PR target/82862
25452         * config/i386/i386.c (ix86_builtin_vectorization_cost):
25453         Adjust vec_construct for the fact we need additional higher latency
25454         128bit inserts for AVX256 and AVX512 vector builds.
25455         (ix86_add_stmt_cost): Scale vector construction cost for
25456         elementwise loads.
25458 2018-02-16  Richard Biener  <rguenther@suse.de>
25460         PR tree-optimization/84417
25461         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
25462         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
25463         (non_rewritable_lvalue_p): Likewise, use poly-ints.
25465 2018-02-16  Martin Liska  <mliska@suse.cz>
25467         PR sanitizer/84307
25468         * internal-fn.def (ASAN_CHECK): Set proper flags.
25469         (ASAN_MARK): Likewise.
25471 2018-02-16  Julia Koval  <julia.koval@intel.com>
25473         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
25474         from PTA_CANNONLAKE.
25476 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
25478         PR target/84272
25479         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
25480         Use ++iter rather than iter++ for std::list iterators.
25481         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
25482         defer deleting them until all nodes in the forest are processed.  Do
25483         free even leaf nodes.  Change to_process into auto_vec.
25485         PR bootstrap/84405
25486         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
25487         * vec.h (vec_default_construct): Use memset instead of placement new
25488         if BROKEN_VALUE_INITIALIZATION is defined.
25489         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
25490         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
25491         is defined.
25493         PR rtl-optimization/83723
25494         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
25495         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
25496         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
25497         recursive calls.
25498         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
25499         callers.
25500         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
25502 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
25504         PR rtl-optimization/81443
25505         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
25506         from inner REGs to paradoxical SUBREGs.
25508 2018-02-16  Richard Biener  <rguenther@suse.de>
25510         PR tree-optimization/84399
25511         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
25512         For operands we can analyze at their definition make sure we can
25513         analyze them at each use as well.
25515 2018-02-16  Richard Biener  <rguenther@suse.de>
25517         PR tree-optimization/84190
25518         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
25519         volatile accesses if the decl isn't volatile.
25521 2018-02-15  Jason Merrill  <jason@redhat.com>
25523         PR c++/84314 - ICE with templates and fastcall attribute.
25524         * attribs.c (build_type_attribute_qual_variant): Don't clobber
25525         TYPE_CANONICAL on an existing type.
25527 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
25529         PR tree-optimization/84383
25530         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
25531         dstoff nor call operand_equal_p if dstbase is NULL.
25533         PR tree-optimization/84334
25534         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
25535         also a CONSTANT_CLASS_P, punt.
25537 2018-02-14  Jim Wilson  <jimw@sifive.com>
25539         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
25540         first SMALL_OPERAND check.  New local min_second_step.  Move assert
25541         to where locals are set.  Add TARGET_RVC support.
25542         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
25544 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
25546         * doc/invoke.texi: Correct -Wformat-overflow code sample.
25548 2018-02-14  Martin Sebor  <msebor@redhat.com>
25550         PR tree-optimization/83698
25551         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
25552         arrays constrain the offset range to their bounds.
25553         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
25554         (builtin_access::overlap): Avoid setting the size of overlap if it's
25555         already been set.
25556         (maybe_diag_overlap): Also consider arrays when deciding what values
25557         of offsets to include in diagnostics.
25559 2018-02-14  Martin Sebor  <msebor@redhat.com>
25561         PR c/84108
25562         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
25563         that correspond to the kind of a declaration.
25565 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
25567         PR target/83984
25568         * config/pa/pa.md: Load address of PIC label using the linkage table
25569         if the label is nonlocal.
25571 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25573         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
25574         warning message if user requests -maltivec=be.
25575         * doc/invoke.texi: Document deprecation of -maltivec=be.
25577 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
25579         PR target/84220
25580         * config/rs6000/rs6000-c.c: Update definitions for
25581         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
25582         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
25584 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25586         PR target/84239
25587         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
25588         add _get_ssp intrinsics. Remove argument from
25589         __builtin_ia32_rdssp[d|q].
25590         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
25591         * config/i386/i386-builtin.def: Remove argument from
25592         __builtin_ia32_rdssp[d|q].
25593         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
25594         ix86_expand_special_args_builtin for _rdssp[d|q].
25595         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
25596         Clear register before usage.
25597         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
25598         Add documentation for new _get_ssp and _inc_ssp intrinsics.
25600 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
25602         PR tree-optimization/84357
25603         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
25604         operand 1 of an ARRAY_REF too.
25606 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
25608         PR target/83831
25609         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
25610         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25611         declarations.
25612         (set_of_reg): New struct.
25613         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
25614         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
25615         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25616         functions.
25617         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
25618         Split into bitclr, bitset, bitinvert patterns if appropriate.
25619         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
25620         use rx_fuse_in_memory_bitop.
25621         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
25622         to named insn, correct maximum insn length.
25624 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
25626         PR target/79242
25627         * machmode.def: Define a complex mode for PARTIAL_INT.
25628         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
25629         MODE_PARTIAL_INT.
25630         * doc/rtl.texi: Document CSPImode.
25631         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
25632         handling.
25633         (msp430_hard_regno_nregs_with_padding): Likewise.
25635 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
25637         PR target/84279
25638         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
25640 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
25642         PR rtl-optimization/84169
25643         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
25644         we generated a parallel as new i3 and we split that to new i2 and i3
25645         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
25646         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
25647         those to i2, not i1.  Partially rewrite this scan code.
25649 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25651         PR c/82210
25652         * stor-layout.c (place_field): For variable length fields, adjust
25653         offset_align afterwards not just based on the field's alignment,
25654         but also on the size.
25656         PR middle-end/84309
25657         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
25658         of exps and logs in the use_exp2 case.
25660 2018-02-13  Jeff Law  <law@redhat.com>
25662         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
25663         entry for "vector".
25665         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
25666         ARGS as unused.
25668 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
25670         PR debug/84342
25671         PR debug/84319
25672         * common.opt (gas-loc-support, gas-locview-support): New.
25673         (ginline-points, ginternal-reset-location-views): New.
25674         * doc/invoke.texi: Document them.  Use @itemx where intended.
25675         (gvariable-location-views): Adjust.
25676         * target.def (reset_location_view): New.
25677         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
25678         (TARGET_RESET_LOCATION_VIEW): New.
25679         * doc/tm.texi: Rebuilt.
25680         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
25681         (dwarf2out_default_as_locview_support): New.
25682         (output_asm_line_debug_info): Use option variables.
25683         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
25684         (output_loc_list): Likewise.
25685         (add_high_low_attributes): Check option variables.
25686         Don't output entry view attribute in strict mode.
25687         (gen_inlined_subroutine_die): Check option variables.
25688         (dwarf2out_inline_entry): Likewise.
25689         (init_sections_and_labels): Likewise.
25690         (dwarf2out_early_finish): Likewise.
25691         (maybe_reset_location_view): New, from...
25692         (dwarf2out_var_location): ... here.  Call it.
25693         * debug.h (dwarf2out_default_as_loc_support): Declare.
25694         (dwarf2out_default_as_locview_support): Declare.
25695         * hooks.c (hook_int_rtx_insn_0): New.
25696         * hooks.h (hook_int_rtx_insn_0): Declare.
25697         * toplev.c (process_options): Take -gas-loc-support and
25698         -gas-locview-support from dwarf2out.  Enable
25699         -gvariable-location-views by default only with locview
25700         assembler support.  Enable -ginternal-reset-location-views by
25701         default only if the target defines the corresponding hook.
25702         Enable -ginline-points by default if location views are
25703         enabled; force it disabled if statement frontiers are
25704         disabled.
25705         * tree-inline.c (expand_call_inline): Check option variables.
25706         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
25708 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
25710         PR tree-optimization/84321
25711         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
25712         handling.  Also check whether the anti-range contains any values
25713         that satisfy the mask; switch to a VR_RANGE if not.
25715 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
25717         PR sanitizer/84340
25718         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
25720 2018-02-13  Martin Jambor  <mjambor@suse.cz>
25722         PR c++/83990
25723         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
25724         of call statements, also set location of a load to a temporary.
25726 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
25728         * config/rl78/rl78.c (add_vector_labels): New function.
25729         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
25730         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
25731         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
25732         which checks that no arguments are passed.
25733         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
25734         * doc/extend.texi: Documentation for the new attribute.
25736 2018-02-13  Andreas Schwab  <schwab@suse.de>
25738         * config/riscv/linux.h (CPP_SPEC): Define.
25740 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25742         PR target/84335
25743         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
25744         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
25745         OPTION_MASK_ISA_AES as first argument to def_builtin_const
25746         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
25747         instead of OPTION_MASK_ISA_PCLMUL as first argument to
25748         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
25749         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
25750         temporarily for AES and PCLMUL builtins.
25752         PR tree-optimization/84339
25753         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
25754         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
25755         Formatting fixes.
25757         PR middle-end/84309
25758         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
25759         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
25760         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
25761         inline function.
25762         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
25763         inline function.
25764         * omp-simd-clone.h: New file.
25765         * omp-simd-clone.c: Include omp-simd-clone.h.
25766         (expand_simd_clones): No longer static.
25767         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
25768         cgraph.h and omp-simd-clone.h.
25769         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
25770         (vect_recog_widen_shift_pattern): Formatting fix.
25771         (vect_pattern_recog_1): Don't check optab for calls.
25773         PR target/84336
25774         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
25775         operands[2] into a REG before using gen_lowpart on it.
25777 2018-02-12  Jeff Law  <law@redhat.com>
25779         PR target/83760
25780         * config/sh/sh.c (find_barrier): Consider a sibling call
25781         a barrier as well.
25783         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
25784         successfully back substituting a reg.
25786 2018-02-12  Richard Biener  <rguenther@suse.de>
25788         PR tree-optimization/84037
25789         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
25790         parameter, move visited init to caller.
25791         (vect_slp_analyze_operations): Separate cost from validity
25792         check, initialize visited once for all instances.
25793         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
25794         for all instances.
25795         * tree-vect-stmts.c (vect_model_simple_cost): Make early
25796         out an assert.
25797         (vect_model_promotion_demotion_cost): Likewise.
25798         (vectorizable_bswap): Guard cost modeling with !slp_node
25799         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
25800         SLP stmts.
25801         (vectorizable_call): Likewise.
25802         (vectorizable_conversion): Likewise.
25803         (vectorizable_assignment): Likewise.
25804         (vectorizable_shift): Likewise.
25805         (vectorizable_operation): Likewise.
25806         (vectorizable_store): Likewise.
25807         (vectorizable_load): Likewise.
25808         (vectorizable_condition): Likewise.
25809         (vectorizable_comparison): Likewise.
25811 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
25813         PR sanitizer/84307
25814         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
25815         (ASAN_MARK): Fix fnspec to account for return value, change pointer
25816         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
25818 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
25820         PR middle-end/83665
25821         * params.def (inline-min-speedup): Increase from 8 to 15.
25822         (max-inline-insns-auto): Decrease from 40 to 30.
25823         * ipa-split.c (consider_split): Add some buffer for function to
25824         be considered inlining candidate.
25825         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
25826         default values.
25828 2018-02-12  Richard Biener  <rguenther@suse.de>
25830         PR tree-optimization/84037
25831         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
25832         matched stmts if we cannot swap the non-matched ones.
25834 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25836         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
25837         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
25838         _mm_maskz_scalef_round_ss): New intrinsics.
25839         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
25840         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
25841         __builtin_ia32_scalefss_round): Remove.
25842         (__builtin_ia32_scalefsd_mask_round,
25843         __builtin_ia32_scalefss_mask_round): New intrinsics.
25844         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
25845         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
25846         ((match_operand:VF_128 2 "<round_nimm_predicate>"
25847         "<round_constraint>")): Changed to ...
25848         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
25849         "<round_scalar_constraint>")): ... this.
25850         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
25851         %0, %1, %2<round_op3>}"): Changed to ...
25852         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
25853         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
25854         %2<round_scalar_mask_op3>}"): ... this.
25855         * config/i386/subst.md (round_scalar_nimm_predicate): New.
25857 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25859         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
25860         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
25861         (_mm_maskz_sqrt_round_ss): New intrinsics.
25862         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
25863         (__builtin_ia32_sqrtsd_mask_round)
25864         (__builtin_ia32_sqrtss_mask_round): New builtins.
25865         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
25866         (__builtin_ia32_sqrtss_round): Remove.
25867         (__builtin_ia32_sqrtsd_mask_round)
25868         (__builtin_ia32_sqrtss_mask_round): New builtins.
25869         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
25870         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
25871         ((match_operand:VF_128 1 "vector_operand"
25872         "xBm,<round_constraint>")): Changed to ...
25873         ((match_operand:VF_128 1 "vector_operand"
25874         "xBm,<round_scalar_constraint>")): ... this.
25875         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
25876         %0, %2, %<iptr>1<round_op3>}): Changed to ...
25877         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
25878         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
25879         %<iptr>1<round_scalar_mask_op3>}): ... this.
25880         ((set_attr "prefix" "<round_prefix>")): Changed to ...
25881         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
25883 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
25885         PR target/84266
25886         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
25887         Cast vec_cmpeq result to correct type.
25888         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
25889         Cast vec_cmpgt result to correct type.
25891 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
25893         * final.c (final_scan_insn_1): Renamed from...
25894         (final_scan_insn): ... this.  New wrapper, to recover
25895         seen from the outermost call in recursive ones.
25896         * config/sparc/sparc.c (output_return): Drop seen from call.
25897         (output_sibcall): Likewise.
25898         * config/visium/visium.c (output_branch): Likewise.
25900 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
25902         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
25903         function label.
25905 2018-02-10  Alan Modra  <amodra@gmail.com>
25907         PR target/84300
25908         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
25909         Specify LR as an input.
25911 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
25913         PR sanitizer/83987
25914         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
25915         remove_member_access_dummy_vars): New functions.
25916         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
25917         lower_omp_1, execute_lower_omp): Use them.
25919         PR rtl-optimization/84308
25920         * shrink-wrap.c (spread_components): Release todo vector.
25922 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
25924         PR rtl-optimization/57193
25925         * ira-color.c (struct allocno_color_data): Add member
25926         conflict_allocno_hard_prefs.
25927         (update_conflict_allocno_hard_prefs): New.
25928         (bucket_allocno_compare_func): Add a preference based on
25929         conflict_allocno_hard_prefs.
25930         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
25931         (color_allocnos): Remove a dead code.  Initiate
25932         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
25934 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25936         PR target/84226
25937         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
25938         constraint from =wa to wa.  Avoid a subreg on the output operand,
25939         instead use a pseudo and subreg it in a move.
25940         (p9_xxbrd_<mode>): Changed to ...
25941         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
25942         (p9_xxbrd_v2df): New expander.
25943         (p9_xxbrw_<mode>): Changed to ...
25944         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
25945         (p9_xxbrw_v4sf): New expander.
25947 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25949         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
25951 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
25953         PR target/83926
25954         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
25955         multiply in 32-bit mode.
25956         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
25957         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
25958         mode.
25960 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25962         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
25963         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
25964         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
25965         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
25967 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
25969         PR lto/84213
25970         * dwarf2out.c (is_trivial_indirect_ref): New function.
25971         (dwarf2out_late_global_decl): Do not generate a location
25972         attribute for variables that have a non-trivial DECL_VALUE_EXPR
25973         and that are not defined in the current unit.
25975 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25977         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
25978         instead of a libcall for UNORDERED.
25980 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
25982         PR target/82641
25983         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
25984         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
25986 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25988         PR target/PR84295
25989         * config/s390/s390.c (s390_set_current_function): Invoke
25990         s390_indirect_branch_settings also if fndecl didn't change.
25992 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25994         * config/rs6000/rs6000.md (blockage): Set length to zero.
25996 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25998         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
26000 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
26002         PR sanitizer/84285
26003         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
26004         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
26005         -static-lib*san.
26007         PR debug/84252
26008         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
26009         PARALLEL incoming that failed vt_get_decl_and_offset check.
26011         PR middle-end/84237
26012         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
26013         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
26014         TREE_READONLY bit.
26015         (get_variable_section): For decls in named .bss* sections pass true as
26016         second argument to bss_initializer_p.
26018 2018-02-09  Marek Polacek  <polacek@redhat.com>
26019             Jakub Jelinek  <jakub@redhat.com>
26021         PR c++/83659
26022         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
26023         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
26024         Sync some changes from cxx_fold_indirect_ref.
26026 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
26028         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
26029         markers.
26030         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
26031         (BLOCK_INLINE_ENTRY_LABEL): New.
26032         (dwarf2out_var_location): Disregard inline entry markers.
26033         (inline_entry_data): New struct.
26034         (inline_entry_data_hasher): New hashtable type.
26035         (inline_entry_data_hasher::hash): New.
26036         (inline_entry_data_hasher::equal): New.
26037         (inline_entry_data_table): New variable.
26038         (add_high_low_attributes): Add DW_AT_entry_pc and
26039         DW_AT_GNU_entry_view attributes if a pending entry is found
26040         in inline_entry_data_table.  Add old entry_pc attribute only
26041         if debug nonbinding markers are disabled.
26042         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
26043         markers are enabled.
26044         (block_within_block_p, dwarf2out_inline_entry): New.
26045         (dwarf2out_finish): Check that no entries remained in
26046         inline_entry_data_table.
26047         * final.c (reemit_insn_block_notes): Handle inline entry notes.
26048         (final_scan_insn, notice_source_line): Likewise.
26049         (rest_of_clean_state): Skip inline entry markers.
26050         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
26051         markers.
26052         * gimple.c (gimple_build_debug_inline_entry): New.
26053         * gimple.h (enum gimple_debug_subcode): Add
26054         GIMPLE_DEBUG_INLINE_ENTRY.
26055         (gimple_build_debug_inline_entry): Declare.
26056         (gimple_debug_inline_entry_p): New.
26057         (gimple_debug_nonbind_marker_p): Adjust.
26058         * insn-notes.def (INLINE_ENTRY): New.
26059         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
26060         inline entry marker notes.
26061         (print_insn): Likewise.
26062         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
26063         (INSN_DEBUG_MARKER_KIND): Likewise.
26064         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
26065         * tree-inline.c (expand_call_inline): Build and insert
26066         debug_inline_entry stmt.
26067         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
26068         inline entry blocks early, if nonbind markers are enabled.
26069         (dump_scope_block): Dump fragment info.
26070         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
26071         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
26072         (gimple_build_debug_inline_entry): New.
26073         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
26074         Enable/disable inline entry points too.
26075         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
26076         (DEBUG_INSN): Describe inline entry markers.
26078         * common.opt (gvariable-location-views): New.
26079         (gvariable-location-views=incompat5): New.
26080         * config.in: Rebuilt.
26081         * configure: Rebuilt.
26082         * configure.ac: Test assembler for view support.
26083         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
26084         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
26085         * dwarf2out.c (var_loc_view): New typedef.
26086         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
26087         (dwarf2out_locviews_in_attribute): New.
26088         (dwarf2out_locviews_in_loclist): New.
26089         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
26090         (enum dw_line_info_opcode): Add LI_adv_address.
26091         (struct dw_line_info_table): Add view.
26092         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
26093         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
26094         (zero_view_p): New variable.
26095         (ZERO_VIEW_P): New macro.
26096         (output_asm_line_debug_info): New.
26097         (struct var_loc_node): Add view.
26098         (add_AT_view_list, AT_loc_list): New.
26099         (add_var_loc_to_decl): Add view param.  Test it against last.
26100         (new_loc_list): Add view params.  Record them.
26101         (AT_loc_list_ptr): Handle loc and view lists.
26102         (view_list_to_loc_list_val_node): New.
26103         (print_dw_val): Handle dw_val_class_view_list.
26104         (size_of_die): Likewise.
26105         (value_format): Likewise.
26106         (loc_list_has_views): New.
26107         (gen_llsym): Set vl_symbol too.
26108         (maybe_gen_llsym, skip_loc_list_entry): New.
26109         (dwarf2out_maybe_output_loclist_view_pair): New.
26110         (output_loc_list): Output view list or entries too.
26111         (output_view_list_offset): New.
26112         (output_die): Handle dw_val_class_view_list.
26113         (output_dwarf_version): New.
26114         (output_compilation_unit_header): Use it.
26115         (output_skeleton_debug_sections): Likewise.
26116         (output_rnglists, output_line_info): Likewise.
26117         (output_pubnames, output_aranges): Update version comments.
26118         (output_one_line_info_table): Output view numbers in asm comments.
26119         (dw_loc_list): Determine current endview, pass it to new_loc_list.
26120         Call maybe_gen_llsym.
26121         (loc_list_from_tree_1): Adjust.
26122         (add_AT_location_description): Create view list attribute if
26123         needed, check it's absent otherwise.
26124         (convert_cfa_to_fb_loc_list): Adjust.
26125         (maybe_emit_file): Call output_asm_line_debug_info for test.
26126         (dwarf2out_var_location): Reset views as needed.  Precompute
26127         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
26128         attribute.  Set view.
26129         (new_line_info_table): Reset next view.
26130         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
26131         (dwarf2out_source_line): Likewise.  Output view resets and labels to
26132         the assembler, or select appropriate line info opcodes.
26133         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
26134         (optimize_string_length): Catch it.  Adjust.
26135         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
26136         dw_val_class_view_list, and remove it if no longer needed.
26137         (hash_loc_list): Hash view numbers.
26138         (loc_list_hasher::equal): Compare them.
26139         (optimize_location_lists): Check whether a view list symbol is
26140         needed, and whether the locview attribute is present, and
26141         whether they match.  Remove the locview attribute if no longer
26142         needed.
26143         (index_location_lists): Call skip_loc_list_entry for test.
26144         (dwarf2out_finish): Call output_asm_line_debug_info for test.
26145         Use output_dwarf_version.
26146         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
26147         (struct dw_val_node): Add val_view_list.
26148         * final.c (SEEN_NEXT_VIEW): New.
26149         (set_next_view_needed): New.
26150         (clear_next_view_needed): New.
26151         (maybe_output_next_view): New.
26152         (final_start_function): Rename to...
26153         (final_start_function_1): ... this.  Take pointer to FIRST,
26154         add SEEN parameter.  Emit param bindings in the initial view.
26155         (final_start_function): Reintroduce SEEN-less interface.
26156         (final): Rename to...
26157         (final_1): ... this.  Take SEEN parameter.  Output final pending
26158         next view at the end.
26159         (final): Reintroduce seen-less interface.
26160         (final_scan_insn): Output pending next view before switching
26161         sections or ending a block.  Mark the next view as needed when
26162         outputting variable locations.  Notify debug backend of section
26163         changes, and of location view changes.
26164         (rest_of_handle_final): Adjust.
26165         * toplev.c (process_options): Autodetect value for debug variable
26166         location views option.  Warn on incompat5 without -gdwarf-5.
26167         * doc/invoke.texi (gvariable-location-views): New.
26168         (gvariable-location-views=incompat5): New.
26169         (gno-variable-location-views): New.
26171 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
26173         PR tree-optimization/84136
26174         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
26175         that the result of find_edge is non-NULL.
26177 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
26179         PR target/83008
26180         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
26181         storing integer register in SImode.  Fix cost of 256 and 512
26182         byte aligned SSE register store.
26184 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
26186         * config/i386/i386.c (ix86_multiplication_cost): Fix
26187         multiplication cost for TARGET_AVX512DQ.
26189 2018-02-08  Marek Polacek  <polacek@redhat.com>
26191         PR tree-optimization/84238
26192         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
26193         get_range_strlen.
26195 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26197         PR tree-optimization/84265
26198         * tree-vect-stmts.c (vectorizable_store): Don't treat
26199         VMAT_CONTIGUOUS accesses as grouped.
26200         (vectorizable_load): Likewise.
26202 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26204         PR tree-optimization/81635
26205         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
26206         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
26207         (test_round_for_mask): New functions.
26208         (wide_int_cc_tests): Call test_round_for_mask.
26209         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
26210         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
26211         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
26212         range returned by get_range_info.
26214 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
26216         PR ipa/81360
26217         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
26218         * symtab.c: Include builtins.h
26219         (symtab_node::output_to_lto_symbol_table_p): Move here
26220         from lto-streamer-out.c:output_symbol_p.
26221         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
26222         (output_symbol_p): Move all logic to symtab.c
26223         (produce_symtab): Update.
26225 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26227         * config/s390/s390-opts.h (enum indirect_branch): Define.
26228         * config/s390/s390-protos.h (s390_return_addr_from_memory)
26229         (s390_indirect_branch_via_thunk)
26230         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
26231         (enum s390_indirect_branch_type): Define.
26232         * config/s390/s390.c (struct s390_frame_layout, struct
26233         machine_function): Remove.
26234         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
26235         (indirect_branch_table_label_no, indirect_branch_table_name):
26236         Define variables.
26237         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
26238         (enum s390_indirect_branch_option): Define.
26239         (s390_return_addr_from_memory): New function.
26240         (s390_handle_string_attribute): New function.
26241         (s390_attribute_table): Add new attribute handler.
26242         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
26243         (s390_indirect_branch_via_thunk): New function.
26244         (s390_indirect_branch_via_inline_thunk): New function.
26245         (s390_function_ok_for_sibcall): When jumping via thunk disallow
26246         sibling call optimization for non z10 compiles.
26247         (s390_emit_call): Force indirect branch target to be a single
26248         register.  Add r1 clobber for non-z10 compiles.
26249         (s390_emit_epilogue): Emit return jump via return_use expander.
26250         (s390_reorg): Handle JUMP_INSNs as execute targets.
26251         (s390_option_override_internal): Perform validity checks for the
26252         new command line options.
26253         (s390_indirect_branch_attrvalue): New function.
26254         (s390_indirect_branch_settings): New function.
26255         (s390_set_current_function): Invoke s390_indirect_branch_settings.
26256         (s390_output_indirect_thunk_function):  New function.
26257         (s390_code_end): Implement target hook.
26258         (s390_case_values_threshold): Implement target hook.
26259         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
26260         macros.
26261         * config/s390/s390.h (struct s390_frame_layout)
26262         (struct machine_function): Move here from s390.c.
26263         (TARGET_INDIRECT_BRANCH_NOBP_RET)
26264         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
26265         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
26266         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
26267         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
26268         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
26269         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
26270         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
26271         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
26272         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
26273         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
26274         (mnemonic attribute): Add values which aren't recognized
26275         automatically.
26276         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
26277         pattern for branch conversion.  Fix mnemonic attribute.
26278         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
26279         indirect branch via thunk if requested.
26280         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
26281         ("*indirect_jump"): Disable for branch conversion using out of
26282         line thunks.
26283         ("indirect_jump_via_thunk<mode>_z10")
26284         ("indirect_jump_via_thunk<mode>")
26285         ("indirect_jump_via_inlinethunk<mode>_z10")
26286         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
26287         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
26288         ("casesi_jump_via_inlinethunk<mode>_z10")
26289         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
26290         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
26291         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
26292         ("*indirect2_jump"): Disable for branch conversion.
26293         ("casesi_jump"): Turn into expander and expand patterns for branch
26294         conversion.
26295         ("return_use"): New expander.
26296         ("*return"): Emit return via thunk and rename it to ...
26297         ("*return<mode>"): ... this one.
26298         * config/s390/s390.opt: Add new options and and enum for the
26299         option values.
26301 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26303         * lra-constraints.c (match_reload): Unconditionally use
26304         gen_lowpart_SUBREG, rather than selecting between that
26305         and equivalent gen_rtx_SUBREG code.
26307 2018-02-08  Richard Biener  <rguenther@suse.de>
26309         PR tree-optimization/84233
26310         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
26311         changed flag instead of boguously re-using phi_inserted.
26313 2018-02-08  Martin Jambor  <mjambor@suse.cz>
26315         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
26316         static local variables.
26318 2018-02-08  Richard Biener  <rguenther@suse.de>
26320         PR tree-optimization/84278
26321         * tree-vect-stmts.c (vectorizable_store): When looking for
26322         smaller vector types to perform grouped strided loads/stores
26323         make sure the mode is supported by the target.
26324         (vectorizable_load): Likewise.
26326 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26328         * config/aarch64/aarch64.c (aarch64_components_for_bb):
26329         Increase LDP/STP opportunities by adding adjacent callee-saves.
26331 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26333         PR rtl-optimization/84068
26334         PR rtl-optimization/83459
26335         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
26337 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
26339         PR tree-optimization/84224
26340         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
26341         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
26342         non-zero arguments.
26344 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
26346         PR target/84113
26347         * config/rs6000/altivec.md (*restore_world): Remove LR use.
26348         * config/rs6000/predicates.md (restore_world_operation): Adjust op
26349         count, remove one USE.
26351 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
26353         * doc/install.texi (Configuration): Document the
26354         --with-long-double-format={ibm,ieee} PowerPC configuration
26355         options.
26357         PR target/84154
26358         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
26359         Convert from define_expand to be define_insn_and_split.  Rework
26360         float/double/_Float128 conversions to QI/HI/SImode to work with
26361         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
26362         conversions to QI/HImode types did a store and then a load to
26363         truncate the value.  For conversions to VSX registers, don't split
26364         the insn, instead emit the code directly.  Use the code iterator
26365         any_fix to combine signed and unsigned conversions.
26366         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
26367         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
26368         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26369         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
26370         (fix_<mode>di2_hw): Likewise.
26371         (fixuns_<mode>di2_hw): Likewise.
26372         (fix_<mode>si2_hw): Likewise.
26373         (fixuns_<mode>si2_hw): Likewise.
26374         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
26375         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26376         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
26377         fix<uns>_trunc<SFDF:mode>si2_p8.
26378         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
26379         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
26380         (fix<uns>_<mode>_mem): Likewise.
26381         (fctiw<u>z_<mode>_mem): Likewise.
26382         (fix<uns>_<mode>_mem): Likewise.
26383         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
26384         the register allocator from doing a direct move to the GPRs to do
26385         a store, and instead use the ISA 3.0 store byte/half-word from
26386         vector register instruction.  For IEEE 128-bit floating point,
26387         also optimize stores of 32-bit ints.
26388         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
26390 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
26392         * genextract.c (push_pathstr_operand): New function to support
26393         [a-zA-Z].
26394         (walk_rtx): Call push_pathstr_operand.
26395         (print_path): Support [a-zA-Z].
26397 2018-02-07  Richard Biener  <rguenther@suse.de>
26399         PR tree-optimization/84037
26400         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
26401         (cse_and_gimplify_to_preheader): Declare.
26402         (vect_get_place_in_interleaving_chain): Likewise.
26403         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26404         ivexpr_map.
26405         (_loop_vec_info::~_loop_vec_info): Delete it.
26406         (cse_and_gimplify_to_preheader): New function.
26407         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
26408         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
26409         (vectorizable_load): Likewise.  For grouped stores always base
26410         the IV on the first element.
26411         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
26412         condition before gimplifying.
26414 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
26416         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
26417         *DIV_EXPR and *MOD_EXPR.
26419 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
26421         PR target/84248
26422         * config/i386/i386.c (ix86_option_override_internal): Mask out
26423         the CF_SET bit when checking -fcf-protection.
26425 2018-02-07  Tom de Vries  <tom@codesourcery.com>
26427         PR libgomp/84217
26428         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
26429         enough.
26431 2018-02-07  Richard Biener  <rguenther@suse.de>
26433         PR tree-optimization/84204
26434         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
26435         this place.
26437         PR tree-optimization/84205
26438         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
26439         special-case isl_ast_op_zdiv_r.
26441         PR tree-optimization/84223
26442         * graphite-scop-detection.c (gather_bbs::before_dom_children):
26443         Only add conditions from within the region.
26444         (gather_bbs::after_dom_children): Adjust.
26446 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
26448         PR target/84209
26449         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
26450         * config/avr/avr.md: Only post-reload split REG-REG moves if
26451         either register is GENERAL_REG_P.
26453 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
26455         PR tree-optimization/84235
26456         * tree-ssa-scopedtables.c
26457         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
26458         if the subtraction is performed in floating point type where NaNs are
26459         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
26460         build 1.  Formatting fix.
26462 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
26464         PR target/84146
26465         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
26466         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
26467         and skip it regardless of bb boundaries.  Use CALL_P macro,
26468         don't test INSN_P (insn) together with CALL_P or JUMP_P check
26469         unnecessarily, formatting fix.
26471 2018-02-06  Michael Collison  <michael.collison@arm.com>
26473         * config/arm/thumb2.md:
26474         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
26475         (*thumb_mov_notscc): Ditto.
26477 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
26479         PR target/84154
26480         * config/rs6000/rs6000.md (su code attribute): Use "u" for
26481         unsigned_fix, not "s".
26483 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26485         * configure.ac (gcc_fn_eh_frame_ro): New function.
26486         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
26487         correct .eh_frame permissions.
26488         * configure: Regenerate.
26490 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
26492         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
26493         irrelevant options.
26495 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26497         * config/rs6000/rs6000.c (rs6000_option_override_internal):
26498         Display warning message for -mno-speculate-indirect-jumps.
26500 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
26502         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
26503         Undocumented.
26504         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
26506 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
26508         PR tree-optimization/84225
26509         * tree-eh.c (find_trapping_overflow): Only call
26510         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
26512 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26514         PR target/84145
26515         * config/i386/i386.c: Reimplement the check of possible options
26516         -mibt/-mshstk conbination. Change error messages.
26517         * doc/invoke.texi: Fix a typo: remove extra '='.
26519 2018-02-06  Marek Polacek  <polacek@redhat.com>
26521         PR tree-optimization/84228
26522         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
26524 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
26526         PR target/82641
26527         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
26528         emitted arch directives.
26529         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
26530         __ARM_FEATURE_COPROC before changing architectures.
26532 2018-02-06  Richard Biener  <rguenther@suse.de>
26534         * config/i386/i386.c (print_reg): Fix typo.
26535         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
26537 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
26539         * configure: Regenerate.
26541 2018-02-05  Martin Sebor  <msebor@redhat.com>
26543         PR tree-optimization/83369
26544         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
26545         inlining context.
26547 2018-02-05  Martin Liska  <mliska@suse.cz>
26549         * doc/invoke.texi: Cherry-pick upstream r323995.
26551 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
26553         * ira.c (ira_init_register_move_cost): Adjust comment.
26555 2018-02-05  Martin Liska  <mliska@suse.cz>
26557         PR gcov-profile/84137
26558         * doc/gcov.texi: Fix typo in documentation.
26560 2018-02-05  Martin Liska  <mliska@suse.cz>
26562         PR gcov-profile/83879
26563         * doc/gcov.texi: Document necessity of --dynamic-list-data when
26564         using dlopen functionality.
26566 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
26568         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
26569         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
26570         _mm_maskz_range_ss, _mm_mask_range_round_ss,
26571         _mm_maskz_range_round_ss): New intrinsics.
26572         (__builtin_ia32_rangesd128_round)
26573         (__builtin_ia32_rangess128_round): Remove.
26574         (__builtin_ia32_rangesd128_mask_round,
26575         __builtin_ia32_rangess128_mask_round): New builtins.
26576         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
26577         __builtin_ia32_rangess128_round): Remove.
26578         (__builtin_ia32_rangesd128_mask_round,
26579         __builtin_ia32_rangess128_mask_round): New builtins.
26580         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
26581         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
26582         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
26583         "<round_saeonly_constraint>")): Changed to ...
26584         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
26585         "<round_saeonly_scalar_constraint>")): ... this.
26586         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
26587         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
26588         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
26589         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
26590         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
26592 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
26594         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
26595         options.
26596         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
26597         Remove all values except native, 8540 and 8548.
26599 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
26601         * config/i386/i386.c (ix86_output_function_return): Pass
26602         INVALID_REGNUM, instead of -1, as invalid register number to
26603         indirect_thunk_name and output_indirect_thunk.
26605 2018-02-02  Julia Koval  <julia.koval@intel.com>
26607         * config.gcc: Add -march=icelake.
26608         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
26609         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
26610         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
26611         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
26612         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
26613         (processor_target_table): Add icelake.
26614         (ix86_option_override_internal): Handle new PTAs.
26615         (get_builtin_code_for_version): Handle icelake.
26616         (M_INTEL_COREI7_ICELAKE): New.
26617         (fold_builtin_cpu): Handle icelake.
26618         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
26619         * doc/invoke.texi: Add -march=icelake.
26621 2018-02-02  Julia Koval  <julia.koval@intel.com>
26623         * config/i386/i386.c (ix86_option_override_internal): Change flags type
26624         to wide_int_bitmask.
26625         * wide-int-bitmask.h: New.
26627 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26629         PR target/84066
26630         * config/i386/i386.md: Replace Pmode with word_mode in
26631         builtin_setjmp_setup and builtin_longjmp to support x32.
26633 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
26635         PR target/56010
26636         PR target/83743
26637         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
26638         #include "opts.h".
26639         (rs6000_supported_cpu_names): New static variable.
26640         (linux_cpu_translation_table): Likewise.
26641         (elf_platform) <cpu>: Define new static variable and use it.
26642         Translate kernel AT_PLATFORM name to canonical name if needed.
26643         Error if platform name is unknown.
26645 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
26647         PR target/84089
26648         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
26650 2018-02-01  Jeff Law  <law@redhat.com>
26652         PR target/84128
26653         * config/i386/i386.c (release_scratch_register_on_entry): Add new
26654         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
26655         the scratch if RELEASE_VIA_POP is false.
26656         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
26657         If we have to save a temporary register, decrement SIZE appropriately.
26658         Pass new arguments to release_scratch_register_on_entry.
26659         (ix86_adjust_stack_and_probe): Likewise.
26660         (ix86_emit_probe_stack_range): Pass new arguments to
26661         release_scratch_register_on_entry.
26663 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
26665         PR rtl-optimization/84157
26666         * combine.c (change_zero_ext): Use REG_P predicate in
26667         front of HARD_REGISTER_P predicate.
26669 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
26671         * config/avr/avr.c (avr_option_override): Move disabling of
26672         -fdelete-null-pointer-checks to...
26673         * common/config/avr/avr-common.c (avr_option_optimization_table):
26674         ...here.
26676 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26678         PR tree-optimization/81635
26679         * tree-data-ref.c (split_constant_offset_1): For types that
26680         wrap on overflow, try to use range info to prove that wrapping
26681         cannot occur.
26683 2018-02-01  Renlin Li  <renlin.li@arm.com>
26685         PR target/83370
26686         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
26687         TAILCALL_ADDR_REGS.
26688         (aarch64_register_move_cost): Likewise.
26689         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
26690         TAILCALL_ADDR_REGS.
26691         (REG_CLASS_NAMES): Likewise.
26692         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
26693         TAILCALL_ADDR_REGS. Remove IP registers.
26694         * config/aarch64/aarch64.md (Ucs): Update register constraint.
26696 2018-02-01  Richard Biener  <rguenther@suse.de>
26698         * domwalk.h (dom_walker::dom_walker): Add additional constructor
26699         for specifying RPO order and allow NULL for that.
26700         * domwalk.c (dom_walker::dom_walker): Likewise.
26701         (dom_walker::walk): Handle NULL RPO order.
26702         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
26703         in RPO order.
26704         (rewrite_update_dom_walker): Likewise.
26705         (mark_def_dom_walker): Likewise.
26707 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26709         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
26710         (aarch64_maybe_expand_sve_subreg_move): Declare.
26711         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
26712         * config/aarch64/predicates.md (aarch64_any_register_operand): New
26713         predicate.
26714         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
26715         that are semantically a reverse operation.
26716         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
26717         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
26718         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
26719         functions.
26720         (aarch64_can_change_mode_class): For big-endian, forbid changes
26721         between two SVE modes if they have different element sizes.
26723 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26725         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
26726         the TImode handling for big-endian targets.
26728 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26730         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
26731         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
26732         not just bytes.
26733         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
26734         Remove BSWAP handing for big-endian targets and use the form of
26735         LD1RQ appropariate for the mode.
26737 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26739         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
26740         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
26741         duplicated element.
26743 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26745         PR tearget/83845
26746         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
26747         check for operands that need to go through aarch64_sve_reload_be.
26749 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
26751         PR tree-optimization/81661
26752         PR tree-optimization/84117
26753         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
26754         * tree-eh.c: Include gimplify.h.
26755         (find_trapping_overflow, replace_trapping_overflow,
26756         rewrite_to_non_trapping_overflow): New functions.
26757         * tree-vect-loop.c: Include tree-eh.h.
26758         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
26759         * tree-data-ref.c: Include tree-eh.h.
26760         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
26762 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
26764         PR rtl-optimization/84123
26765         * combine.c (change_zero_ext): Check if hard register satisfies
26766         can_change_dest_mode before calling gen_lowpart_SUBREG.
26768 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
26770         PR target/82444
26771         * ira.c (ira_init_register_move_cost): Remove assert.
26773 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26775         PR rtl-optimization/84071
26776         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
26777         * doc/tm.texi: Regenerate.
26779 2018-01-31  Richard Biener  <rguenther@suse.de>
26781         PR tree-optimization/84132
26782         * tree-data-ref.c (analyze_miv_subscript): Properly
26783         check whether evolution_function_is_affine_multivariate_p
26784         before calling gcd_of_steps_may_divide_p.
26786 2018-01-31  Julia Koval  <julia.koval@intel.com>
26788         PR target/83618
26789         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
26790         * config/i386/i386.md (rdpid_rex64) New.
26791         (rdpid): Make 32bit only.
26793 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
26795         PR lto/84105
26796         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
26797         an IDENTIFIER_NODE for FUNCTION_TYPE's.
26799 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26801         Revert
26802         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26804         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26806 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26808         PR rtl-optimization/84071
26809         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
26810         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
26812 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26814         * config/arc/arc.c (arc_handle_aux_attribute): New function.
26815         (arc_attribute_table): Add 'aux' attribute.
26816         (arc_in_small_data_p): Consider aux like variables.
26817         (arc_is_aux_reg_p): New function.
26818         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
26819         (arc_get_aux_arg): New function.
26820         (prepare_move_operands): Handle aux-register access.
26821         (arc_handle_aux_attribute): New function.
26822         * doc/extend.texi (ARC Variable attributes): Add subsection.
26824 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26826         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
26827         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
26828         (arc_attribute_table): Add 'uncached' attribute.
26829         (arc_print_operand): Print '.di' flag for uncached memory
26830         accesses.
26831         (arc_in_small_data_p): Do not consider for small data the uncached
26832         types.
26833         (arc_is_uncached_mem_p): New function.
26834         * config/arc/predicates.md (compact_store_memory_operand): Check
26835         for uncached memory accesses.
26836         (nonvol_nonimm_operand): Likewise.
26837         * doc/extend.texi (ARC Type Attribute): New subsection.
26839 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
26841         PR c/84100
26842         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
26843         falign-loops=): Add Optimization flag.
26845 2018-01-30  Jeff Law  <law@redhat.com>
26847         PR target/84064
26848         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
26849         INT_REGISTERS_SAVED.  Check it prior to calling
26850         get_scratch_register_on_entry.
26851         (ix86_adjust_stack_and_probe): Similarly.
26852         (ix86_emit_probe_stack_range): Similarly.
26853         (ix86_expand_prologue): Corresponding changes.
26855 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26857         PR target/40411
26858         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
26859         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
26861 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
26863         PR target/84112
26864         * lra-constraints.c (curr_insn_transform): Process AND in the
26865         address.
26867 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
26869         PR rtl-optimization/83986
26870         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
26871         dependence against last_pending_memory_flush in addition to
26872         pending_jump_insns.
26874 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
26876         PR tree-optimization/81611
26877         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
26878         copies.
26880 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26882         PR target/83758
26883         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
26884         a reg rtx.
26886 2018-01-30  Richard Biener  <rguenther@suse.de>
26887             Jakub Jelinek  <jakub@redhat.com>
26889         PR tree-optimization/84111
26890         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
26891         inner loops added during recursion, as they don't have up-to-date
26892         SSA form.
26894 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26896         PR ipa/81360
26897         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
26898         (can_inline_edge_by_limits_p): ... here.
26899         (can_early_inline_edge_p, check_callers,
26900         update_caller_keys, update_callee_keys, recursive_inlining,
26901         add_new_edges_to_heap, speculation_useful_p,
26902         inline_small_functions,
26903         inline_small_functions, flatten_function,
26904         inline_to_all_callers_1): Update.
26906 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26908         * profile-count.c (profile_count::combine_with_ipa_count): Handle
26909         zeros correctly.
26911 2018-01-30  Richard Biener  <rguenther@suse.de>
26913         PR tree-optimization/83008
26914         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
26915         invariant and constant vector uses in stmts when they need
26916         more than one stmt.
26918 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26920         PR bootstrap/84017
26921         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
26922         * configure: Regenerate.
26924 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26926         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
26927         pattern.
26928         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
26929         Use gen_rtx_REG rather than gen_lowpart.
26931 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26933         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
26934         rather than 0 when creating partial subregs.
26936 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26938         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
26939         of usage.
26941 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
26943         PR target/81550
26944         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
26945         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
26946         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
26947         flags.  This restores the settings used before the 2017-07-24.
26948         Turning off pre increment/decrement/modify allows IVOPTS to
26949         optimize DF/SF loops where the index is an int.
26951 2018-01-29  Richard Biener  <rguenther@suse.de>
26952             Kelvin Nilsen  <kelvin@gcc.gnu.org>
26954         PR bootstrap/80867
26955         * tree-vect-stmts.c (vectorizable_call): Don't call
26956         targetm.vectorize_builtin_md_vectorized_function if callee is
26957         NULL.
26959 2018-01-22  Carl Love  <cel@us.ibm.com>
26961         * doc/extend.tex: Fix typo in second arg in
26962         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
26964 2018-01-29  Richard Biener  <rguenther@suse.de>
26966         PR tree-optimization/84086
26967         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
26968         (flush_ssaname_freelist): When SSA names were released reset
26969         the SCEV hash table.
26971 2018-01-29  Richard Biener  <rguenther@suse.de>
26973         PR tree-optimization/84057
26974         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
26975         removed paths when removing edges.
26977 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
26979         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
26980         -mfunction-return=@var{choice}.
26982 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
26984         PR diagnostic/84034
26985         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
26986         Handle CR like TAB.
26987         (layout::print_source_line): Likewise.
26988         (test_get_line_width_without_trailing_whitespace): Add test cases.
26990 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
26992         PR middle-end/84040
26993         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
26994         debug insns.
26996 2018-01-26  Jim Wilson  <jimw@sifive.com>
26998         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
27000         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
27001         specified.
27003 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27005         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
27006         and CMP + SUB-immediate -> SUBS.
27008 2018-01-26  Martin Sebor  <msebor@redhat.com>
27010         PR tree-optimization/83896
27011         * tree-ssa-strlen.c (get_string_len): Rename...
27012         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
27013         Avoid assuming length is constant.
27014         (handle_char_store): Use HOST_WIDE_INT for string length.
27016 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
27018         PR target/81763
27019         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
27020         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
27022 2018-01-26  Richard Biener  <rguenther@suse.de>
27024         PR rtl-optimization/84003
27025         * dse.c (record_store): Only record redundant stores when
27026         the earlier store aliases at least all accesses the later one does.
27028 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
27030         PR rtl-optimization/83985
27031         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
27032         REG_CFA_RESTORE insns.
27033         (delete_unmarked_insns): Don't ignore separate shrink wrapping
27034         REG_CFA_RESTORE insns here.
27036         PR c/83989
27037         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
27038         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
27040 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27042         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
27043         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
27044         (arc_init): Likewise.
27045         (arc_override_options): Likewise.
27046         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
27047         value.
27048         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
27049         support.
27050         * config/arc/arc.h (TARGET_DBNZ): Define.
27051         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
27052         properly set the tune attribute.
27053         (dbnz): Use TARGET_DBNZ guard.
27054         * config/arc/arc.opt (mtune): Add core3 option.
27056 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27058         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
27059         recognize new pic like addresses.
27060         (arc_delegitimize_address): Clean up.
27062 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27064         * config/arc/arc-arches.def: Option mrf16 valid for all
27065         architectures.
27066         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
27067         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
27068         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
27069         * config/arc/arc-tables.opt: Regenerate.
27070         * config/arc/arc.c (arc_conditional_register_usage): Handle
27071         reduced register file case.
27072         (arc_file_start): Set must have build attributes.
27073         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
27074         mrf16 option value.
27075         * config/arc/arc.opt (mrf16): Add new option.
27076         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
27077         * config/arc/genmultilib.awk: Handle new mrf16 option.
27078         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
27079         * config/arc/t-multilib: Regenerate.
27080         * doc/invoke.texi (ARC Options): Document mrf16 option.
27082 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27084         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
27085         * config/arc/arc.c (arc_handle_secure_attribute): New function.
27086         (arc_attribute_table): Add 'secure_call' attribute.
27087         (arc_print_operand): Print secure call operand.
27088         (arc_function_ok_for_sibcall): Don't optimize tail calls when
27089         secure.
27090         (arc_is_secure_call_p): New function.  * config/arc/arc.md
27091         (call_i): Add support for sjli instruction.
27092         (call_value_i): Likewise.
27093         * config/arc/constraints.md (Csc): New constraint.
27095 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27096             John Eric Martin  <John.Martin@emmicro-us.com>
27098         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
27099         * config/arc/arc.c (_arc_jli_section): New struct.
27100         (arc_jli_section): New type.
27101         (rc_jli_sections): New static variable.
27102         (arc_handle_jli_attribute): New function.
27103         (arc_attribute_table): Add jli_always and jli_fixed attribute.
27104         (arc_file_end): New function.
27105         (TARGET_ASM_FILE_END): Define.
27106         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
27107         (arc_add_jli_section): New function.
27108         (jli_call_scan): Likewise.
27109         (arc_reorg): Call jli_call_scan.
27110         (arc_output_addsi): Remove 'S' from printing asm operand.
27111         (arc_is_jli_call_p): New function.
27112         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
27113         operand.
27114         (movhi_insn): Likewise.
27115         (movsi_insn): Likewise.
27116         (movsi_set_cc_insn): Likewise.
27117         (loadqi_update): Likewise.
27118         (load_zeroextendqisi_update): Likewise.
27119         (load_signextendqisi_update): Likewise.
27120         (loadhi_update): Likewise.
27121         (load_zeroextendhisi_update): Likewise.
27122         (load_signextendhisi_update): Likewise.
27123         (loadsi_update): Likewise.
27124         (loadsf_update): Likewise.
27125         (movsicc_insn): Likewise.
27126         (bset_insn): Likewise.
27127         (bxor_insn): Likewise.
27128         (bclr_insn): Likewise.
27129         (bmsk_insn): Likewise.
27130         (bicsi3_insn): Likewise.
27131         (cmpsi_cc_c_insn): Likewise.
27132         (movsi_ne): Likewise.
27133         (movsi_cond_exec): Likewise.
27134         (clrsbsi2): Likewise.
27135         (norm_f): Likewise.
27136         (normw): Likewise.
27137         (swap): Likewise.
27138         (divaw): Likewise.
27139         (flag): Likewise.
27140         (sr): Likewise.
27141         (kflag): Likewise.
27142         (ffs): Likewise.
27143         (ffs_f): Likewise.
27144         (fls): Likewise.
27145         (call_i): Remove 'S' asm letter, add jli instruction.
27146         (call_value_i): Likewise.
27147         * config/arc/arc.op (mjli-always): New option.
27148         * config/arc/constraints.md (Cji): New constraint.
27149         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
27150         operand.
27151         (subsf3_fpx): Likewise.
27152         (mulsf3_fpx): Likewise.
27153         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
27154         asm operand.
27155         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
27156         function attrbutes.
27157         * doc/invoke.texi (ARC): Document mjli-always option.
27159 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
27161         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
27162         avoid addition with 0 and use incw and decw where possible.
27164 2018-01-26  Richard Biener  <rguenther@suse.de>
27166         PR tree-optimization/81082
27167         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
27168         association if it requires casting to unsigned.
27169         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
27170         from fold_plusminus_mult_expr to catch important cases late when
27171         range info is available.
27173 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27175         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
27176         * configure.ac (hidden_linkonce): New test.
27177         * configure: Regenerate.
27178         * config.in: Regenerate.
27180 2018-01-26  Julia Koval  <julia.koval@intel.com>
27182         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
27183         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
27184         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
27185         _mm_mask_bitshuffle_epi64_mask): Fix type.
27186         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
27187         USI_FTYPE_V4DI_V4DI_USI): Remove.
27188         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
27189         __builtin_ia32_vpshufbitqmb256_mask,
27190         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
27191         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
27192         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
27194 2018-01-26  Alan Modra  <amodra@gmail.com>
27196         PR target/84033
27197         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
27198         UNSPEC_VBPERMQ.  Sort other unspecs.
27200 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
27202         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
27204 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
27206         PR middle-end/83055
27207         * predict.c (drop_profile): Do not push/pop cfun; update also
27208         node->count.
27209         (handle_missing_profiles): Fix logic looking for zero profiles.
27211 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
27213         PR middle-end/83977
27214         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
27215         on functions with #pragma omp declare simd or functions with simd
27216         attribute.
27217         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
27218         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
27219         Remove trailing \n from warning_at calls.
27221 2018-01-25  Tom de Vries  <tom@codesourcery.com>
27223         PR target/84028
27224         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27225         for neutered workers.
27227 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
27229         PR target/68467
27230         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
27231         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
27233 2018-01-24  Jeff Law  <law@redhat.com>
27235         PR target/83994
27236         * i386.c (get_probe_interval): Move to earlier point.
27237         (ix86_compute_frame_layout): If -fstack-clash-protection and
27238         the frame is larger than the probe interval, then use pushes
27239         to save registers rather than reg->mem moves.
27240         (ix86_expand_prologue): Remove conditional for int_registers_saved
27241         assertion.
27243 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
27245         PR target/84014
27246         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
27247         min/max for never referenced object.
27249 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
27251         PR middle-end/83977
27252         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
27253         here.
27254         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
27255         attributes from DECL_ATTRIBUTES (decl) without affecting
27256         DECL_ATTRIBUTES (current_function_decl).
27257         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
27258         functions with non-NULL DECL_ABSTRACT_ORIGIN.
27260 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
27262         PR tree-optimization/83979
27263         * fold-const.c (fold_comparison): Use constant_boolean_node
27264         instead of boolean_{true,false}_node.
27266 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
27268         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
27269         with zero counts.
27271 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27273         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27274         Simplify the clause that sets the length attribute.
27275         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27276         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
27277         clause that sets the length attribute.
27278         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27280 2018-01-24  Tom de Vries  <tom@codesourcery.com>
27282         PR target/83589
27283         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
27284         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
27285         Add strict parameter.
27286         (prevent_branch_around_nothing): Insert dummy insn between branch to
27287         label and label with no ptx insn inbetween.
27288         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
27290 2018-01-24  Tom de Vries  <tom@codesourcery.com>
27292         PR target/81352
27293         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27294         for neutered threads in warp.
27295         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
27297 2018-01-24  Richard Biener  <rguenther@suse.de>
27299         PR tree-optimization/83176
27300         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
27301         operands.
27303 2018-01-24  Richard Biener  <rguenther@suse.de>
27305         PR tree-optimization/82819
27306         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
27307         code generating pluses that are no-ops in the target precision.
27309 2018-01-24  Richard Biener  <rguenther@suse.de>
27311         PR middle-end/84000
27312         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
27314 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27316         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
27317         to merge probabilities.
27318         * predict.c (probably_never_executed): Also mark as cold functions
27319         with global 0 profile and guessed local profile.
27320         * profile-count.c (profile_probability::combine_with_count): New
27321         member function.
27322         * profile-count.h (profile_probability::operator*,
27323         profile_probability::operator*=, profile_probability::operator/,
27324         profile_probability::operator/=): Reduce precision to adjusted
27325         and set value to guessed on contradictory divisions.
27326         (profile_probability::combine_with_freq): Remove.
27327         (profile_probability::combine_wiht_count): Declare.
27328         (profile_count::force_nonzero):: Set to adjusted.
27329         (profile_count::probability_in):: Set quality to adjusted.
27330         * tree-ssa-tail-merge.c (replace_block_by): Use
27331         combine_with_count.
27333 2018-01-23  Andrew Waterman  <andrew@sifive.com>
27334             Jim Wilson  <jimw@sifive.com>
27336         * config/riscv/riscv.c (riscv_stack_boundary): New.
27337         (riscv_option_override): Set riscv_stack_boundary.  Handle
27338         riscv_preferred_stack_boundary_arg.
27339         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
27340         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
27341         (STACK_BOUNDARY): Set to riscv_stack_boundary.
27342         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
27343         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
27344         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
27346 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
27348         PR target/83905
27349         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
27350         of struct ix86_frame.
27351         (ix86_expand_epilogue): Likewise.  Add a local variable for
27352         the reg_save_offset field in struct ix86_frame.
27354 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
27356         PR tree-optimization/82604
27357         * tree-loop-distribution.c (enum partition_kind): New enum item
27358         PKIND_PARTIAL_MEMSET.
27359         (partition_builtin_p): Support above new enum item.
27360         (generate_code_for_partition): Ditto.
27361         (compute_access_range): Differentiate cases that equality can be
27362         proven at all loops, the innermost loops or no loops.
27363         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
27364         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
27365         (finalize_partitions, distribute_loop): Don't fuse partition of
27366         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
27367         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
27368         parloop is enabled.
27370 2018-01-23  Martin Liska  <mliska@suse.cz>
27372         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
27373         order to ignore the predictor.
27374         (PRED_POLYMORPHIC_CALL): Likewise.
27375         (PRED_RECURSIVE_CALL): Likewise.
27377 2018-01-23  Martin Liska  <mliska@suse.cz>
27379         * tree-profile.c (tree_profiling): Print function header to
27380         aware reader which function we are working on.
27381         * value-prof.c (gimple_find_values_to_profile): Do not print
27382         not interesting value histograms.
27384 2018-01-23  Martin Liska  <mliska@suse.cz>
27386         * profile-count.h (enum profile_quality): Add
27387         profile_uninitialized as the first value. Do not number values
27388         as they are zero based.
27389         (profile_count::verify): Update sanity check.
27390         (profile_probability::verify): Likewise.
27392 2018-01-23  Nathan Sidwell  <nathan@acm.org>
27394         * doc/invoke.texi (ffor-scope): Deprecate.
27396 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27398         PR tree-optimization/83510
27399         * domwalk.c (set_all_edges_as_executable): New function.
27400         (dom_walker::dom_walker): Convert bool param
27401         "skip_unreachable_blocks" to enum reachability.  Move setup of
27402         edge flags to set_all_edges_as_executable and only do it when
27403         reachability is REACHABLE_BLOCKS.
27404         * domwalk.h (enum dom_walker::reachability): New enum.
27405         (dom_walker::dom_walker): Convert bool param
27406         "skip_unreachable_blocks" to enum reachability.
27407         (set_all_edges_as_executable): New decl.
27408         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
27409         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
27410         "reachability".
27411         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
27412         but converting true to REACHABLE_BLOCKS.
27413         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
27414         * tree-vrp.c
27415         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
27416         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
27417         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
27418         REACHABLE_BLOCKS.
27419         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
27420         if check_all_array_refs will be called.
27422 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27424         * tree.c (selftest::test_location_wrappers): Add more test
27425         coverage.
27427 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27429         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
27430         (selftest::test_bit_in_range): Likewise.
27432 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27434         PR testsuite/83888
27435         * doc/sourcebuild.texi (vect_float): Say that the selector
27436         only describes the situation when -funsafe-math-optimizations is on.
27437         (vect_float_strict): Document.
27439 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27441         PR tree-optimization/83965
27442         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
27443         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
27444         instead of checking only for a reduction.
27445         (vect_recog_widen_sum_pattern): Likewise.
27447 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27449         * predict.c (probably_never_executed): Only use precise profile info.
27450         (compute_function_frequency): Skip after inlining hack since we now
27451         have quality checking.
27453 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27455         * profile-count.h (profile_probability::very_unlikely,
27456         profile_probability::unlikely, profile_probability::even): Set
27457         precision to guessed.
27459 2018-01-23  Richard Biener  <rguenther@suse.de>
27461         PR tree-optimization/83963
27462         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
27463         Properly terminate dominator walk when crossing the exit edge not
27464         when visiting its source block.
27466 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
27468         PR c++/83918
27469         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
27470         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
27472 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
27474         PR tree-optimization/83957
27475         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
27476         semicolon after for body surrounded by braces.
27478         PR tree-optimization/83081
27479         * profile-count.h (profile_probability::split): New method.
27480         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
27481         Use profile_probability::split.
27482         (do_compare_rtx_and_jump): Fix adjustment of probabilities
27483         when splitting a single conditional jump into 2.
27485 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
27487         PR tree-optimization/69452
27488         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
27489         decl.
27491 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27493         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
27494         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
27495         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
27497 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27499         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
27500         declaration.
27501         * config/rl78/rl78.md (movdi): New define_expand.
27502         * config/rl78/rl78.c (rl78_split_movdi): New function.
27504 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
27506         PR target/83862
27507         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
27508         no longer used.
27509         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
27510         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
27511         128-bit to produce an UNSPEC move to get the double word with the
27512         signbit and then a shift directly to do signbit.
27513         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
27514         implementation with a new version that just does either a direct
27515         move or a regular move.  Move memory interface to separate insns.
27516         Move insns so they are next to the expander.
27517         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
27518         with signbit move.  Split big and little endian case.
27519         (signbit<mode>2_dm_mem_le): Likewise.
27520         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
27521         (signbit<mode>2_dm2): Likewise.
27523 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27525         * config/rl78/rl78.md (anddi3): New define_expand.
27527 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27529         * config/rl78/rl78.md (umindi3): New define_expand.
27531 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27533         * config/rl78/rl78.md (smindi3): New define_expand.
27535 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27537         * config/rl78/rl78.md (smaxdi3): New define_expand.
27539 2018-01-22  Carl Love  <cel@us.ibm.com>
27541         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
27542         LVX_V1TI): Add macro expansion.
27543         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
27544         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
27545         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
27546         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
27547         Change check to determine if the instruction is a byte reversing
27548         entry.  Fix typo in comment.
27549         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
27550         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
27551         Add def_builtin calls for new builtins.
27552         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
27553         Add define_insn expansion.
27555 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27557         * config/rl78/rl78.md (umaxdi3): New define_expand.
27559 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27561         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
27562         for non-QImode registers.
27564 2018-01-22  Richard Biener  <rguenther@suse.de>
27566         PR tree-optimization/83963
27567         * graphite-scop-detection.c (scop_detection::get_sese): Delay
27568         including the loop exit block.
27569         (scop_detection::merge_sese): Likewise.
27570         (scop_detection::add_scop): Do it here instead.
27572 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27574         * doc/sourcebuild.texi (arm_softfloat): Document.
27576 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
27578         PR gcc/77734
27579         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
27580         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
27581         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
27583 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27584             David Edelsohn  <dje.gcc@gmail.com>
27586         PR target/83946
27587         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27588         Change "crset eq" to "crset 2".
27589         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27590         (*call_indirect_aix<mode>_nospec): Likewise.
27591         (*call_value_indirect_aix<mode>_nospec): Likewise.
27592         (*call_indirect_elfv2<mode>_nospec): Likewise.
27593         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
27594         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
27595         change assembly output from . to $.
27596         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27597         (indirect_jump<mode>_nospec): Change assembly output from . to $.
27598         (*tablejump<mode>_internal1_nospec): Likewise.
27600 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
27602         PR target/80870
27603         * config/sh/sh_optimize_sett_clrt.cc:
27604         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
27606 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
27608         PR tree-optimization/83940
27609         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
27610         offset_dt to vect_constant_def rather than vect_unknown_def_type.
27611         (vect_check_load_store_mask): Add a mask_dt_out parameter and
27612         use it to pass back the definition type.
27613         (vect_check_store_rhs): Likewise rhs_dt_out.
27614         (vect_build_gather_load_calls): Add a mask_dt argument and use
27615         it instead of a call to vect_is_simple_use.
27616         (vectorizable_store): Update calls to vect_check_load_store_mask
27617         and vect_check_store_rhs.  Use the dt returned by the latter instead
27618         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
27619         instead of calls to vect_is_simple_use.  Pass the scalar rather
27620         than the vector operand to vect_is_simple_use when handling
27621         second and subsequent copies of an rhs value.
27622         (vectorizable_load): Update calls to vect_check_load_store_mask
27623         and vect_build_gather_load_calls.  Use the cached mask_dt and
27624         gs_info.offset_dt instead of calls to vect_is_simple_use.
27626 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
27628         PR middle-end/83945
27629         * tree-emutls.c: Include gimplify.h.
27630         (lower_emutls_2): New function.
27631         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
27632         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
27633         it before further processing.
27635         PR target/83930
27636         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
27637         UINTVAL (trueop1) instead of INTVAL (op1).
27639 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
27641         PR debug/81570
27642         PR debug/83728
27643         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
27644         INCOMING_FRAME_SP_OFFSET if not defined.
27645         (scan_trace): Add ENTRY argument.  If true and
27646         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
27647         emit a note to adjust the CFA offset.
27648         (create_cfi_notes): Adjust scan_trace callers.
27649         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
27650         INCOMING_FRAME_SP_OFFSET in the CIE.
27651         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
27652         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
27653         Likewise.
27654         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
27655         * doc/tm.texi: Regenerated.
27657 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27659         PR rtl-optimization/83147
27660         * lra-constraints.c (remove_inheritance_pseudos): Use
27661         lra_substitute_pseudo_within_insn.
27663 2018-01-19  Tom de Vries  <tom@codesourcery.com>
27664             Cesar Philippidis  <cesar@codesourcery.com>
27666         PR target/83920
27667         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
27669 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
27671         PR target/83790
27672         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
27673         spaces for function labels.
27675 2018-01-19  Martin Liska  <mliska@suse.cz>
27677         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
27678         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
27679         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
27680         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
27681         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
27682         (PRED_CONST_RETURN): Change from 69 to 65.
27683         (PRED_NULL_RETURN): Change from 91 to 71.
27684         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
27685         (PRED_LOOP_GUARD): Change from 66 to 73.
27687 2018-01-19  Martin Liska  <mliska@suse.cz>
27689         * predict.c (predict_insn_def): Add new assert.
27690         (struct branch_predictor): Change type to signed integer.
27691         (test_prediction_value_range): Amend test to cover
27692         PROB_UNINITIALIZED.
27693         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
27694         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
27695         (PRED_LOOP_ITERATIONS_MAX): Likewise.
27696         (PRED_LOOP_IV_COMPARE): Likewise.
27697         * predict.h (PROB_UNINITIALIZED): Define new constant.
27699 2018-01-19  Martin Liska  <mliska@suse.cz>
27701         * predict.c (dump_prediction): Add new format for
27702         analyze_brprob.py script which is enabled with -details
27703         suboption.
27704         * profile-count.h (precise_p): New function.
27706 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27708         PR tree-optimization/83922
27709         * tree-vect-loop.c (vect_verify_full_masking): Return false if
27710         there are no statements that need masking.
27711         (vect_active_double_reduction_p): New function.
27712         (vect_analyze_loop_operations): Use it when handling phis that
27713         are not in the loop header.
27715 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27717         PR tree-optimization/83914
27718         * tree-vect-loop.c (vectorizable_induction): Don't convert
27719         init_expr or apply the peeling adjustment for inductions
27720         that are nested within the vectorized loop.
27722 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27724         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
27725         instead of NEG.
27727 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27729         PR sanitizer/81715
27730         PR testsuite/83882
27731         * function.h (gimplify_parameters): Add gimple_seq * argument.
27732         * function.c: Include gimple.h and options.h.
27733         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
27734         for the added local temporaries if needed.
27735         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
27736         if there are any parameter cleanups, wrap whole body into a
27737         try/finally with the cleanups.
27739 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
27741         PR target/82964
27742         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27743         Use GET_MODE_CLASS for scalar floating point.
27745 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27747         PR ipa/82256
27748         patch by PaX Team
27749         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
27750         Fix call of call_cgraph_insertion_hooks.
27752 2018-01-18  Martin Sebor  <msebor@redhat.com>
27754         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
27756 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27758         PR ipa/83619
27759         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
27760         frequencies.
27762 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
27764         PR other/70268
27765         * common.opt: (-ffile-prefix-map): New option.
27766         * opts.c (common_handle_option): Defer it.
27767         * opts-global.c (handle_common_deferred_options): Handle it.
27768         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
27769         * file-prefix-map.h: New file.
27770         (remap_debug_filename, add_debug_prefix_map): ...here.
27771         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
27772         * final.c (debug_prefix_map, add_debug_prefix_map
27773         remap_debug_filename): Move to...
27774         * file-prefix-map.c: New file.
27775         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
27776         generalize, get rid of alloca(), use strrchr() instead of strchr().
27777         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
27778         Implement in terms of add_prefix_map().
27779         (remap_macro_filename, remap_debug_filename): Implement in term of
27780         remap_filename().
27781         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
27782         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
27783         * dbxout.c: Include file-prefix-map.h.
27784         * varasm.c: Likewise.
27785         * vmsdbgout.c: Likewise.
27786         * xcoffout.c: Likewise.
27787         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
27788         * doc/cppopts.texi (-fmacro-prefix-map): Document.
27789         * doc/invoke.texi (-ffile-prefix-map): Document.
27790         (-fdebug-prefix-map): Update description.
27792 2018-01-18  Martin Liska  <mliska@suse.cz>
27794         * config/i386/i386.c (indirect_thunk_name): Document that also
27795         lfence is emitted.
27796         (output_indirect_thunk): Document why both instructions
27797         (pause and lfence) are generated.
27799 2018-01-18  Richard Biener  <rguenther@suse.de>
27801         PR tree-optimization/83887
27802         * graphite-scop-detection.c
27803         (scop_detection::get_nearest_dom_with_single_entry): Remove.
27804         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
27805         (scop_detection::merge_sese): Re-implement with a flood-fill
27806         algorithm that properly finds a SESE region if it exists.
27808 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27810         PR c/61240
27811         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
27812         pointer_diff optimizations use view_convert instead of convert.
27814 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27816         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27817         Generate different code for -mno-speculate-indirect-jumps.
27818         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27819         (*call_indirect_aix<mode>): Disable for
27820         -mno-speculate-indirect-jumps.
27821         (*call_indirect_aix<mode>_nospec): New define_insn.
27822         (*call_value_indirect_aix<mode>): Disable for
27823         -mno-speculate-indirect-jumps.
27824         (*call_value_indirect_aix<mode>_nospec): New define_insn.
27825         (*sibcall_nonlocal_sysv<mode>): Generate different code for
27826         -mno-speculate-indirect-jumps.
27827         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27829 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
27831         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
27832         long double type, set the flags for noting the default long double
27833         type, even if we don't pass or return a long double type.
27835 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
27837         PR ipa/83051
27838         * ipa-inline.c (flatten_function): Do not overwrite final inlining
27839         failure.
27841 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
27843         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
27844         support for merge[hl].
27845         (fold_mergehl_helper): New helper function.
27846         (tree-vector-builder.h): New #include for tree_vector_builder usage.
27847         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
27848         (altivec_vmrglw_direct): Add xxmrglw insn.
27850 2018-01-17  Andrew Waterman  <andrew@sifive.com>
27852         * config/riscv/riscv.c (riscv_conditional_register_usage): If
27853         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
27855 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
27857         PR lto/83121
27858         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
27859         call the lto_location_cache before reading the
27860         DECL_SOURCE_LOCATION of the types.
27862 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
27863             Richard Sandiford  <richard.sandiford@linaro.org>
27865         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
27866         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
27867         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
27868         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
27869         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
27870         Add declaration.
27871         * config/aarch64/constraints.md (aarch64_movti_operand):
27872         Limit immediates.
27873         * config/aarch64/predicates.md (Uti): Add new constraint.
27875 2018-01-17  Carl Love  <cel@us.ibm.com>
27877         * config/rs6000/vsx.md (define_expand xl_len_r,
27878         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
27879         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
27880         lxvll.
27881         (define_expand, define_insn): Move the shift left from  the
27882         define_insn to the define_expand for lxvl and stxvl instructions.
27883         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
27884         and XL_LEN_R definitions to PURE.
27886 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
27888         * config/i386/i386.c (indirect_thunk_name): Declare regno
27889         as unsigned int.  Compare regno with INVALID_REGNUM.
27890         (output_indirect_thunk): Ditto.
27891         (output_indirect_thunk_function): Ditto.
27892         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
27893         in the call to output_indirect_thunk_function.
27895 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
27897         PR middle-end/83884
27898         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
27899         rather than the size of inner_type to determine the stack slot size
27900         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
27902 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
27904         PR target/83546
27905         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
27906         to PTA_SILVERMONT.
27908 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
27910         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
27911         endian Linux systems to optionally enable multilibs for selecting
27912         the long double type if the user configured an explicit type.
27913         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
27914         have no long double multilibs if not defined.
27915         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
27916         warn if the user used -mabi={ieee,ibm}longdouble and we built
27917         multilibs for long double.
27918         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
27919         appropriate multilib option.
27920         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
27921         multilib options.
27922         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
27923         for building long double multilibs.
27924         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
27926 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
27928         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
27929         copies.
27931         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
27932         64 bits.
27933         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
27934         128 bits.
27936         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
27937         variables.
27939         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
27940         return value.
27942 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27944         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
27945         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
27947 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27949         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
27950         different rtl trees depending on TARGET_64BIT.
27951         (rs6000_gen_lvx): Likewise.
27953 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27955         * config/visium/visium.md (nop): Tweak comment.
27956         (hazard_nop): Likewise.
27958 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27960         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
27961         -mspeculate-indirect-jumps.
27962         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
27963         for -mno-speculate-indirect-jumps.
27964         (*call_indirect_elfv2<mode>_nospec): New define_insn.
27965         (*call_value_indirect_elfv2<mode>): Disable for
27966         -mno-speculate-indirect-jumps.
27967         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
27968         (indirect_jump): Emit different RTL for
27969         -mno-speculate-indirect-jumps.
27970         (*indirect_jump<mode>): Disable for
27971         -mno-speculate-indirect-jumps.
27972         (*indirect_jump<mode>_nospec): New define_insn.
27973         (tablejump): Emit different RTL for
27974         -mno-speculate-indirect-jumps.
27975         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
27976         (tablejumpsi_nospec): New define_expand.
27977         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
27978         (tablejumpdi_nospec): New define_expand.
27979         (*tablejump<mode>_internal1): Disable for
27980         -mno-speculate-indirect-jumps.
27981         (*tablejump<mode>_internal1_nospec): New define_insn.
27982         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
27983         option.
27985 2018-01-16  Artyom Skrobov tyomitch@gmail.com
27987         * caller-save.c (insert_save): Drop unnecessary parameter.  All
27988         callers updated.
27990 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27991             Richard Biener  <rguenth@suse.de>
27993         PR libgomp/83590
27994         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
27995         return early, inline manually is_gimple_sizepos.  Make sure if we
27996         call gimplify_expr we don't end up with a gimple constant.
27997         * tree.c (variably_modified_type_p): Don't return true for
27998         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
27999         * gimplify.h (is_gimple_sizepos): Remove.
28001 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28003         PR tree-optimization/83857
28004         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
28005         vectorizable_live_operation for pure SLP statements.
28006         (vectorizable_live_operation): Handle PHIs.
28008 2018-01-16  Richard Biener  <rguenther@suse.de>
28010         PR tree-optimization/83867
28011         * tree-vect-stmts.c (vect_transform_stmt): Precompute
28012         nested_in_vect_loop_p since the scalar stmt may get invalidated.
28014 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
28016         PR c/83844
28017         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
28018         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
28019         If off is not INTEGER_CST, issue a may not be aligned warning
28020         rather than isn't aligned.  Use isn%'t rather than isn't.
28021         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
28022         into MULT_EXPR.
28023         <case MULT_EXPR>: Improve the case when bottom and one of the
28024         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
28025         operand, in that case check if the other operand is multiple of
28026         bottom divided by the INTEGER_CST operand.
28028 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28030         PR target/83858
28031         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
28032         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
28033         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
28034         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
28035         * config/pa/pa.c (pa_function_arg_advance): Likewise.
28036         (pa_function_arg, pa_arg_partial_bytes): Likewise.
28037         (pa_function_arg_size): New function.
28039 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28041         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
28042         in a separate statement.
28044 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28046         PR tree-optimization/83847
28047         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
28048         group gathers and scatters.
28050 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
28052         PR rtl-optimization/86620
28053         * params.def (max-sched-ready-insns): Bump minimum value to 1.
28055         PR rtl-optimization/83213
28056         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
28057         to last if both are JUMP_INSNs.
28059         PR tree-optimization/83843
28060         * gimple-ssa-store-merging.c
28061         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
28062         store_immediate_info for bswap/nop orig_stores.
28064 2018-01-15  Andrew Waterman  <andrew@sifive.com>
28066         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
28067         !TARGET_MUL.
28068         <UDIV>: Increase cost if !TARGET_DIV.
28070 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
28072         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
28073         (define_attr "cr_logical_3op"): New.
28074         (cceq_ior_compare): Adjust.
28075         (cceq_ior_compare_complement): Adjust.
28076         (*cceq_rev_compare): Adjust.
28077         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
28078         (is_cracked_insn): Adjust.
28079         (insn_must_be_first_in_group): Adjust.
28080         * config/rs6000/40x.md: Adjust.
28081         * config/rs6000/440.md: Adjust.
28082         * config/rs6000/476.md: Adjust.
28083         * config/rs6000/601.md: Adjust.
28084         * config/rs6000/603.md: Adjust.
28085         * config/rs6000/6xx.md: Adjust.
28086         * config/rs6000/7450.md: Adjust.
28087         * config/rs6000/7xx.md: Adjust.
28088         * config/rs6000/8540.md: Adjust.
28089         * config/rs6000/cell.md: Adjust.
28090         * config/rs6000/e300c2c3.md: Adjust.
28091         * config/rs6000/e500mc.md: Adjust.
28092         * config/rs6000/e500mc64.md: Adjust.
28093         * config/rs6000/e5500.md: Adjust.
28094         * config/rs6000/e6500.md: Adjust.
28095         * config/rs6000/mpc.md: Adjust.
28096         * config/rs6000/power4.md: Adjust.
28097         * config/rs6000/power5.md: Adjust.
28098         * config/rs6000/power6.md: Adjust.
28099         * config/rs6000/power7.md: Adjust.
28100         * config/rs6000/power8.md: Adjust.
28101         * config/rs6000/power9.md: Adjust.
28102         * config/rs6000/rs64.md: Adjust.
28103         * config/rs6000/titan.md: Adjust.
28105 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28107         * config/i386/predicates.md (indirect_branch_operand): Rewrite
28108         ix86_indirect_branch_register logic.
28110 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28112         * config/i386/constraints.md (Bs): Update
28113         ix86_indirect_branch_register check.  Don't check
28114         ix86_indirect_branch_register with GOT_memory_operand.
28115         (Bw): Likewise.
28116         * config/i386/predicates.md (GOT_memory_operand): Don't check
28117         ix86_indirect_branch_register here.
28118         (GOT32_symbol_operand): Likewise.
28120 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28122         * config/i386/predicates.md (constant_call_address_operand):
28123         Rewrite ix86_indirect_branch_register logic.
28124         (sibcall_insn_operand): Likewise.
28126 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28128         * config/i386/constraints.md (Bs): Replace
28129         ix86_indirect_branch_thunk_register with
28130         ix86_indirect_branch_register.
28131         (Bw): Likewise.
28132         * config/i386/i386.md (indirect_jump): Likewise.
28133         (tablejump): Likewise.
28134         (*sibcall_memory): Likewise.
28135         (*sibcall_value_memory): Likewise.
28136         Peepholes of indirect call and jump via memory: Likewise.
28137         * config/i386/i386.opt: Likewise.
28138         * config/i386/predicates.md (indirect_branch_operand): Likewise.
28139         (GOT_memory_operand): Likewise.
28140         (call_insn_operand): Likewise.
28141         (sibcall_insn_operand): Likewise.
28142         (GOT32_symbol_operand): Likewise.
28144 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
28146         PR middle-end/83837
28147         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
28148         type rather than type addr's type points to.
28149         (expand_omp_atomic_mutex): Likewise.
28150         (expand_omp_atomic): Likewise.
28152 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28154         PR target/83839
28155         * config/i386/i386.c (output_indirect_thunk_function): Use
28156         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
28157         for  __x86_return_thunk.
28159 2018-01-15  Richard Biener  <rguenther@suse.de>
28161         PR middle-end/83850
28162         * expmed.c (extract_bit_field_1): Fix typo.
28164 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28166         PR target/83687
28167         * config/arm/iterators.md (VF): New mode iterator.
28168         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
28169         Remove integer-related logic from pattern.
28170         (neon_vabd<mode>_3): Likewise.
28172 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
28174         PR middle-end/82694
28175         * common.opt (fstrict-overflow): No longer an alias.
28176         (fwrapv-pointer): New option.
28177         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
28178         also for pointer types based on flag_wrapv_pointer.
28179         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
28180         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
28181         opts->x_flag_wrapv got set.
28182         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
28183         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
28184         POINTER_TYPE_OVERFLOW_UNDEFINED.
28185         * match.pd: Likewise in address comparison pattern.
28186         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
28188 2018-01-15  Richard Biener  <rguenther@suse.de>
28190         PR lto/83804
28191         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
28192         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
28193         Reset type names to their identifier if their TYPE_DECL doesn't
28194         have linkage (and thus is used for ODR and devirt).
28195         (save_debug_info_for_decl): Remove.
28196         (save_debug_info_for_type): Likewise.
28197         (add_tree_to_fld_list): Adjust.
28198         * tree-pretty-print.c (dump_generic_node): Make dumping of
28199         type names more robust.
28201 2018-01-15  Richard Biener  <rguenther@suse.de>
28203         * BASE-VER: Bump to 8.0.1.
28205 2018-01-14  Martin Sebor  <msebor@redhat.com>
28207         PR other/83508
28208         * builtins.c (check_access): Avoid warning when the no-warning bit
28209         is set.
28211 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
28213         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
28214         * ira-color (allocno_hard_regs_compare): Likewise.
28216 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
28218         PR target/83013
28219         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
28220         Use .pushsection/.popsection.
28222 2018-01-14  Martin Sebor  <msebor@redhat.com>
28224         PR c++/81327
28225         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
28227 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
28229         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
28230         entry from extra_headers.
28231         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
28232         extra_headers, make the list bitwise identical to the i?86-*-* one.
28234 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28236         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
28237         -mcmodel=large with -mindirect-branch=thunk,
28238         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
28239         -mfunction-return=thunk-extern.
28240         * doc/invoke.texi: Document -mcmodel=large is incompatible with
28241         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
28242         -mfunction-return=thunk and -mfunction-return=thunk-extern.
28244 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28246         * config/i386/i386.c (print_reg): Print the name of the full
28247         integer register without '%'.
28248         (ix86_print_operand): Handle 'V'.
28249         * doc/extend.texi: Document 'V' modifier.
28251 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28253         * config/i386/constraints.md (Bs): Disallow memory operand for
28254         -mindirect-branch-register.
28255         (Bw): Likewise.
28256         * config/i386/predicates.md (indirect_branch_operand): Likewise.
28257         (GOT_memory_operand): Likewise.
28258         (call_insn_operand): Likewise.
28259         (sibcall_insn_operand): Likewise.
28260         (GOT32_symbol_operand): Likewise.
28261         * config/i386/i386.md (indirect_jump): Call convert_memory_address
28262         for -mindirect-branch-register.
28263         (tablejump): Likewise.
28264         (*sibcall_memory): Likewise.
28265         (*sibcall_value_memory): Likewise.
28266         Disallow peepholes of indirect call and jump via memory for
28267         -mindirect-branch-register.
28268         (*call_pop): Replace m with Bw.
28269         (*call_value_pop): Likewise.
28270         (*sibcall_pop_memory): Replace m with Bs.
28271         * config/i386/i386.opt (mindirect-branch-register): New option.
28272         * doc/invoke.texi: Document -mindirect-branch-register option.
28274 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28276         * config/i386/i386-protos.h (ix86_output_function_return): New.
28277         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
28278         set function_return_type.
28279         (indirect_thunk_name): Add ret_p to indicate thunk for function
28280         return.
28281         (output_indirect_thunk_function): Pass false to
28282         indirect_thunk_name.
28283         (ix86_output_indirect_branch_via_reg): Likewise.
28284         (ix86_output_indirect_branch_via_push): Likewise.
28285         (output_indirect_thunk_function): Create alias for function
28286         return thunk if regno < 0.
28287         (ix86_output_function_return): New function.
28288         (ix86_handle_fndecl_attribute): Handle function_return.
28289         (ix86_attribute_table): Add function_return.
28290         * config/i386/i386.h (machine_function): Add
28291         function_return_type.
28292         * config/i386/i386.md (simple_return_internal): Use
28293         ix86_output_function_return.
28294         (simple_return_internal_long): Likewise.
28295         * config/i386/i386.opt (mfunction-return=): New option.
28296         (indirect_branch): Mention -mfunction-return=.
28297         * doc/extend.texi: Document function_return function attribute.
28298         * doc/invoke.texi: Document -mfunction-return= option.
28300 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28302         * config/i386/i386-opts.h (indirect_branch): New.
28303         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
28304         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
28305         with local indirect jump when converting indirect call and jump.
28306         (ix86_set_indirect_branch_type): New.
28307         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
28308         (indirectlabelno): New.
28309         (indirect_thunk_needed): Likewise.
28310         (indirect_thunk_bnd_needed): Likewise.
28311         (indirect_thunks_used): Likewise.
28312         (indirect_thunks_bnd_used): Likewise.
28313         (INDIRECT_LABEL): Likewise.
28314         (indirect_thunk_name): Likewise.
28315         (output_indirect_thunk): Likewise.
28316         (output_indirect_thunk_function): Likewise.
28317         (ix86_output_indirect_branch_via_reg): Likewise.
28318         (ix86_output_indirect_branch_via_push): Likewise.
28319         (ix86_output_indirect_branch): Likewise.
28320         (ix86_output_indirect_jmp): Likewise.
28321         (ix86_code_end): Call output_indirect_thunk_function if needed.
28322         (ix86_output_call_insn): Call ix86_output_indirect_branch if
28323         needed.
28324         (ix86_handle_fndecl_attribute): Handle indirect_branch.
28325         (ix86_attribute_table): Add indirect_branch.
28326         * config/i386/i386.h (machine_function): Add indirect_branch_type
28327         and has_local_indirect_jump.
28328         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
28329         to true.
28330         (tablejump): Likewise.
28331         (*indirect_jump): Use ix86_output_indirect_jmp.
28332         (*tablejump_1): Likewise.
28333         (simple_return_indirect_internal): Likewise.
28334         * config/i386/i386.opt (mindirect-branch=): New option.
28335         (indirect_branch): New.
28336         (keep): Likewise.
28337         (thunk): Likewise.
28338         (thunk-inline): Likewise.
28339         (thunk-extern): Likewise.
28340         * doc/extend.texi: Document indirect_branch function attribute.
28341         * doc/invoke.texi: Document -mindirect-branch= option.
28343 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
28345         PR ipa/83051
28346         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
28348 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
28350         * ipa-inline.c (want_inline_small_function_p): Return false if
28351         inlining has already failed with CIF_FINAL_ERROR.
28352         (update_caller_keys): Call want_inline_small_function_p before
28353         can_inline_edge_p.
28354         (update_callee_keys): Likewise.
28356 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
28358         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
28359         New function.
28360         (rs6000_quadword_masked_address_p): Likewise.
28361         (quad_aligned_load_p): Likewise.
28362         (quad_aligned_store_p): Likewise.
28363         (const_load_sequence_p): Add comment to describe the outer-most loop.
28364         (mimic_memory_attributes_and_flags): New function.
28365         (rs6000_gen_stvx): Likewise.
28366         (replace_swapped_aligned_store): Likewise.
28367         (rs6000_gen_lvx): Likewise.
28368         (replace_swapped_aligned_load): Likewise.
28369         (replace_swapped_load_constant): Capitalize argument name in
28370         comment describing this function.
28371         (rs6000_analyze_swaps): Add a third pass to search for vector loads
28372         and stores that access quad-word aligned addresses and replace
28373         with stvx or lvx instructions when appropriate.
28374         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
28375         New function prototype.
28376         (rs6000_quadword_masked_address_p): Likewise.
28377         (rs6000_gen_lvx): Likewise.
28378         (rs6000_gen_stvx): Likewise.
28379         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
28380         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
28381         when memory address is aligned.
28382         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
28383         this split to select lvx instruction when memory address is aligned.
28384         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
28385         instruction when memory address is aligned.
28386         (*vsx_le_perm_load_v16qi): Likewise.
28387         (four unnamed splitters): Modify to select the stvx instruction
28388         when memory is aligned.
28390 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
28392         * predict.c (determine_unlikely_bbs): Handle correctly BBs
28393         which appears in the queue multiple times.
28395 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28396             Alan Hayward  <alan.hayward@arm.com>
28397             David Sherwood  <david.sherwood@arm.com>
28399         * tree-vectorizer.h (vec_lower_bound): New structure.
28400         (_loop_vec_info): Add check_nonzero and lower_bounds.
28401         (LOOP_VINFO_CHECK_NONZERO): New macro.
28402         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
28403         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
28404         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
28405         fields.  Make seg_len the distance travelled, not including the
28406         access size.
28407         (dr_direction_indicator): Declare.
28408         (dr_zero_step_indicator): Likewise.
28409         (dr_known_forward_stride_p): Likewise.
28410         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
28411         tree-ssanames.h.
28412         (runtime_alias_check_p): Allow runtime alias checks with
28413         variable strides.
28414         (operator ==): Compare access_size and align.
28415         (prune_runtime_alias_test_list): Rework for new distinction between
28416         the access_size and seg_len.
28417         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
28418         segment lengths.
28419         (get_segment_min_max): New function.
28420         (create_intersect_range_checks): Use it.
28421         (dr_step_indicator): New function.
28422         (dr_direction_indicator): Likewise.
28423         (dr_zero_step_indicator): Likewise.
28424         (dr_known_forward_stride_p): Likewise.
28425         * tree-loop-distribution.c (data_ref_segment_size): Return
28426         DR_STEP * (niters - 1).
28427         (compute_alias_check_pairs): Update call to the dr_with_seg_len
28428         constructor.
28429         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
28430         (vect_preserves_scalar_order_p): New function, split out from...
28431         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
28432         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
28433         (vect_vfa_access_size): New function.
28434         (vect_vfa_align): Likewise.
28435         (vect_compile_time_alias): Take access_size_a and access_b arguments.
28436         (dump_lower_bound): New function.
28437         (vect_check_lower_bound): Likewise.
28438         (vect_small_gap_p): Likewise.
28439         (vectorizable_with_step_bound_p): Likewise.
28440         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
28441         depencies if the vectorization factor is 1.  Convert the checks
28442         for nonzero steps into checks on the bounds of DR_STEP.  Try using
28443         a bunds check for variable steps if the minimum required step is
28444         relatively small. Update calls to the dr_with_seg_len
28445         constructor and to vect_compile_time_alias.
28446         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
28447         function.
28448         (vect_loop_versioning): Call it.
28449         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
28450         when retrying.
28451         (vect_estimate_min_profitable_iters): Account for any bounds checks.
28453 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28454             Alan Hayward  <alan.hayward@arm.com>
28455             David Sherwood  <david.sherwood@arm.com>
28457         * doc/sourcebuild.texi (vect_scatter_store): Document.
28458         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
28459         optabs.
28460         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
28461         Document.
28462         * genopinit.c (main): Add supports_vec_scatter_store and
28463         supports_vec_scatter_store_cached to target_optabs.
28464         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
28465         IFN_MASK_SCATTER_STORE.
28466         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
28467         functions.
28468         * internal-fn.h (internal_store_fn_p): Declare.
28469         (internal_fn_stored_value_index): Likewise.
28470         * internal-fn.c (scatter_store_direct): New macro.
28471         (expand_scatter_store_optab_fn): New function.
28472         (direct_scatter_store_optab_supported_p): New macro.
28473         (internal_store_fn_p): New function.
28474         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
28475         IFN_MASK_SCATTER_STORE.
28476         (internal_fn_mask_index): Likewise.
28477         (internal_fn_stored_value_index): New function.
28478         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
28479         for scatter stores.
28480         * optabs-query.h (supports_vec_scatter_store_p): Declare.
28481         * optabs-query.c (supports_vec_scatter_store_p): New function.
28482         * tree-vectorizer.h (vect_get_store_rhs): Declare.
28483         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
28484         true for scatter stores.
28485         (vect_gather_scatter_fn_p): Handle scatter stores too.
28486         (vect_check_gather_scatter): Consider using scatter stores if
28487         supports_vec_scatter_store_p.
28488         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
28489         scatter stores too.
28490         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28491         internal_fn_stored_value_index.
28492         (check_load_store_masking): Handle scatter stores too.
28493         (vect_get_store_rhs): Make public.
28494         (vectorizable_call): Use internal_store_fn_p.
28495         (vectorizable_store): Handle scatter store internal functions.
28496         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
28497         when deciding whether the end of the group has been reached.
28498         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
28499         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
28500         (mask_scatter_store<mode>): New insns.
28502 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28503             Alan Hayward  <alan.hayward@arm.com>
28504             David Sherwood  <david.sherwood@arm.com>
28506         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
28507         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
28508         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
28509         function.
28510         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
28511         Use vect_truncate_gather_scatter_offset if we can't treat the
28512         operation as a normal gather load or scatter store.
28513         (get_group_load_store_type): Take the gather_scatter_info
28514         as argument.  Try using a gather load or scatter store for
28515         single-element groups.
28516         (get_load_store_type): Update calls to get_group_load_store_type
28517         and vect_use_strided_gather_scatters_p.
28519 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28520             Alan Hayward  <alan.hayward@arm.com>
28521             David Sherwood  <david.sherwood@arm.com>
28523         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
28524         optional tree argument.
28525         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
28526         null target hooks.
28527         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
28528         but continue to use the current value as a fallback.
28529         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
28530         to compare the updates.
28531         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
28532         (get_load_store_type): Use it when handling a strided access.
28533         (vect_get_strided_load_store_ops): New function.
28534         (vect_get_data_ptr_increment): Likewise.
28535         (vectorizable_load): Handle strided gather loads.  Always pass
28536         a step to vect_create_data_ref_ptr and bump_vector_ptr.
28538 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28539             Alan Hayward  <alan.hayward@arm.com>
28540             David Sherwood  <david.sherwood@arm.com>
28542         * doc/md.texi (gather_load@var{m}): Document.
28543         (mask_gather_load@var{m}): Likewise.
28544         * genopinit.c (main): Add supports_vec_gather_load and
28545         supports_vec_gather_load_cached to target_optabs.
28546         * optabs-tree.c (init_tree_optimization_optabs): Use
28547         ggc_cleared_alloc to allocate target_optabs.
28548         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
28549         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
28550         functions.
28551         * internal-fn.h (internal_load_fn_p): Declare.
28552         (internal_gather_scatter_fn_p): Likewise.
28553         (internal_fn_mask_index): Likewise.
28554         (internal_gather_scatter_fn_supported_p): Likewise.
28555         * internal-fn.c (gather_load_direct): New macro.
28556         (expand_gather_load_optab_fn): New function.
28557         (direct_gather_load_optab_supported_p): New macro.
28558         (direct_internal_fn_optab): New function.
28559         (internal_load_fn_p): Likewise.
28560         (internal_gather_scatter_fn_p): Likewise.
28561         (internal_fn_mask_index): Likewise.
28562         (internal_gather_scatter_fn_supported_p): Likewise.
28563         * optabs-query.c (supports_at_least_one_mode_p): New function.
28564         (supports_vec_gather_load_p): Likewise.
28565         * optabs-query.h (supports_vec_gather_load_p): Declare.
28566         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
28567         and memory_type field.
28568         (NUM_PATTERNS): Bump to 15.
28569         * tree-vect-data-refs.c: Include internal-fn.h.
28570         (vect_gather_scatter_fn_p): New function.
28571         (vect_describe_gather_scatter_call): Likewise.
28572         (vect_check_gather_scatter): Try using internal functions for
28573         gather loads.  Recognize existing calls to a gather load function.
28574         (vect_analyze_data_refs): Consider using gather loads if
28575         supports_vec_gather_load_p.
28576         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
28577         (vect_get_gather_scatter_offset_type): Likewise.
28578         (vect_convert_mask_for_vectype): Likewise.
28579         (vect_add_conversion_to_patterm): Likewise.
28580         (vect_try_gather_scatter_pattern): Likewise.
28581         (vect_recog_gather_scatter_pattern): New pattern recognizer.
28582         (vect_vect_recog_func_ptrs): Add it.
28583         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28584         internal_fn_mask_index and internal_gather_scatter_fn_p.
28585         (check_load_store_masking): Take the gather_scatter_info as an
28586         argument and handle gather loads.
28587         (vect_get_gather_scatter_ops): New function.
28588         (vectorizable_call): Check internal_load_fn_p.
28589         (vectorizable_load): Likewise.  Handle gather load internal
28590         functions.
28591         (vectorizable_store): Update call to check_load_store_masking.
28592         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
28593         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
28594         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
28595         (aarch64_gather_scale_operand_d): New predicates.
28596         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
28597         (mask_gather_load<mode>): New insns.
28599 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28600             Alan Hayward  <alan.hayward@arm.com>
28601             David Sherwood  <david.sherwood@arm.com>
28603         * optabs.def (fold_left_plus_optab): New optab.
28604         * doc/md.texi (fold_left_plus_@var{m}): Document.
28605         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
28606         * internal-fn.c (fold_left_direct): Define.
28607         (expand_fold_left_optab_fn): Likewise.
28608         (direct_fold_left_optab_supported_p): Likewise.
28609         * fold-const-call.c (fold_const_fold_left): New function.
28610         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
28611         * tree-parloops.c (valid_reduction_p): New function.
28612         (gather_scalar_reductions): Use it.
28613         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
28614         (vect_finish_replace_stmt): Declare.
28615         * tree-vect-loop.c (fold_left_reduction_fn): New function.
28616         (needs_fold_left_reduction_p): New function, split out from...
28617         (vect_is_simple_reduction): ...here.  Accept reductions that
28618         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
28619         (vect_force_simple_reduction): Also store the reduction type in
28620         the assignment's STMT_VINFO_REDUC_TYPE.
28621         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
28622         (merge_with_identity): New function.
28623         (vect_expand_fold_left): Likewise.
28624         (vectorize_fold_left_reduction): Likewise.
28625         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
28626         scalar phi in place for it.  Check for target support and reject
28627         cases that would reassociate the operation.  Defer the transform
28628         phase to vectorize_fold_left_reduction.
28629         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
28630         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
28631         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
28633 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28635         * tree-if-conv.c (predicate_mem_writes): Remove redundant
28636         call to ifc_temp_var.
28638 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28639             Alan Hayward  <alan.hayward@arm.com>
28640             David Sherwood  <david.sherwood@arm.com>
28642         * target.def (legitimize_address_displacement): Take the original
28643         offset as a poly_int.
28644         * targhooks.h (default_legitimize_address_displacement): Update
28645         accordingly.
28646         * targhooks.c (default_legitimize_address_displacement): Likewise.
28647         * doc/tm.texi: Regenerate.
28648         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
28649         as an argument, moving assert of ad->disp == ad->disp_term to...
28650         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
28651         Try calling targetm.legitimize_address_displacement before expanding
28652         the address rather than afterwards, and adjust for the new interface.
28653         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
28654         Match the new hook interface.  Handle SVE addresses.
28655         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
28656         new hook interface.
28658 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28660         * Makefile.in (OBJS): Add early-remat.o.
28661         * target.def (select_early_remat_modes): New hook.
28662         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
28663         * doc/tm.texi: Regenerate.
28664         * targhooks.h (default_select_early_remat_modes): Declare.
28665         * targhooks.c (default_select_early_remat_modes): New function.
28666         * timevar.def (TV_EARLY_REMAT): New timevar.
28667         * passes.def (pass_early_remat): New pass.
28668         * tree-pass.h (make_pass_early_remat): Declare.
28669         * early-remat.c: New file.
28670         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
28671         function.
28672         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
28674 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28675             Alan Hayward  <alan.hayward@arm.com>
28676             David Sherwood  <david.sherwood@arm.com>
28678         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
28679         vfm1 with a bound_epilog parameter.
28680         (vect_do_peeling): Update calls accordingly, and move the prologue
28681         call earlier in the function.  Treat the base bound_epilog as 0 for
28682         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
28683         this base when peeling for gaps.
28684         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
28685         with fully-masked loops.
28686         (vect_estimate_min_profitable_iters): Handle the single peeled
28687         iteration in that case.
28689 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28690             Alan Hayward  <alan.hayward@arm.com>
28691             David Sherwood  <david.sherwood@arm.com>
28693         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
28694         single-element interleaving even if the size is not a power of 2.
28695         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
28696         accesses for single-element interleaving if the group size is
28697         not a power of 2.
28699 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28700             Alan Hayward  <alan.hayward@arm.com>
28701             David Sherwood  <david.sherwood@arm.com>
28703         * doc/md.texi (fold_extract_last_@var{m}): Document.
28704         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
28705         * optabs.def (fold_extract_last_optab): New optab.
28706         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
28707         * internal-fn.c (fold_extract_direct): New macro.
28708         (expand_fold_extract_optab_fn): Likewise.
28709         (direct_fold_extract_optab_supported_p): Likewise.
28710         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
28711         * tree-vect-loop.c (vect_model_reduction_cost): Handle
28712         EXTRACT_LAST_REDUCTION.
28713         (get_initial_def_for_reduction): Do not create an initial vector
28714         for EXTRACT_LAST_REDUCTION reductions.
28715         (vectorizable_reduction): Leave the scalar phi in place for
28716         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
28717         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
28718         epilogue code for EXTRACT_LAST_REDUCTION and defer the
28719         transform phase to vectorizable_condition.
28720         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
28721         split out from...
28722         (vect_finish_stmt_generation): ...here.
28723         (vect_finish_replace_stmt): New function.
28724         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
28725         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
28726         pattern.
28727         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
28729 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28730             Alan Hayward  <alan.hayward@arm.com>
28731             David Sherwood  <david.sherwood@arm.com>
28733         * doc/md.texi (extract_last_@var{m}): Document.
28734         * optabs.def (extract_last_optab): New optab.
28735         * internal-fn.def (EXTRACT_LAST): New internal function.
28736         * internal-fn.c (cond_unary_direct): New macro.
28737         (expand_cond_unary_optab_fn): Likewise.
28738         (direct_cond_unary_optab_supported_p): Likewise.
28739         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
28740         loops using EXTRACT_LAST.
28741         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
28742         (extract_last_<mode>): ...this optab.
28743         (vec_extract<mode><Vel>): Update accordingly.
28745 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28746             Alan Hayward  <alan.hayward@arm.com>
28747             David Sherwood  <david.sherwood@arm.com>
28749         * target.def (empty_mask_is_expensive): New hook.
28750         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
28751         * doc/tm.texi: Regenerate.
28752         * targhooks.h (default_empty_mask_is_expensive): Declare.
28753         * targhooks.c (default_empty_mask_is_expensive): New function.
28754         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
28755         if the target says that empty masks are expensive.
28756         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
28757         New function.
28758         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
28760 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28761             Alan Hayward  <alan.hayward@arm.com>
28762             David Sherwood  <david.sherwood@arm.com>
28764         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
28765         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
28766         (vect_use_loop_mask_for_alignment_p): New function.
28767         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
28768         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
28769         niters_skip argument.  Make sure that the first niters_skip elements
28770         of the first iteration are inactive.
28771         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
28772         Update call to vect_set_loop_masks_directly.
28773         (get_misalign_in_elems): New function, split out from...
28774         (vect_gen_prolog_loop_niters): ...here.
28775         (vect_update_init_of_dr): Take a code argument that specifies whether
28776         the adjustment should be added or subtracted.
28777         (vect_update_init_of_drs): Likewise.
28778         (vect_prepare_for_masked_peels): New function.
28779         (vect_do_peeling): Skip prologue peeling if we're using a mask
28780         instead.  Update call to vect_update_inits_of_drs.
28781         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28782         mask_skip_niters.
28783         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
28784         alignment.  Do not include the number of peeled iterations in
28785         the minimum threshold in that case.
28786         (vectorizable_induction): Adjust the start value down by
28787         LOOP_VINFO_MASK_SKIP_NITERS iterations.
28788         (vect_transform_loop): Call vect_prepare_for_masked_peels.
28789         Take the number of skipped iterations into account when calculating
28790         the loop bounds.
28791         * tree-vect-stmts.c (vect_gen_while_not): New function.
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/sourcebuild.texi (vect_fully_masked): Document.
28798         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
28799         default value to 0.
28800         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
28801         split out from...
28802         (vect_analyze_loop_2): ...here. Don't check the vectorization
28803         factor against the number of loop iterations if the loop is
28804         fully-masked.
28806 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28807             Alan Hayward  <alan.hayward@arm.com>
28808             David Sherwood  <david.sherwood@arm.com>
28810         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
28811         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
28812         (dump_groups): Update accordingly.
28813         (iv_use::mem_type): New member variable.
28814         (address_p): New function.
28815         (record_use): Add a mem_type argument and initialize the new
28816         mem_type field.
28817         (record_group_use): Add a mem_type argument.  Use address_p.
28818         Remove obsolete null checks of base_object.  Update call to record_use.
28819         (find_interesting_uses_op): Update call to record_group_use.
28820         (find_interesting_uses_cond): Likewise.
28821         (find_interesting_uses_address): Likewise.
28822         (get_mem_type_for_internal_fn): New function.
28823         (find_address_like_use): Likewise.
28824         (find_interesting_uses_stmt): Try find_address_like_use before
28825         calling find_interesting_uses_op.
28826         (addr_offset_valid_p): Use the iv mem_type field as the type
28827         of the addressed memory.
28828         (add_autoinc_candidates): Likewise.
28829         (get_address_cost): Likewise.
28830         (split_small_address_groups_p): Use address_p.
28831         (split_address_groups): Likewise.
28832         (add_iv_candidate_for_use): Likewise.
28833         (autoinc_possible_for_pair): Likewise.
28834         (rewrite_groups): Likewise.
28835         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
28836         (determine_group_iv_cost): Update after split of USE_ADDRESS.
28837         (get_alias_ptr_type_for_ptr_address): New function.
28838         (rewrite_use_address): Rewrite address uses in calls that were
28839         identified by find_address_like_use.
28841 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28842             Alan Hayward  <alan.hayward@arm.com>
28843             David Sherwood  <david.sherwood@arm.com>
28845         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
28846         TARGET_MEM_REFs.
28847         * gimple-expr.h (is_gimple_addressable: Likewise.
28848         * gimple-expr.c (is_gimple_address): Likewise.
28849         * internal-fn.c (expand_call_mem_ref): New function.
28850         (expand_mask_load_optab_fn): Use it.
28851         (expand_mask_store_optab_fn): Likewise.
28853 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28854             Alan Hayward  <alan.hayward@arm.com>
28855             David Sherwood  <david.sherwood@arm.com>
28857         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
28858         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
28859         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
28860         (cond_umax@var{mode}): Document.
28861         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
28862         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
28863         (cond_umin_optab, cond_umax_optab): New optabs.
28864         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
28865         (COND_IOR, COND_XOR): New internal functions.
28866         * internal-fn.h (get_conditional_internal_fn): Declare.
28867         * internal-fn.c (cond_binary_direct): New macro.
28868         (expand_cond_binary_optab_fn): Likewise.
28869         (direct_cond_binary_optab_supported_p): Likewise.
28870         (get_conditional_internal_fn): New function.
28871         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
28872         Cope with reduction statements that are vectorized as calls rather
28873         than assignments.
28874         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
28875         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
28876         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
28877         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
28878         (UNSPEC_COND_EOR): New unspecs.
28879         (optab): Add mappings for them.
28880         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
28881         (sve_int_op, sve_fp_op): New int attributes.
28883 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28884             Alan Hayward  <alan.hayward@arm.com>
28885             David Sherwood  <david.sherwood@arm.com>
28887         * optabs.def (while_ult_optab): New optab.
28888         * doc/md.texi (while_ult@var{m}@var{n}): Document.
28889         * internal-fn.def (WHILE_ULT): New internal function.
28890         * internal-fn.h (direct_internal_fn_supported_p): New override
28891         that takes two types as argument.
28892         * internal-fn.c (while_direct): New macro.
28893         (expand_while_optab_fn): New function.
28894         (convert_optab_supported_p): Likewise.
28895         (direct_while_optab_supported_p): New macro.
28896         * wide-int.h (wi::udiv_ceil): New function.
28897         * tree-vectorizer.h (rgroup_masks): New structure.
28898         (vec_loop_masks): New typedef.
28899         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
28900         and fully_masked_p.
28901         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
28902         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
28903         (vect_max_vf): New function.
28904         (slpeel_make_loop_iterate_ntimes): Delete.
28905         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
28906         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
28907         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
28908         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
28909         internal-fn.h, stor-layout.h and optabs-query.h.
28910         (vect_set_loop_mask): New function.
28911         (add_preheader_seq): Likewise.
28912         (add_header_seq): Likewise.
28913         (interleave_supported_p): Likewise.
28914         (vect_maybe_permute_loop_masks): Likewise.
28915         (vect_set_loop_masks_directly): Likewise.
28916         (vect_set_loop_condition_masked): Likewise.
28917         (vect_set_loop_condition_unmasked): New function, split out from
28918         slpeel_make_loop_iterate_ntimes.
28919         (slpeel_make_loop_iterate_ntimes): Rename to..
28920         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
28921         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
28922         (vect_do_peeling): Update call accordingly.
28923         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
28924         loops.
28925         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28926         mask_compare_type, can_fully_mask_p and fully_masked_p.
28927         (release_vec_loop_masks): New function.
28928         (_loop_vec_info): Use it to free the loop masks.
28929         (can_produce_all_loop_masks_p): New function.
28930         (vect_get_max_nscalars_per_iter): Likewise.
28931         (vect_verify_full_masking): Likewise.
28932         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
28933         retries, and free the mask rgroups before retrying.  Check loop-wide
28934         reasons for disallowing fully-masked loops.  Make the final decision
28935         about whether use a fully-masked loop or not.
28936         (vect_estimate_min_profitable_iters): Do not assume that peeling
28937         for the number of iterations will be needed for fully-masked loops.
28938         (vectorizable_reduction): Disable fully-masked loops.
28939         (vectorizable_live_operation): Likewise.
28940         (vect_halve_mask_nunits): New function.
28941         (vect_double_mask_nunits): Likewise.
28942         (vect_record_loop_mask): Likewise.
28943         (vect_get_loop_mask): Likewise.
28944         (vect_transform_loop): Handle the case in which the final loop
28945         iteration might handle a partial vector.  Call vect_set_loop_condition
28946         instead of slpeel_make_loop_iterate_ntimes.
28947         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
28948         (check_load_store_masking): New function.
28949         (prepare_load_store_mask): Likewise.
28950         (vectorizable_store): Handle fully-masked loops.
28951         (vectorizable_load): Likewise.
28952         (supportable_widening_operation): Use vect_halve_mask_nunits for
28953         booleans.
28954         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
28955         (vect_gen_while): New function.
28956         * config/aarch64/aarch64.md (umax<mode>3): New expander.
28957         (aarch64_uqdec<mode>): New insn.
28959 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28960             Alan Hayward  <alan.hayward@arm.com>
28961             David Sherwood  <david.sherwood@arm.com>
28963         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
28964         (reduc_xor_scal_optab): New optabs.
28965         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
28966         (reduc_xor_scal_@var{m}): Document.
28967         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
28968         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
28969         internal functions.
28970         * fold-const-call.c (fold_const_call): Handle them.
28971         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
28972         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
28973         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
28974         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
28975         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
28976         (UNSPEC_XORV): New unspecs.
28977         (optab): Add entries for them.
28978         (BITWISEV): New int iterator.
28979         (bit_reduc_op): New int attributes.
28981 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28982             Alan Hayward  <alan.hayward@arm.com>
28983             David Sherwood  <david.sherwood@arm.com>
28985         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
28986         * internal-fn.def (VEC_SHL_INSERT): New internal function.
28987         * optabs.def (vec_shl_insert_optab): New optab.
28988         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
28989         (duplicate_and_interleave): Likewise.
28990         * tree-vect-loop.c: Include internal-fn.h.
28991         (neutral_op_for_slp_reduction): New function, split out from
28992         get_initial_defs_for_reduction.
28993         (get_initial_def_for_reduction): Handle option 2 for variable-length
28994         vectors by loading the neutral value into a vector and then shifting
28995         the initial value into element 0.
28996         (get_initial_defs_for_reduction): Replace the code argument with
28997         the neutral value calculated by neutral_op_for_slp_reduction.
28998         Use gimple_build_vector for constant-length vectors.
28999         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
29000         but the first group_size elements have a neutral value.
29001         Use duplicate_and_interleave otherwise.
29002         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
29003         Update call to get_initial_defs_for_reduction.  Handle SLP
29004         reductions for variable-length vectors by creating one vector
29005         result for each scalar result, with the elements associated
29006         with other scalar results stubbed out with the neutral value.
29007         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
29008         Require IFN_VEC_SHL_INSERT for double reductions on
29009         variable-length vectors, or SLP reductions that have
29010         a neutral value.  Require can_duplicate_and_interleave_p
29011         support for variable-length unchained SLP reductions if there
29012         is no neutral value, such as for MIN/MAX reductions.  Also require
29013         the number of vector elements to be a multiple of the number of
29014         SLP statements when doing variable-length unchained SLP reductions.
29015         Update call to vect_create_epilog_for_reduction.
29016         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
29017         and remove initial values.
29018         (duplicate_and_interleave): Make public.
29019         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
29020         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
29022 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29023             Alan Hayward  <alan.hayward@arm.com>
29024             David Sherwood  <david.sherwood@arm.com>
29026         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
29027         (can_duplicate_and_interleave_p): New function.
29028         (vect_get_and_check_slp_defs): Take the vector of statements
29029         rather than just the current one.  Remove excess parentheses.
29030         Restriction rejectinon of vect_constant_def and vect_external_def
29031         for variable-length vectors to boolean types, or types for which
29032         can_duplicate_and_interleave_p is false.
29033         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
29034         (duplicate_and_interleave): New function.
29035         (vect_get_constant_vectors): Use gimple_build_vector for
29036         constant-length vectors and suitable variable-length constant
29037         vectors.  Use duplicate_and_interleave for other variable-length
29038         vectors.  Don't defer the update when inserting new statements.
29040 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29041             Alan Hayward  <alan.hayward@arm.com>
29042             David Sherwood  <david.sherwood@arm.com>
29044         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
29045         min_profitable_iters doesn't go negative.
29047 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29048             Alan Hayward  <alan.hayward@arm.com>
29049             David Sherwood  <david.sherwood@arm.com>
29051         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
29052         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
29053         * optabs.def (vec_mask_load_lanes_optab): New optab.
29054         (vec_mask_store_lanes_optab): Likewise.
29055         * internal-fn.def (MASK_LOAD_LANES): New internal function.
29056         (MASK_STORE_LANES): Likewise.
29057         * internal-fn.c (mask_load_lanes_direct): New macro.
29058         (mask_store_lanes_direct): Likewise.
29059         (expand_mask_load_optab_fn): Handle masked operations.
29060         (expand_mask_load_lanes_optab_fn): New macro.
29061         (expand_mask_store_optab_fn): Handle masked operations.
29062         (expand_mask_store_lanes_optab_fn): New macro.
29063         (direct_mask_load_lanes_optab_supported_p): Likewise.
29064         (direct_mask_store_lanes_optab_supported_p): Likewise.
29065         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
29066         parameter.
29067         (vect_load_lanes_supported): Likewise.
29068         * tree-vect-data-refs.c (strip_conversion): New function.
29069         (can_group_stmts_p): Likewise.
29070         (vect_analyze_data_ref_accesses): Use it instead of checking
29071         for a pair of assignments.
29072         (vect_store_lanes_supported): Take a masked_p parameter.
29073         (vect_load_lanes_supported): Likewise.
29074         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
29075         vect_store_lanes_supported and vect_load_lanes_supported.
29076         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
29077         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
29078         parameter.  Don't allow gaps for masked accesses.
29079         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
29080         and vect_load_lanes_supported.
29081         (get_load_store_type): Take a masked_p parameter and update
29082         call to get_group_load_store_type.
29083         (vectorizable_store): Update call to get_load_store_type.
29084         Handle IFN_MASK_STORE_LANES.
29085         (vectorizable_load): Update call to get_load_store_type.
29086         Handle IFN_MASK_LOAD_LANES.
29088 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29089             Alan Hayward  <alan.hayward@arm.com>
29090             David Sherwood  <david.sherwood@arm.com>
29092         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
29093         modes for SVE.
29094         * config/aarch64/aarch64-protos.h
29095         (aarch64_sve_struct_memory_operand_p): Declare.
29096         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
29097         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
29098         (VPRED, vpred): Handle SVE structure modes.
29099         * config/aarch64/constraints.md (Utx): New constraint.
29100         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
29101         (aarch64_sve_struct_nonimmediate_operand): New predicates.
29102         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
29103         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
29104         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
29105         structure modes.  Split into pieces after RA.
29106         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
29107         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
29108         New patterns.
29109         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
29110         SVE structure modes.
29111         (aarch64_classify_address): Likewise.
29112         (sizetochar): Move earlier in file.
29113         (aarch64_print_operand): Handle SVE register lists.
29114         (aarch64_array_mode): New function.
29115         (aarch64_sve_struct_memory_operand_p): Likewise.
29116         (TARGET_ARRAY_MODE): Redefine.
29118 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29119             Alan Hayward  <alan.hayward@arm.com>
29120             David Sherwood  <david.sherwood@arm.com>
29122         * target.def (array_mode): New target hook.
29123         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
29124         * doc/tm.texi: Regenerate.
29125         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
29126         * hooks.c (hook_optmode_mode_uhwi_none): New function.
29127         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
29128         targetm.array_mode.
29129         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
29130         type sizes.
29132 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29133             Alan Hayward  <alan.hayward@arm.com>
29134             David Sherwood  <david.sherwood@arm.com>
29136         * fold-const.c (fold_binary_loc): Check the argument types
29137         rather than the result type when testing for a vector operation.
29139 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29141         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
29142         * doc/tm.texi: Regenerate.
29144 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29145             Alan Hayward  <alan.hayward@arm.com>
29146             David Sherwood  <david.sherwood@arm.com>
29148         * doc/invoke.texi (-msve-vector-bits=): Document new option.
29149         (sve): Document new AArch64 extension.
29150         * doc/md.texi (w): Extend the description of the AArch64
29151         constraint to include SVE vectors.
29152         (Upl, Upa): Document new AArch64 predicate constraints.
29153         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
29154         enum.
29155         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
29156         (msve-vector-bits=): New option.
29157         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
29158         SVE when these are disabled.
29159         (sve): New extension.
29160         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
29161         modes.  Adjust their number of units based on aarch64_sve_vg.
29162         (MAX_BITSIZE_MODE_ANY_MODE): Define.
29163         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
29164         aarch64_addr_query_type.
29165         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
29166         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
29167         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
29168         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
29169         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
29170         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
29171         (aarch64_simd_imm_zero_p): Delete.
29172         (aarch64_check_zero_based_sve_index_immediate): Declare.
29173         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29174         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29175         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29176         (aarch64_sve_float_mul_immediate_p): Likewise.
29177         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29178         rather than an rtx.
29179         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
29180         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
29181         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
29182         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
29183         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
29184         (aarch64_regmode_natural_size): Likewise.
29185         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
29186         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
29187         left one place.
29188         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
29189         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
29190         for VG and the SVE predicate registers.
29191         (V_ALIASES): Add a "z"-prefixed alias.
29192         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
29193         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
29194         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
29195         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
29196         (REG_CLASS_NAMES): Add entries for them.
29197         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
29198         and the predicate registers.
29199         (aarch64_sve_vg): Declare.
29200         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
29201         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
29202         (REGMODE_NATURAL_SIZE): Define.
29203         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
29204         SVE macros.
29205         * config/aarch64/aarch64.c: Include cfgrtl.h.
29206         (simd_immediate_info): Add a constructor for series vectors,
29207         and an associated step field.
29208         (aarch64_sve_vg): New variable.
29209         (aarch64_dbx_register_number): Handle VG and the predicate registers.
29210         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
29211         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
29212         (VEC_ANY_DATA, VEC_STRUCT): New constants.
29213         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
29214         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
29215         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
29216         (aarch64_get_mask_mode): New functions.
29217         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
29218         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
29219         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
29220         predicate modes and predicate registers.  Explicitly restrict
29221         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
29222         to store a vector mode if it is recognized by
29223         aarch64_classify_vector_mode.
29224         (aarch64_regmode_natural_size): New function.
29225         (aarch64_hard_regno_caller_save_mode): Return the original mode
29226         for predicates.
29227         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
29228         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
29229         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
29230         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
29231         functions.
29232         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
29233         does not overlap dest if the function is frame-related.  Handle
29234         SVE constants.
29235         (aarch64_split_add_offset): New function.
29236         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
29237         them aarch64_add_offset.
29238         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
29239         and update call to aarch64_sub_sp.
29240         (aarch64_add_cfa_expression): New function.
29241         (aarch64_expand_prologue): Pass extra temporary registers to the
29242         functions above.  Handle the case in which we need to emit new
29243         DW_CFA_expressions for registers that were originally saved
29244         relative to the stack pointer, but now have to be expressed
29245         relative to the frame pointer.
29246         (aarch64_output_mi_thunk): Pass extra temporary registers to the
29247         functions above.
29248         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
29249         IP0 and IP1 values for SVE frames.
29250         (aarch64_expand_vec_series): New function.
29251         (aarch64_expand_sve_widened_duplicate): Likewise.
29252         (aarch64_expand_sve_const_vector): Likewise.
29253         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
29254         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
29255         into the register, rather than emitting a SET directly.
29256         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
29257         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
29258         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
29259         (offset_9bit_signed_scaled_p): New functions.
29260         (aarch64_replicate_bitmask_imm): New function.
29261         (aarch64_bitmask_imm): Use it.
29262         (aarch64_cannot_force_const_mem): Reject expressions involving
29263         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
29264         (aarch64_classify_index): Handle SVE indices, by requiring
29265         a plain register index with a scale that matches the element size.
29266         (aarch64_classify_address): Handle SVE addresses.  Assert that
29267         the mode of the address is VOIDmode or an integer mode.
29268         Update call to aarch64_classify_symbol.
29269         (aarch64_classify_symbolic_expression): Update call to
29270         aarch64_classify_symbol.
29271         (aarch64_const_vec_all_in_range_p): New function.
29272         (aarch64_print_vector_float_operand): Likewise.
29273         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
29274         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
29275         and the FP immediates 1.0 and 0.5.
29276         (aarch64_print_address_internal): Handle SVE addresses.
29277         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
29278         (aarch64_regno_regclass): Handle predicate registers.
29279         (aarch64_secondary_reload): Handle big-endian reloads of SVE
29280         data modes.
29281         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
29282         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
29283         (aarch64_convert_sve_vector_bits): New function.
29284         (aarch64_override_options): Use it to handle -msve-vector-bits=.
29285         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29286         rather than an rtx.
29287         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
29288         Handle SVE vector and predicate modes.  Accept VL-based constants
29289         that need only one temporary register, and VL offsets that require
29290         no temporary registers.
29291         (aarch64_conditional_register_usage): Mark the predicate registers
29292         as fixed if SVE isn't available.
29293         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
29294         Return true for SVE vector and predicate modes.
29295         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
29296         rather than an unsigned int.  Handle SVE modes.
29297         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
29298         SVE modes.
29299         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
29300         if SVE is enabled.
29301         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29302         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29303         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29304         (aarch64_sve_float_mul_immediate_p): New functions.
29305         (aarch64_sve_valid_immediate): New function.
29306         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
29307         Explicitly reject structure modes.  Check for INDEX constants.
29308         Handle PTRUE and PFALSE constants.
29309         (aarch64_check_zero_based_sve_index_immediate): New function.
29310         (aarch64_simd_imm_zero_p): Delete.
29311         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
29312         vector modes.  Accept constants in the range of CNT[BHWD].
29313         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
29314         ask for an Advanced SIMD mode.
29315         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
29316         (aarch64_simd_vector_alignment): Handle SVE predicates.
29317         (aarch64_vectorize_preferred_vector_alignment): New function.
29318         (aarch64_simd_vector_alignment_reachable): Use it instead of
29319         the vector size.
29320         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
29321         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
29322         functions.
29323         (MAX_VECT_LEN): Delete.
29324         (expand_vec_perm_d): Add a vec_flags field.
29325         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
29326         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
29327         (aarch64_evpc_ext): Don't apply a big-endian lane correction
29328         for SVE modes.
29329         (aarch64_evpc_rev): Rename to...
29330         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
29331         (aarch64_evpc_rev_global): New function.
29332         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
29333         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
29334         MAX_VECT_LEN.
29335         (aarch64_evpc_sve_tbl): New function.
29336         (aarch64_expand_vec_perm_const_1): Update after rename of
29337         aarch64_evpc_rev.  Handle SVE permutes too, trying
29338         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
29339         than aarch64_evpc_tbl.
29340         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
29341         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
29342         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
29343         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
29344         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
29345         (aarch64_expand_sve_vcond): New functions.
29346         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
29347         of aarch64_vector_mode_p.
29348         (aarch64_dwarf_poly_indeterminate_value): New function.
29349         (aarch64_compute_pressure_classes): Likewise.
29350         (aarch64_can_change_mode_class): Likewise.
29351         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
29352         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
29353         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
29354         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
29355         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
29356         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
29357         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
29358         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
29359         constraints.
29360         (Dn, Dl, Dr): Accept const as well as const_vector.
29361         (Dz): Likewise.  Compare against CONST0_RTX.
29362         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
29363         of "vector" where appropriate.
29364         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
29365         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
29366         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
29367         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
29368         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
29369         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
29370         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
29371         (v_int_equiv): Extend to SVE modes.
29372         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
29373         mode attributes.
29374         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
29375         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
29376         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
29377         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
29378         (SVE_COND_FP_CMP): New int iterators.
29379         (perm_hilo): Handle the new unpack unspecs.
29380         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
29381         attributes.
29382         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
29383         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
29384         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
29385         (aarch64_equality_operator, aarch64_constant_vector_operand)
29386         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
29387         (aarch64_sve_nonimmediate_operand): Likewise.
29388         (aarch64_sve_general_operand): Likewise.
29389         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
29390         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
29391         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
29392         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
29393         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
29394         (aarch64_sve_float_arith_immediate): Likewise.
29395         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
29396         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
29397         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
29398         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
29399         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
29400         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
29401         (aarch64_sve_float_arith_operand): Likewise.
29402         (aarch64_sve_float_arith_with_sub_operand): Likewise.
29403         (aarch64_sve_float_mul_operand): Likewise.
29404         (aarch64_sve_vec_perm_operand): Likewise.
29405         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
29406         (aarch64_mov_operand): Accept const_poly_int and const_vector.
29407         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
29408         as well as const_vector.
29409         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
29410         in file.  Use CONST0_RTX and CONSTM1_RTX.
29411         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
29412         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
29413         Use aarch64_simd_imm_zero.
29414         * config/aarch64/aarch64-sve.md: New file.
29415         * config/aarch64/aarch64.md: Include it.
29416         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
29417         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
29418         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
29419         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
29420         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
29421         (sve): New attribute.
29422         (enabled): Disable instructions with the sve attribute unless
29423         TARGET_SVE.
29424         (movqi, movhi): Pass CONST_POLY_INT operaneds through
29425         aarch64_expand_mov_immediate.
29426         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
29427         CNT[BHSD] immediates.
29428         (movti): Split CONST_POLY_INT moves into two halves.
29429         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
29430         Split additions that need a temporary here if the destination
29431         is the stack pointer.
29432         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
29433         (*add<mode>3_poly_1): New instruction.
29434         (set_clobber_cc): New expander.
29436 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29438         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
29439         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
29440         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
29441         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
29442         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
29443         Change innermode from fixed_mode_size to machine_mode.
29444         (simplify_subreg): Update call accordingly.  Handle a constant-sized
29445         subreg of a variable-length CONST_VECTOR.
29447 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29448             Alan Hayward  <alan.hayward@arm.com>
29449             David Sherwood  <david.sherwood@arm.com>
29451         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
29452         (add_offset_to_base): New function, split out from...
29453         (create_mem_ref): ...here.  When handling a scale other than 1,
29454         check first whether the address is valid without the offset.
29455         Add it into the base if so, leaving the index and scale as-is.
29457 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29459         PR c++/83778
29460         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
29461         fold_for_warn before checking if arg2 is INTEGER_CST.
29463 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
29465         * config/rs6000/predicates.md (load_multiple_operation): Delete.
29466         (store_multiple_operation): Delete.
29467         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
29468         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
29469         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
29470         guarded by TARGET_STRING.
29471         (rs6000_output_load_multiple): Delete.
29472         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29473         OPTION_MASK_STRING / TARGET_STRING handling.
29474         (print_operand) <'N', 'O'>: Add comment that these are unused now.
29475         (const rs6000_opt_masks) <"string">: Change mask to 0.
29476         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
29477         (MASK_STRING): Delete.
29478         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
29479         parts.  Simplify.
29480         (load_multiple): Delete.
29481         (*ldmsi8): Delete.
29482         (*ldmsi7): Delete.
29483         (*ldmsi6): Delete.
29484         (*ldmsi5): Delete.
29485         (*ldmsi4): Delete.
29486         (*ldmsi3): Delete.
29487         (store_multiple): Delete.
29488         (*stmsi8): Delete.
29489         (*stmsi7): Delete.
29490         (*stmsi6): Delete.
29491         (*stmsi5): Delete.
29492         (*stmsi4): Delete.
29493         (*stmsi3): Delete.
29494         (movmemsi_8reg): Delete.
29495         (corresponding unnamed define_insn): Delete.
29496         (movmemsi_6reg): Delete.
29497         (corresponding unnamed define_insn): Delete.
29498         (movmemsi_4reg): Delete.
29499         (corresponding unnamed define_insn): Delete.
29500         (movmemsi_2reg): Delete.
29501         (corresponding unnamed define_insn): Delete.
29502         (movmemsi_1reg): Delete.
29503         (corresponding unnamed define_insn): Delete.
29504         * config/rs6000/rs6000.opt (mno-string): New.
29505         (mstring): Replace by deprecation warning stub.
29506         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
29508 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29510         * regrename.c (regrename_do_replace): If replacing the same
29511         reg multiple times, try to reuse last created gen_raw_REG.
29513         PR debug/81155
29514         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
29515         main to workaround a bug in GDB.
29517 2018-01-12  Tom de Vries  <tom@codesourcery.com>
29519         PR target/83737
29520         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
29522 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
29524         PR rtl-optimization/80481
29525         * ira-color.c (get_cap_member): New function.
29526         (allocnos_conflict_by_live_ranges_p): Use it.
29527         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
29528         (setup_slot_coalesced_allocno_live_ranges): Ditto.
29530 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
29532         PR target/83628
29533         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
29534         (*saddl_se_1): Ditto.
29535         (*ssubsi_1): Ditto.
29536         (*ssubl_se_1): Ditto.
29538 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29540         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
29541         rather than wi::to_widest for DR_INITs.
29542         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
29543         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
29544         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
29545         INTEGER_CSTs.
29546         (vect_analyze_group_access_1): Note that here.
29548 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29550         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
29551         polynomial type sizes.
29553 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29555         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
29556         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
29557         (gimple_add_tmp_var): Likewise.
29559 2018-01-12  Martin Liska  <mliska@suse.cz>
29561         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
29562         (gimple_alloc_sizes): Likewise.
29563         (dump_gimple_statistics): Use PRIu64 in printf format.
29564         * gimple.h: Change uint64_t to int.
29566 2018-01-12  Martin Liska  <mliska@suse.cz>
29568         * tree-core.h: Use uint64_t instead of int.
29569         * tree.c (tree_node_counts): Likewise.
29570         (tree_node_sizes): Likewise.
29571         (dump_tree_statistics): Use PRIu64 in printf format.
29573 2018-01-12  Martin Liska  <mliska@suse.cz>
29575         * Makefile.in: As qsort_chk is implemented in vec.c, add
29576         vec.o to linkage of gencfn-macros.
29577         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
29578         passing the info to record_node_allocation_statistics.
29579         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
29580         and pass the info.
29581         * ggc-common.c (struct ggc_usage): Add operator== and use
29582         it in operator< and compare function.
29583         * mem-stats.h (struct mem_usage): Likewise.
29584         * vec.c (struct vec_usage): Remove operator< and compare
29585         function. Can be simply inherited.
29587 2018-01-12  Martin Jambor  <mjambor@suse.cz>
29589         PR target/81616
29590         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
29591         * tree-ssa-math-opts.c: Include domwalk.h.
29592         (convert_mult_to_fma_1): New function.
29593         (fma_transformation_info): New type.
29594         (fma_deferring_state): Likewise.
29595         (cancel_fma_deferring): New function.
29596         (result_of_phi): Likewise.
29597         (last_fma_candidate_feeds_initial_phi): Likewise.
29598         (convert_mult_to_fma): Added deferring logic, split actual
29599         transformation to convert_mult_to_fma_1.
29600         (math_opts_dom_walker): New type.
29601         (math_opts_dom_walker::after_dom_children): New method, body moved
29602         here from pass_optimize_widening_mul::execute, added deferring logic
29603         bits.
29604         (pass_optimize_widening_mul::execute): Moved most of code to
29605         math_opts_dom_walker::after_dom_children.
29606         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
29607         * config/i386/i386.c (ix86_option_override_internal): Added
29608         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
29610 2018-01-12  Richard Biener  <rguenther@suse.de>
29612         PR debug/83157
29613         * dwarf2out.c (gen_variable_die): Do not reset old_die for
29614         inline instance vars.
29616 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
29618         PR target/81819
29619         * config/rx/rx.c (rx_is_restricted_memory_address):
29620         Handle SUBREG case.
29622 2018-01-12  Richard Biener  <rguenther@suse.de>
29624         PR tree-optimization/80846
29625         * target.def (split_reduction): New target hook.
29626         * targhooks.c (default_split_reduction): New function.
29627         * targhooks.h (default_split_reduction): Declare.
29628         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
29629         target requests first reduce vectors by combining low and high
29630         parts.
29631         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
29632         (get_vectype_for_scalar_type_and_size): Export.
29633         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
29634         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
29635         * doc/tm.texi: Regenerate.
29636         * config/i386/i386.c (ix86_split_reduction): Implement
29637         TARGET_VECTORIZE_SPLIT_REDUCTION.
29639 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29641         PR target/83368
29642         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
29643         in PIC mode except for TARGET_VXWORKS_RTP.
29644         * config/sparc/sparc.c: Include cfgrtl.h.
29645         (TARGET_INIT_PIC_REG): Define.
29646         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
29647         (sparc_pic_register_p): New predicate.
29648         (sparc_legitimate_address_p): Use it.
29649         (sparc_legitimize_pic_address): Likewise.
29650         (sparc_delegitimize_address): Likewise.
29651         (sparc_mode_dependent_address_p): Likewise.
29652         (gen_load_pcrel_sym): Remove 4th parameter.
29653         (load_got_register): Adjust call to above.  Remove obsolete stuff.
29654         (sparc_expand_prologue): Do not call load_got_register here.
29655         (sparc_flat_expand_prologue): Likewise.
29656         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
29657         (sparc_use_pseudo_pic_reg): New function.
29658         (sparc_init_pic_reg): Likewise.
29659         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
29660         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
29662 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
29664         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
29665         Add item for branch_cost.
29667 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29669         PR rtl-optimization/83565
29670         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
29671         not extend the result to a larger mode for rotate operations.
29672         (num_sign_bit_copies1): Likewise.
29674 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29676         PR target/40411
29677         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
29678         -symbolic.
29679         Use values-Xc.o for -pedantic.
29680         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
29682 2018-01-12  Martin Liska  <mliska@suse.cz>
29684         PR ipa/83054
29685         * ipa-devirt.c (final_warning_record::grow_type_warnings):
29686         New function.
29687         (possible_polymorphic_call_targets): Use it.
29688         (ipa_devirt): Likewise.
29690 2018-01-12  Martin Liska  <mliska@suse.cz>
29692         * profile-count.h (enum profile_quality): Use 0 as invalid
29693         enum value of profile_quality.
29695 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
29697         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
29698         -mext-string options.
29700 2018-01-12  Richard Biener  <rguenther@suse.de>
29702         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
29703         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
29704         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
29705         Likewise.
29706         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
29708 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
29710         * configure.ac (--with-long-double-format): Add support for the
29711         configuration option to change the default long double format on
29712         PowerPC systems.
29713         * config.gcc (powerpc*-linux*-*): Likewise.
29714         * configure: Regenerate.
29715         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
29716         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
29717         used without modification.
29719 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29721         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
29722         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
29723         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
29724         MISC_BUILTIN_SPEC_BARRIER.
29725         (rs6000_init_builtins): Likewise.
29726         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
29727         enum value.
29728         (speculation_barrier): New define_insn.
29729         * doc/extend.texi: Document __builtin_speculation_barrier.
29731 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29733         PR target/83203
29734         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
29735         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
29736         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
29737         iterators.
29738         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
29739         integral modes instead of "ss" and "sd".
29740         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
29741         vectors with 32-bit and 64-bit elements.
29742         (vecdupssescalarmodesuffix): New mode attribute.
29743         (vec_dup<mode>): Use it.
29745 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
29747         PR target/83330
29748         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
29749         frame if argument is passed on stack.
29751 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29753         PR target/82682
29754         * ree.c (combine_reaching_defs): Optimize also
29755         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
29756         reg2=any_extend(exp); reg1=reg2;, formatting fix.
29758 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29760         PR middle-end/83189
29761         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
29763 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29765         PR middle-end/83718
29766         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
29767         after they are computed.
29769 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
29771         PR tree-optimization/83695
29772         * gimple-loop-linterchange.cc
29773         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
29774         reset cached scev information after interchange.
29775         (pass_linterchange::execute): Remove call to scev_reset_htab.
29777 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29779         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
29780         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
29781         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
29782         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
29783         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
29784         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
29785         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
29786         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
29787         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
29788         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
29789         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
29790         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
29791         (V_lane_reg): Likewise.
29792         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
29793         New define_expand.
29794         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
29795         (vfmal_lane_low<mode>_intrinsic,
29796         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
29797         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
29798         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
29799         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
29800         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
29801         vfmsl_lane_high<mode>_intrinsic): New define_insns.
29803 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29805         * config/arm/arm-cpus.in (fp16fml): New feature.
29806         (ALL_SIMD): Add fp16fml.
29807         (armv8.2-a): Add fp16fml as an option.
29808         (armv8.3-a): Likewise.
29809         (armv8.4-a): Add fp16fml as part of fp16.
29810         * config/arm/arm.h (TARGET_FP16FML): Define.
29811         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
29812         when appropriate.
29813         * config/arm/arm-modes.def (V2HF): Define.
29814         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
29815         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
29816         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
29817         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
29818         vfmsl_low, vfmsl_high): New set of builtins.
29819         * config/arm/iterators.md (PLUSMINUS): New code iterator.
29820         (vfml_op): New code attribute.
29821         (VFMLHALVES): New int iterator.
29822         (VFML, VFMLSEL): New mode attributes.
29823         (V_reg): Define mapping for V2HF.
29824         (V_hi, V_lo): New mode attributes.
29825         (VF_constraint): Likewise.
29826         (vfml_half, vfml_half_selector): New int attributes.
29827         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
29828         define_expand.
29829         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
29830         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
29831         New define_insn.
29832         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
29833         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
29834         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
29835         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
29836         documentation.
29837         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
29838         Document new effective target and option set.
29840 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29842         * config/arm/arm-cpus.in (armv8_4): New feature.
29843         (ARMv8_4a): New fgroup.
29844         (armv8.4-a): New arch.
29845         * config/arm/arm-tables.opt: Regenerate.
29846         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
29847         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
29848         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
29849         Add matching rules for -march=armv8.4-a and extensions.
29850         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
29852 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
29854         PR target/81821
29855         * config/rx/rx.md (BW): New mode attribute.
29856         (sync_lock_test_and_setsi): Add mode suffix to insn output.
29858 2018-01-11  Richard Biener  <rguenther@suse.de>
29860         PR tree-optimization/83435
29861         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
29862         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
29863         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
29865 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29866             Alan Hayward  <alan.hayward@arm.com>
29867             David Sherwood  <david.sherwood@arm.com>
29869         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
29870         field.
29871         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
29872         (aarch64_print_address_internal): Use it to check for a zero offset.
29874 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29875             Alan Hayward  <alan.hayward@arm.com>
29876             David Sherwood  <david.sherwood@arm.com>
29878         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
29879         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
29880         Return a poly_int64 rather than a HOST_WIDE_INT.
29881         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
29882         rather than a HOST_WIDE_INT.
29883         * config/aarch64/aarch64.h (aarch64_frame): Protect with
29884         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
29885         hard_fp_offset, frame_size, initial_adjust, callee_offset and
29886         final_offset from HOST_WIDE_INT to poly_int64.
29887         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
29888         to_constant when getting the number of units in an Advanced SIMD
29889         mode.
29890         (aarch64_builtin_vectorized_function): Check for a constant number
29891         of units.
29892         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
29893         GET_MODE_SIZE.
29894         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
29895         attribute instead of GET_MODE_NUNITS.
29896         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
29897         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
29898         GET_MODE_SIZE for fixed-size registers.
29899         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
29900         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
29901         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
29902         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
29903         (aarch64_print_operand, aarch64_print_address_internal)
29904         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
29905         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
29906         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
29907         Handle polynomial GET_MODE_SIZE.
29908         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
29909         wider than SImode without modification.
29910         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
29911         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
29912         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
29913         passing and returning SVE modes.
29914         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
29915         rather than GEN_INT.
29916         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
29917         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
29918         (aarch64_allocate_and_probe_stack_space): Likewise.
29919         (aarch64_layout_frame): Cope with polynomial offsets.
29920         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
29921         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
29922         polynomial offsets.
29923         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
29924         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
29925         poly_int64 rather than a HOST_WIDE_INT.
29926         (aarch64_get_separate_components, aarch64_process_components)
29927         (aarch64_expand_prologue, aarch64_expand_epilogue)
29928         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
29929         (aarch64_anchor_offset): New function, split out from...
29930         (aarch64_legitimize_address): ...here.
29931         (aarch64_builtin_vectorization_cost): Handle polynomial
29932         TYPE_VECTOR_SUBPARTS.
29933         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
29934         GET_MODE_NUNITS.
29935         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
29936         number of elements from the PARALLEL rather than the mode.
29937         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
29938         rather than GET_MODE_BITSIZE.
29939         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
29940         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
29941         (aarch64_expand_vec_perm_const_1): Handle polynomial
29942         d->perm.length () and d->perm elements.
29943         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
29944         Apply to_constant to d->perm elements.
29945         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
29946         polynomial CONST_VECTOR_NUNITS.
29947         (aarch64_move_pointer): Take amount as a poly_int64 rather
29948         than an int.
29949         (aarch64_progress_pointer): Avoid temporary variable.
29950         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
29951         the mode attribute instead of GET_MODE.
29953 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29954             Alan Hayward  <alan.hayward@arm.com>
29955             David Sherwood  <david.sherwood@arm.com>
29957         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
29958         x exists before using it.
29959         (aarch64_add_constant_internal): Rename to...
29960         (aarch64_add_offset_1): ...this.  Replace regnum with separate
29961         src and dest rtxes.  Handle the case in which they're different,
29962         including when the offset is zero.  Replace scratchreg with an rtx.
29963         Use 2 additions if there is no spare register into which we can
29964         move a 16-bit constant.
29965         (aarch64_add_constant): Delete.
29966         (aarch64_add_offset): Replace reg with separate src and dest
29967         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
29968         Use aarch64_add_offset_1.
29969         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
29970         an rtx rather than an int.  Take the delta as a poly_int64
29971         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
29972         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
29973         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
29974         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
29975         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
29976         and aarch64_add_sp.
29977         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
29978         aarch64_add_constant.
29980 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29982         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
29983         Use scalar_float_mode.
29985 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29987         * config/aarch64/aarch64-simd.md
29988         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
29989         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
29990         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
29991         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
29992         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
29993         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
29994         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
29995         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
29996         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
29997         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
29999 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
30001         PR target/83514
30002         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
30003         targ_options->x_arm_arch_string is non NULL.
30005 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
30007         * config/aarch64/aarch64.h
30008         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
30010 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
30012         PR target/82096
30013         * expmed.c (emit_store_flag_force): Swap if const op0
30014         and change VOIDmode to mode of op0.
30016 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
30018         PR rtl-optimization/83761
30019         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
30020         than bytes to mode_for_size.
30022 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30024         PR middle-end/83189
30025         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
30026         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
30027         profile.
30029 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30031         PR middle-end/83575
30032         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
30033         when in layout mode.
30034         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
30035         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
30036         partition fixup.
30038 2018-01-10  Michael Collison  <michael.collison@arm.com>
30040         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
30041         * config/aarch64/aarch64-option-extension.def: Add
30042         AARCH64_OPT_EXTENSION of 'fp16fml'.
30043         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30044         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
30045         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
30046         * config/aarch64/constraints.md (Ui7): New constraint.
30047         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
30048         (VFMLA_SEL_W): Ditto.
30049         (f16quad): Ditto.
30050         (f16mac1): Ditto.
30051         (VFMLA16_LOW): New int iterator.
30052         (VFMLA16_HIGH): Ditto.
30053         (UNSPEC_FMLAL): New unspec.
30054         (UNSPEC_FMLSL): Ditto.
30055         (UNSPEC_FMLAL2): Ditto.
30056         (UNSPEC_FMLSL2): Ditto.
30057         (f16mac): New code attribute.
30058         * config/aarch64/aarch64-simd-builtins.def
30059         (aarch64_fmlal_lowv2sf): Ditto.
30060         (aarch64_fmlsl_lowv2sf): Ditto.
30061         (aarch64_fmlalq_lowv4sf): Ditto.
30062         (aarch64_fmlslq_lowv4sf): Ditto.
30063         (aarch64_fmlal_highv2sf): Ditto.
30064         (aarch64_fmlsl_highv2sf): Ditto.
30065         (aarch64_fmlalq_highv4sf): Ditto.
30066         (aarch64_fmlslq_highv4sf): Ditto.
30067         (aarch64_fmlal_lane_lowv2sf): Ditto.
30068         (aarch64_fmlsl_lane_lowv2sf): Ditto.
30069         (aarch64_fmlal_laneq_lowv2sf): Ditto.
30070         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
30071         (aarch64_fmlalq_lane_lowv4sf): Ditto.
30072         (aarch64_fmlsl_lane_lowv4sf): Ditto.
30073         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
30074         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
30075         (aarch64_fmlal_lane_highv2sf): Ditto.
30076         (aarch64_fmlsl_lane_highv2sf): Ditto.
30077         (aarch64_fmlal_laneq_highv2sf): Ditto.
30078         (aarch64_fmlsl_laneq_highv2sf): Ditto.
30079         (aarch64_fmlalq_lane_highv4sf): Ditto.
30080         (aarch64_fmlsl_lane_highv4sf): Ditto.
30081         (aarch64_fmlalq_laneq_highv4sf): Ditto.
30082         (aarch64_fmlsl_laneq_highv4sf): Ditto.
30083         * config/aarch64/aarch64-simd.md:
30084         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
30085         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
30086         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
30087         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
30088         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
30089         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
30090         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
30091         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
30092         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
30093         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
30094         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
30095         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
30096         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
30097         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
30098         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
30099         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
30100         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
30101         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
30102         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
30103         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
30104         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
30105         (vfmlsl_low_u32): Ditto.
30106         (vfmlalq_low_u32): Ditto.
30107         (vfmlslq_low_u32): Ditto.
30108         (vfmlal_high_u32): Ditto.
30109         (vfmlsl_high_u32): Ditto.
30110         (vfmlalq_high_u32): Ditto.
30111         (vfmlslq_high_u32): Ditto.
30112         (vfmlal_lane_low_u32): Ditto.
30113         (vfmlsl_lane_low_u32): Ditto.
30114         (vfmlal_laneq_low_u32): Ditto.
30115         (vfmlsl_laneq_low_u32): Ditto.
30116         (vfmlalq_lane_low_u32): Ditto.
30117         (vfmlslq_lane_low_u32): Ditto.
30118         (vfmlalq_laneq_low_u32): Ditto.
30119         (vfmlslq_laneq_low_u32): Ditto.
30120         (vfmlal_lane_high_u32): Ditto.
30121         (vfmlsl_lane_high_u32): Ditto.
30122         (vfmlal_laneq_high_u32): Ditto.
30123         (vfmlsl_laneq_high_u32): Ditto.
30124         (vfmlalq_lane_high_u32): Ditto.
30125         (vfmlslq_lane_high_u32): Ditto.
30126         (vfmlalq_laneq_high_u32): Ditto.
30127         (vfmlslq_laneq_high_u32): Ditto.
30128         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
30129         (AARCH64_FL_FOR_ARCH8_4): New.
30130         (AARCH64_ISA_F16FML): New ISA flag.
30131         (TARGET_F16FML): New feature flag for fp16fml.
30132         (doc/invoke.texi): Document new fp16fml option.
30134 2018-01-10  Michael Collison  <michael.collison@arm.com>
30136         * config/aarch64/aarch64-builtins.c:
30137         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
30138         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30139         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
30140         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
30141         (AARCH64_ISA_SHA3): New ISA flag.
30142         (TARGET_SHA3): New feature flag for sha3.
30143         * config/aarch64/iterators.md (sha512_op): New int attribute.
30144         (CRYPTO_SHA512): New int iterator.
30145         (UNSPEC_SHA512H): New unspec.
30146         (UNSPEC_SHA512H2): Ditto.
30147         (UNSPEC_SHA512SU0): Ditto.
30148         (UNSPEC_SHA512SU1): Ditto.
30149         * config/aarch64/aarch64-simd-builtins.def
30150         (aarch64_crypto_sha512hqv2di): New builtin.
30151         (aarch64_crypto_sha512h2qv2di): Ditto.
30152         (aarch64_crypto_sha512su0qv2di): Ditto.
30153         (aarch64_crypto_sha512su1qv2di): Ditto.
30154         (aarch64_eor3qv8hi): Ditto.
30155         (aarch64_rax1qv2di): Ditto.
30156         (aarch64_xarqv2di): Ditto.
30157         (aarch64_bcaxqv8hi): Ditto.
30158         * config/aarch64/aarch64-simd.md:
30159         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
30160         (aarch64_crypto_sha512su0qv2di): Ditto.
30161         (aarch64_crypto_sha512su1qv2di): Ditto.
30162         (aarch64_eor3qv8hi): Ditto.
30163         (aarch64_rax1qv2di): Ditto.
30164         (aarch64_xarqv2di): Ditto.
30165         (aarch64_bcaxqv8hi): Ditto.
30166         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
30167         (vsha512h2q_u64): Ditto.
30168         (vsha512su0q_u64): Ditto.
30169         (vsha512su1q_u64): Ditto.
30170         (veor3q_u16): Ditto.
30171         (vrax1q_u64): Ditto.
30172         (vxarq_u64): Ditto.
30173         (vbcaxq_u16): Ditto.
30174         * config/arm/types.md (crypto_sha512): New type attribute.
30175         (crypto_sha3): Ditto.
30176         (doc/invoke.texi): Document new sha3 option.
30178 2018-01-10  Michael Collison  <michael.collison@arm.com>
30180         * config/aarch64/aarch64-builtins.c:
30181         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
30182         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30183         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
30184         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
30185         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
30186         (AARCH64_ISA_SM4): New ISA flag.
30187         (TARGET_SM4): New feature flag for sm4.
30188         * config/aarch64/aarch64-simd-builtins.def
30189         (aarch64_sm3ss1qv4si): Ditto.
30190         (aarch64_sm3tt1aq4si): Ditto.
30191         (aarch64_sm3tt1bq4si): Ditto.
30192         (aarch64_sm3tt2aq4si): Ditto.
30193         (aarch64_sm3tt2bq4si): Ditto.
30194         (aarch64_sm3partw1qv4si): Ditto.
30195         (aarch64_sm3partw2qv4si): Ditto.
30196         (aarch64_sm4eqv4si): Ditto.
30197         (aarch64_sm4ekeyqv4si): Ditto.
30198         * config/aarch64/aarch64-simd.md:
30199         (aarch64_sm3ss1qv4si): Ditto.
30200         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
30201         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
30202         (aarch64_sm4eqv4si): Ditto.
30203         (aarch64_sm4ekeyqv4si): Ditto.
30204         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
30205         (sm3part_op): Ditto.
30206         (CRYPTO_SM3TT): Ditto.
30207         (CRYPTO_SM3PART): Ditto.
30208         (UNSPEC_SM3SS1): New unspec.
30209         (UNSPEC_SM3TT1A): Ditto.
30210         (UNSPEC_SM3TT1B): Ditto.
30211         (UNSPEC_SM3TT2A): Ditto.
30212         (UNSPEC_SM3TT2B): Ditto.
30213         (UNSPEC_SM3PARTW1): Ditto.
30214         (UNSPEC_SM3PARTW2): Ditto.
30215         (UNSPEC_SM4E): Ditto.
30216         (UNSPEC_SM4EKEY): Ditto.
30217         * config/aarch64/constraints.md (Ui2): New constraint.
30218         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
30219         * config/arm/types.md (crypto_sm3): New type attribute.
30220         (crypto_sm4): Ditto.
30221         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
30222         (vsm3tt1aq_u32): Ditto.
30223         (vsm3tt1bq_u32): Ditto.
30224         (vsm3tt2aq_u32): Ditto.
30225         (vsm3tt2bq_u32): Ditto.
30226         (vsm3partw1q_u32): Ditto.
30227         (vsm3partw2q_u32): Ditto.
30228         (vsm4eq_u32): Ditto.
30229         (vsm4ekeyq_u32): Ditto.
30230         (doc/invoke.texi): Document new sm4 option.
30232 2018-01-10  Michael Collison  <michael.collison@arm.com>
30234         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
30235         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
30236         (AARCH64_FL_FOR_ARCH8_4): New.
30237         (AARCH64_FL_V8_4): New flag.
30238         (doc/invoke.texi): Document new armv8.4-a option.
30240 2018-01-10  Michael Collison  <michael.collison@arm.com>
30242         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30243         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
30244         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
30245         * config/aarch64/aarch64-option-extension.def: Add
30246         AARCH64_OPT_EXTENSION of 'sha2'.
30247         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
30248         (crypto): Disable sha2 and aes if crypto disabled.
30249         (crypto): Enable aes and sha2 if enabled.
30250         (simd): Disable sha2 and aes if simd disabled.
30251         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
30252         New flags.
30253         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
30254         (TARGET_SHA2): New feature flag for sha2.
30255         (TARGET_AES): New feature flag for aes.
30256         * config/aarch64/aarch64-simd.md:
30257         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
30258         conditional on TARGET_AES.
30259         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
30260         (aarch64_crypto_sha1hsi): Make pattern conditional
30261         on TARGET_SHA2.
30262         (aarch64_crypto_sha1hv4si): Ditto.
30263         (aarch64_be_crypto_sha1hv4si): Ditto.
30264         (aarch64_crypto_sha1su1v4si): Ditto.
30265         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
30266         (aarch64_crypto_sha1su0v4si): Ditto.
30267         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
30268         (aarch64_crypto_sha256su0v4si): Ditto.
30269         (aarch64_crypto_sha256su1v4si): Ditto.
30270         (doc/invoke.texi): Document new aes and sha2 options.
30272 2018-01-10  Martin Sebor  <msebor@redhat.com>
30274         PR tree-optimization/83781
30275         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
30276         as string arrays.
30278 2018-01-11  Martin Sebor  <msebor@gmail.com>
30279             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
30281         PR tree-optimization/83501
30282         PR tree-optimization/81703
30284         * tree-ssa-strlen.c (get_string_cst): Rename...
30285         (get_string_len): ...to this.  Handle global constants.
30286         (handle_char_store): Adjust.
30288 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
30289             Jim Wilson  <jimw@sifive.com>
30291         * config/riscv/riscv-protos.h (riscv_output_return): New.
30292         * config/riscv/riscv.c (struct machine_function): New naked_p field.
30293         (riscv_attribute_table, riscv_output_return),
30294         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
30295         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
30296         (riscv_compute_frame_info): Only compute frame->mask if not a naked
30297         function.
30298         (riscv_expand_prologue): Add early return for naked function.
30299         (riscv_expand_epilogue): Likewise.
30300         (riscv_function_ok_for_sibcall): Return false for naked function.
30301         (riscv_set_current_function): New.
30302         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
30303         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
30304         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
30305         * doc/extend.texi (RISC-V Function Attributes): New.
30307 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
30309         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
30310         check for 128-bit long double before checking TCmode.
30311         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
30312         128-bit long doubles before checking TFmode or TCmode.
30313         (FLOAT128_IBM_P): Likewise.
30315 2018-01-10  Martin Sebor  <msebor@redhat.com>
30317         PR tree-optimization/83671
30318         * builtins.c (c_strlen): Unconditionally return zero for the empty
30319         string.
30320         Use -Warray-bounds for warnings.
30321         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
30322         for non-constant array indices with COMPONENT_REF, arrays of
30323         arrays, and pointers to arrays.
30324         (gimple_fold_builtin_strlen): Determine and set length range for
30325         non-constant character arrays.
30327 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
30329         PR middle-end/81897
30330         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
30331         empty blocks.
30333 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
30335         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
30337 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
30339         PR target/83399
30340         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
30341         VECTOR_MEM_ALTIVEC_OR_VSX_P.
30342         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
30343         indexed_or_indirect_operand predicate.
30344         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
30345         (*vsx_le_perm_load_v8hi): Likewise.
30346         (*vsx_le_perm_load_v16qi): Likewise.
30347         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
30348         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
30349         (*vsx_le_perm_store_v8hi): Likewise.
30350         (*vsx_le_perm_store_v16qi): Likewise.
30351         (eight unnamed splitters): Likewise.
30353 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
30355         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
30356         * config/rs6000/emmintrin.h: Likewise.
30357         * config/rs6000/mmintrin.h: Likewise.
30358         * config/rs6000/xmmintrin.h: Likewise.
30360 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
30362         PR c++/43486
30363         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
30364         "public_flag".
30365         * tree.c (tree_nop_conversion): Return true for location wrapper
30366         nodes.
30367         (maybe_wrap_with_location): New function.
30368         (selftest::check_strip_nops): New function.
30369         (selftest::test_location_wrappers): New function.
30370         (selftest::tree_c_tests): Call it.
30371         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
30372         (maybe_wrap_with_location): New decl.
30373         (EXPR_LOCATION_WRAPPER_P): New macro.
30374         (location_wrapper_p): New inline function.
30375         (tree_strip_any_location_wrapper): New inline function.
30377 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
30379         PR target/83735
30380         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
30381         stack_realign_offset for the largest alignment of stack slot
30382         actually used.
30383         (ix86_find_max_used_stack_alignment): New function.
30384         (ix86_finalize_stack_frame_flags): Use it.  Set
30385         max_used_stack_alignment if we don't realign stack.
30386         * config/i386/i386.h (machine_function): Add
30387         max_used_stack_alignment.
30389 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
30391         * config/arm/arm.opt (-mbranch-cost): New option.
30392         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
30393         account.
30395 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
30397         PR target/83629
30398         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
30399         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
30401 2018-01-10  Richard Biener  <rguenther@suse.de>
30403         PR debug/83765
30404         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
30405         early out so it also covers the case where we have a non-NULL
30406         origin.
30408 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
30410         PR tree-optimization/83753
30411         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
30412         for non-strided grouped accesses if the number of elements is 1.
30414 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30416         PR target/81616
30417         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
30418         * i386.h (TARGET_USE_GATHER): Define.
30419         * x86-tune.def (X86_TUNE_USE_GATHER): New.
30421 2018-01-10  Martin Liska  <mliska@suse.cz>
30423         PR bootstrap/82831
30424         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
30425         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
30426         partitioning.
30427         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
30428         CLEANUP_NO_PARTITIONING is not set.
30430 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
30432         * doc/rtl.texi: Remove documentation of (const ...) wrappers
30433         for vectors, as a partial revert of r254296.
30434         * rtl.h (const_vec_p): Delete.
30435         (const_vec_duplicate_p): Don't test for vector CONSTs.
30436         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
30437         * expmed.c (make_tree): Likewise.
30439         Revert:
30440         * common.md (E, F): Use CONSTANT_P instead of checking for
30441         CONST_VECTOR.
30442         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
30443         checking for CONST_VECTOR.
30445 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30447         PR middle-end/83575
30448         * predict.c (force_edge_cold): Handle in more sane way edges
30449         with no prediction.
30451 2018-01-09  Carl Love  <cel@us.ibm.com>
30453         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
30454         V4SI, V4SF types.
30455         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
30456         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
30457         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
30458         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
30459         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
30460         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
30461         * config/rs6000/rs6000-protos.h: Add extern defition for
30462         rs6000_generate_float2_double_code.
30463         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
30464         function.
30465         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
30466         (float2_v2df): Add define_expand.
30468 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
30470         PR target/83628
30471         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
30472         op_mode in the force_to_mode call.
30474 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30476         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
30477         instead of checking each element individually.
30478         (aarch64_evpc_uzp): Likewise.
30479         (aarch64_evpc_zip): Likewise.
30480         (aarch64_evpc_ext): Likewise.
30481         (aarch64_evpc_rev): Likewise.
30482         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
30483         instead of checking each element individually.  Return true without
30484         generating rtl if
30485         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
30486         whether all selected elements come from the same input, instead of
30487         checking each element individually.  Remove calls to gen_rtx_REG,
30488         start_sequence and end_sequence and instead assert that no rtl is
30489         generated.
30491 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30493         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
30494         order of HIGH and CONST checks.
30496 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30498         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
30499         if the destination isn't an SSA_NAME.
30501 2018-01-09  Richard Biener  <rguenther@suse.de>
30503         PR tree-optimization/83668
30504         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
30505         move prologue...
30506         (canonicalize_loop_form): ... here, renamed from ...
30507         (canonicalize_loop_closed_ssa_form): ... this and amended to
30508         swap successor edges for loop exit blocks to make us use
30509         the RPO order we need for initial schedule generation.
30511 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
30513         PR tree-optimization/64811
30514         * match.pd: When optimizing comparisons with Inf, avoid
30515         introducing or losing exceptions from comparisons with NaN.
30517 2018-01-09  Martin Liska  <mliska@suse.cz>
30519         PR sanitizer/82517
30520         * asan.c (shadow_mem_size): Add gcc_assert.
30522 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
30524         Don't save registers in main().
30526         PR target/83738
30527         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
30528         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
30529         * config/avr/avr.c (avr_set_current_function): Don't error if
30530         naked, OS_task or OS_main are specified at the same time.
30531         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
30532         OS_main.
30533         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
30534         attribute.
30535         * common/config/avr/avr-common.c (avr_option_optimization_table):
30536         Switch on -mmain-is-OS_task for optimizing compilations.
30538 2018-01-09  Richard Biener  <rguenther@suse.de>
30540         PR tree-optimization/83572
30541         * graphite.c: Include cfganal.h.
30542         (graphite_transform_loops): Connect infinite loops to exit
30543         and remove fake edges at the end.
30545 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30547         * ipa-inline.c (edge_badness): Revert accidental checkin.
30549 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30551         PR ipa/80763
30552         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
30553         symbols; not inline clones.
30555 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
30557         PR target/83507
30558         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
30559         hard registers.  Formatting fixes.
30561         PR preprocessor/83722
30562         * gcc.c (try_generate_repro): Pass
30563         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
30564         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
30565         do_report_bug.
30567 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
30568             Kito Cheng  <kito.cheng@gmail.com>
30570         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
30571         (riscv_leaf_function_p): Delete.
30572         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
30574 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30576         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
30577         function.
30578         (do_ifelse): New function.
30579         (do_isel): New function.
30580         (do_sub3): New function.
30581         (do_add3): New function.
30582         (do_load_mask_compare): New function.
30583         (do_overlap_load_compare): New function.
30584         (expand_compare_loop): New function.
30585         (expand_block_compare): Call expand_compare_loop() when appropriate.
30586         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
30587         option description.
30588         (-mblock-compare-inline-loop-limit): New option.
30590 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30592         PR target/83677
30593         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
30594         Reverse order of second and third operands in first alternative.
30595         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
30596         of first and second elements in UNSPEC_VPERMR vector.
30597         (altivec_expand_vec_perm_le): Likewise.
30599 2018-01-08  Jeff Law  <law@redhat.com>
30601         PR rtl-optimizatin/81308
30602         * tree-switch-conversion.c (cfg_altered): New file scoped static.
30603         (process_switch): If group_case_labels makes a change, then set
30604         cfg_altered.
30605         (pass_convert_switch::execute): If a switch is converted, then
30606         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
30608         PR rtl-optimization/81308
30609         * recog.c (split_all_insns): Conditionally cleanup the CFG after
30610         splitting insns.
30612 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
30614         PR target/83663 - Revert r255946
30615         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
30616         generation for cases where splatting a value is not useful.
30617         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
30618         across a vec_duplicate and a paradoxical subreg forming a vector
30619         mode to a vec_concat.
30621 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30623         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
30624         -march=armv8.3-a variants.
30625         * config/arm/t-multilib: Likewise.
30626         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
30628 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30630         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
30631         to generate rtl.
30632         (cceq_ior_compare_complement): Give it a name so I can use it, and
30633         change boolean_or_operator predicate to boolean_operator so it can
30634         be used to generate a crand.
30635         (eqne): New code iterator.
30636         (bd/bd_neg): New code_attrs.
30637         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
30638         a single define_insn.
30639         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
30640         decrement (bdnzt/bdnzf/bdzt/bdzf).
30641         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
30642         with the new names of the branch decrement patterns, and added the
30643         names of the branch decrement conditional patterns.
30645 2018-01-08  Richard Biener  <rguenther@suse.de>
30647         PR tree-optimization/83563
30648         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
30649         cache.
30651 2018-01-08  Richard Biener  <rguenther@suse.de>
30653         PR middle-end/83713
30654         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
30656 2018-01-08  Richard Biener  <rguenther@suse.de>
30658         PR tree-optimization/83685
30659         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
30660         references to abnormals.
30662 2018-01-08  Richard Biener  <rguenther@suse.de>
30664         PR lto/83719
30665         * dwarf2out.c (output_indirect_strings): Handle empty
30666         skeleton_debug_str_hash.
30667         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
30669 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30671         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
30672         (emit_store_direct): Likewise.
30673         (arc_trampoline_adjust_address): Likewise.
30674         (arc_asm_trampoline_template): New function.
30675         (arc_initialize_trampoline): Use asm_trampoline_template.
30676         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
30677         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
30678         * config/arc/arc.md (flush_icache): Delete pattern.
30680 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30682         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
30683         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
30684         munaligned-access.
30686 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30688         PR target/83681
30689         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
30690         by not USED_FOR_TARGET.
30691         (make_pass_resolve_sw_modes): Likewise.
30693 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30695         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
30696         USED_FOR_TARGET.
30698 2018-01-08  Richard Biener  <rguenther@suse.de>
30700         PR middle-end/83580
30701         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
30703 2018-01-08  Richard Biener  <rguenther@suse.de>
30705         PR middle-end/83517
30706         * match.pd ((t * 2) / 2) -> t): Add missing :c.
30708 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
30710         PR middle-end/81897
30711         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
30712         basic blocks with a small number of successors.
30713         (convert_control_dep_chain_into_preds): Improve handling of
30714         forwarder blocks.
30715         (dump_predicates): Split apart into...
30716         (dump_pred_chain): ...here...
30717         (dump_pred_info): ...and here.
30718         (can_one_predicate_be_invalidated_p): Add debugging printfs.
30719         (can_chain_union_be_invalidated_p): Improve check for invalidation
30720         of paths.
30721         (uninit_uses_cannot_happen): Avoid unnecessary if
30722         convert_control_dep_chain_into_preds yielded nothing.
30724 2018-01-06  Martin Sebor  <msebor@redhat.com>
30726         PR tree-optimization/83640
30727         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
30728         subtracting negative offset from size.
30729         (builtin_access::overlap): Adjust offset bounds of the access to fall
30730         within the size of the object if possible.
30732 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
30734         PR rtl-optimization/83699
30735         * expmed.c (extract_bit_field_1): Restrict the vector usage of
30736         extract_bit_field_as_subreg to cases in which the extracted
30737         value is also a vector.
30739         * lra-constraints.c (process_alt_operands): Test for the equivalence
30740         substitutions when detecting a possible reload cycle.
30742 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
30744         PR debug/83480
30745         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
30746         by default if flag_selective_schedling{,2}.  Formatting fixes.
30748         PR rtl-optimization/83682
30749         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
30750         if it has non-VECTOR_MODE element mode.
30751         (vec_duplicate_p): Likewise.
30753         PR middle-end/83694
30754         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
30755         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
30757 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30759         PR target/83604
30760         * config/i386/i386-builtin.def
30761         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
30762         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
30763         Require also OPTION_MASK_ISA_AVX512F in addition to
30764         OPTION_MASK_ISA_GFNI.
30765         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
30766         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
30767         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
30768         to OPTION_MASK_ISA_GFNI.
30769         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
30770         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
30771         OPTION_MASK_ISA_AVX512BW.
30772         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
30773         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
30774         addition to OPTION_MASK_ISA_GFNI.
30775         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
30776         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
30777         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
30778         to OPTION_MASK_ISA_GFNI.
30779         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
30780         a requirement for all ISAs rather than any of them with a few
30781         exceptions.
30782         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
30783         processing.
30784         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
30785         bitmasks to be enabled with 3 exceptions, instead of requiring any
30786         enabled ISA with lots of exceptions.
30787         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
30788         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
30789         Change avx512bw in isa attribute to avx512f.
30790         * config/i386/sgxintrin.h: Add license boilerplate.
30791         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
30792         to __AVX512F__ and __AVX512VL to __AVX512VL__.
30793         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
30794         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
30795         defined.
30796         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
30797         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
30798         temporarily sse2 rather than sse if not enabled already.
30800         PR target/83604
30801         * config/i386/sse.md (VI248_VLBW): Rename to ...
30802         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
30803         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
30804         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
30805         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
30806         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
30807         mode iterator instead of VI248_VLBW.
30809 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
30811         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
30812         (record_modified): Skip clobbers; add debug output.
30813         (param_change_prob): Use sreal frequencies.
30815 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30817         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
30818         punt for user-aligned variables.
30820 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30822         * tree-chrec.c (chrec_contains_symbols): Return true for
30823         POLY_INT_CST.
30825 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
30827         PR target/82439
30828         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
30829         of (x|y) == x for BICS pattern.
30831 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30833         PR tree-optimization/83605
30834         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
30835         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
30836         can throw.
30838 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30840         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
30841         * config/epiphany/rtems.h: New file.
30843 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30844             Uros Bizjak  <ubizjak@gmail.com>
30846         PR target/83554
30847         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
30848         QIreg_operand instead of register_operand predicate.
30849         * config/i386/i386.c (ix86_rop_should_change_byte_p,
30850         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
30851         comments instead of -fmitigate[-_]rop.
30853 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30855         PR bootstrap/81926
30856         * cgraphunit.c (symbol_table::compile): Switch to text_section
30857         before calling assembly_start debug hook.
30858         * run-rtl-passes.c (run_rtl_passes): Likewise.
30859         Include output.h.
30861 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30863         * tree-vrp.c (extract_range_from_binary_expr_1): Check
30864         range_int_cst_p rather than !symbolic_range_p before calling
30865         extract_range_from_multiplicative_op_1.
30867 2018-01-04  Jeff Law  <law@redhat.com>
30869         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
30870         redundant test in assertion.
30872 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30874         * doc/rtl.texi: Document machine_mode wrapper classes.
30876 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30878         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
30879         using tree_to_uhwi.
30881 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30883         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
30884         the VEC_PERM_EXPR fold to fail.
30886 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30888         PR debug/83585
30889         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
30890         to switched_sections.
30892 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30894         PR target/83680
30895         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
30896         test for d.testing.
30898 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
30900         PR target/83387
30901         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
30902         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
30904 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30906         PR debug/83666
30907         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
30908         is BLKmode and bitpos not zero or mode change is needed.
30910 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30912         PR target/83675
30913         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
30914         TARGET_VIS2.
30916 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
30918         PR target/83628
30919         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
30920         instead of MULT rtx.  Update all corresponding splitters.
30921         (*saddl_se): Ditto.
30922         (*ssub<modesuffix>): Ditto.
30923         (*ssubl_se): Ditto.
30924         (*cmp_sadd_di): Update split patterns.
30925         (*cmp_sadd_si): Ditto.
30926         (*cmp_sadd_sidi): Ditto.
30927         (*cmp_ssub_di): Ditto.
30928         (*cmp_ssub_si): Ditto.
30929         (*cmp_ssub_sidi): Ditto.
30930         * config/alpha/predicates.md (const23_operand): New predicate.
30931         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
30932         Look for ASHIFT, not MULT inner operand.
30933         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
30935 2018-01-04  Martin Liska  <mliska@suse.cz>
30937         PR gcov-profile/83669
30938         * gcov.c (output_intermediate_file): Add version to intermediate
30939         gcov file.
30940         * doc/gcov.texi: Document new field 'version' in intermediate
30941         file format. Fix location of '-k' option of gcov command.
30943 2018-01-04  Martin Liska  <mliska@suse.cz>
30945         PR ipa/82352
30946         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
30948 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30950         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
30952 2018-01-03  Martin Sebor  <msebor@redhat.com>
30954         PR tree-optimization/83655
30955         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
30956         checking calls with invalid arguments.
30958 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30960         * tree-vect-stmts.c (vect_get_store_rhs): New function.
30961         (vectorizable_mask_load_store): Delete.
30962         (vectorizable_call): Return false for masked loads and stores.
30963         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
30964         instead of gimple_assign_rhs1.
30965         (vectorizable_load): Handle IFN_MASK_LOAD.
30966         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
30968 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30970         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
30971         split out from..,
30972         (vectorizable_mask_load_store): ...here.
30973         (vectorizable_load): ...and here.
30975 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30977         * tree-vect-stmts.c (vect_build_all_ones_mask)
30978         (vect_build_zero_merge_argument): New functions, split out from...
30979         (vectorizable_load): ...here.
30981 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30983         * tree-vect-stmts.c (vect_check_store_rhs): New function,
30984         split out from...
30985         (vectorizable_mask_load_store): ...here.
30986         (vectorizable_store): ...and here.
30988 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30990         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
30991         split out from...
30992         (vectorizable_mask_load_store): ...here.
30994 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30996         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
30997         (vect_model_store_cost): Take a vec_load_store_type instead of a
30998         vect_def_type.
30999         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
31000         (vect_model_store_cost): Take a vec_load_store_type instead of a
31001         vect_def_type.
31002         (vectorizable_mask_load_store): Update accordingly.
31003         (vectorizable_store): Likewise.
31004         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
31006 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31008         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
31009         IFN_MASK_LOAD calls here rather than...
31010         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
31012 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31013             Alan Hayward  <alan.hayward@arm.com>
31014             David Sherwood  <david.sherwood@arm.com>
31016         * expmed.c (extract_bit_field_1): For vector extracts,
31017         fall back to extract_bit_field_as_subreg if vec_extract
31018         isn't available.
31020 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31021             Alan Hayward  <alan.hayward@arm.com>
31022             David Sherwood  <david.sherwood@arm.com>
31024         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
31025         they are variable or constant sized.
31026         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
31027         slots for constant-sized data.
31029 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31030             Alan Hayward  <alan.hayward@arm.com>
31031             David Sherwood  <david.sherwood@arm.com>
31033         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
31034         handling COND_EXPRs with boolean comparisons, try to find a better
31035         basis for the mask type than the boolean itself.
31037 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31039         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
31040         is calculated and how it can be overridden.
31041         * genmodes.c (max_bitsize_mode_any_mode): New variable.
31042         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
31043         if defined.
31044         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
31045         if nonzero.
31047 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31048             Alan Hayward  <alan.hayward@arm.com>
31049             David Sherwood  <david.sherwood@arm.com>
31051         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
31052         Remove the mode argument.
31053         (aarch64_simd_valid_immediate): Remove the mode and inverse
31054         arguments.
31055         * config/aarch64/iterators.md (bitsize): New iterator.
31056         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
31057         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
31058         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
31059         aarch64_simd_valid_immediate.
31060         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
31061         (aarch64_reg_or_bic_imm): Likewise.
31062         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
31063         with an insn_type enum and msl with a modifier_type enum.
31064         Replace element_width with a scalar_mode.  Change the shift
31065         to unsigned int.  Add constructors for scalar_float_mode and
31066         scalar_int_mode elements.
31067         (aarch64_vect_float_const_representable_p): Delete.
31068         (aarch64_can_const_movi_rtx_p)
31069         (aarch64_simd_scalar_immediate_valid_for_move)
31070         (aarch64_simd_make_constant): Update call to
31071         aarch64_simd_valid_immediate.
31072         (aarch64_advsimd_valid_immediate_hs): New function.
31073         (aarch64_advsimd_valid_immediate): Likewise.
31074         (aarch64_simd_valid_immediate): Remove mode and inverse
31075         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
31076         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
31077         and aarch64_float_const_representable_p on the result.
31078         (aarch64_output_simd_mov_immediate): Remove mode argument.
31079         Update call to aarch64_simd_valid_immediate and use of
31080         simd_immediate_info.
31081         (aarch64_output_scalar_simd_mov_immediate): Update call
31082         accordingly.
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         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
31089         (mode_nunits): Likewise CONST_MODE_NUNITS.
31090         * machmode.def (ADJUST_NUNITS): Document.
31091         * genmodes.c (mode_data::need_nunits_adj): New field.
31092         (blank_mode): Update accordingly.
31093         (adj_nunits): New variable.
31094         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
31095         parameter.
31096         (emit_mode_size_inline): Set need_bytesize_adj for all modes
31097         listed in adj_nunits.
31098         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
31099         listed in adj_nunits.  Don't emit case statements for such modes.
31100         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
31101         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
31102         nothing if adj_nunits is nonnull.
31103         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
31104         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
31105         (emit_mode_fbit): Update use of print_maybe_const_decl.
31106         (emit_move_size): Likewise.  Treat the array as non-const
31107         if adj_nunits.
31108         (emit_mode_adjustments): Handle adj_nunits.
31110 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31112         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
31113         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
31114         (VECTOR_MODES): Use it.
31115         (make_vector_modes): Take the prefix as an argument.
31117 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31118             Alan Hayward  <alan.hayward@arm.com>
31119             David Sherwood  <david.sherwood@arm.com>
31121         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
31122         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
31123         for MODE_VECTOR_BOOL.
31124         * machmode.def (VECTOR_BOOL_MODE): Document.
31125         * genmodes.c (VECTOR_BOOL_MODE): New macro.
31126         (make_vector_bool_mode): New function.
31127         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
31128         MODE_VECTOR_BOOL.
31129         * lto-streamer-in.c (lto_input_mode_table): Likewise.
31130         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
31131         Likewise.
31132         * stor-layout.c (int_mode_for_mode): Likewise.
31133         * tree.c (build_vector_type_for_mode): Likewise.
31134         * varasm.c (output_constant_pool_2): Likewise.
31135         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
31136         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
31137         for MODE_VECTOR_BOOL.
31138         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
31139         of mode class checks.
31140         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
31141         instead of a list of mode class checks.
31142         (expand_vector_scalar_condition): Likewise.
31143         (type_for_widest_vector_mode): Handle BImode as an inner mode.
31145 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31146             Alan Hayward  <alan.hayward@arm.com>
31147             David Sherwood  <david.sherwood@arm.com>
31149         * machmode.h (mode_size): Change from unsigned short to
31150         poly_uint16_pod.
31151         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
31152         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
31153         or if measurement_type is not polynomial.
31154         (fixed_size_mode::includes_p): Check for constant-sized modes.
31155         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
31156         return a poly_uint16 rather than an unsigned short.
31157         (emit_mode_size): Change the type of mode_size from unsigned short
31158         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
31159         (emit_mode_adjustments): Cope with polynomial vector sizes.
31160         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31161         for GET_MODE_SIZE.
31162         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31163         for GET_MODE_SIZE.
31164         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
31165         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
31166         * caller-save.c (setup_save_areas): Likewise.
31167         (replace_reg_with_saved_mem): Likewise.
31168         * calls.c (emit_library_call_value_1): Likewise.
31169         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
31170         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
31171         (gen_lowpart_for_combine): Likewise.
31172         * convert.c (convert_to_integer_1): Likewise.
31173         * cse.c (equiv_constant, cse_insn): Likewise.
31174         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
31175         (cselib_subst_to_values): Likewise.
31176         * dce.c (word_dce_process_block): Likewise.
31177         * df-problems.c (df_word_lr_mark_ref): Likewise.
31178         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
31179         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
31180         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
31181         (rtl_for_decl_location): Likewise.
31182         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
31183         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
31184         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
31185         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
31186         (expand_expr_real_1): Likewise.
31187         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
31188         (pad_below): Likewise.
31189         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31190         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
31191         * ira.c (get_subreg_tracking_sizes): Likewise.
31192         * ira-build.c (ira_create_allocno_objects): Likewise.
31193         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
31194         (ira_sort_regnos_for_alter_reg): Likewise.
31195         * ira-costs.c (record_operand_costs): Likewise.
31196         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
31197         (resolve_simple_move): Likewise.
31198         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
31199         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
31200         (lra_constraints): Likewise.
31201         (CONST_POOL_OK_P): Reject variable-sized modes.
31202         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
31203         (add_pseudo_to_slot, lra_spill): Likewise.
31204         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31205         * optabs-query.c (get_best_extraction_insn): Likewise.
31206         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31207         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
31208         (expand_mult_highpart, valid_multiword_target_p): Likewise.
31209         * recog.c (offsettable_address_addr_space_p): Likewise.
31210         * regcprop.c (maybe_mode_change): Likewise.
31211         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
31212         * regrename.c (build_def_use): Likewise.
31213         * regstat.c (dump_reg_info): Likewise.
31214         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
31215         (find_reloads, find_reloads_subreg_address): Likewise.
31216         * reload1.c (eliminate_regs_1): Likewise.
31217         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
31218         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
31219         (simplify_binary_operation_1, simplify_subreg): Likewise.
31220         * targhooks.c (default_function_arg_padding): Likewise.
31221         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
31222         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
31223         (verify_gimple_assign_ternary): Likewise.
31224         * tree-inline.c (estimate_move_cost): Likewise.
31225         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31226         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
31227         (get_address_cost_ainc): Likewise.
31228         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
31229         (vect_supportable_dr_alignment): Likewise.
31230         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31231         (vectorizable_reduction): Likewise.
31232         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
31233         (vectorizable_operation, vectorizable_load): Likewise.
31234         * tree.c (build_same_sized_truth_vector_type): Likewise.
31235         * valtrack.c (cleanup_auto_inc_dec): Likewise.
31236         * var-tracking.c (emit_note_insn_var_location): Likewise.
31237         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
31238         (ADDR_VEC_ALIGN): Likewise.
31240 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31241             Alan Hayward  <alan.hayward@arm.com>
31242             David Sherwood  <david.sherwood@arm.com>
31244         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
31245         unsigned short.
31246         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
31247         or if measurement_type is polynomial.
31248         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
31249         * combine.c (make_extraction): Likewise.
31250         * dse.c (find_shift_sequence): Likewise.
31251         * dwarf2out.c (mem_loc_descriptor): Likewise.
31252         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
31253         (extract_bit_field, extract_low_bits): Likewise.
31254         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
31255         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
31256         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
31257         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
31258         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31259         * reload.c (find_reloads): Likewise.
31260         * reload1.c (alter_reg): Likewise.
31261         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
31262         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
31263         * tree-if-conv.c (predicate_mem_writes): Likewise.
31264         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
31265         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
31266         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
31267         * valtrack.c (dead_debug_insert_temp): Likewise.
31268         * varasm.c (mergeable_constant_section): Likewise.
31269         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
31271 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31272             Alan Hayward  <alan.hayward@arm.com>
31273             David Sherwood  <david.sherwood@arm.com>
31275         * expr.c (expand_assignment): Cope with polynomial mode sizes
31276         when assigning to a CONCAT.
31278 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31279             Alan Hayward  <alan.hayward@arm.com>
31280             David Sherwood  <david.sherwood@arm.com>
31282         * machmode.h (mode_precision): Change from unsigned short to
31283         poly_uint16_pod.
31284         (mode_to_precision): Return a poly_uint16 rather than an unsigned
31285         short.
31286         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
31287         or if measurement_type is not polynomial.
31288         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
31289         in which the mode is already known to be a scalar_int_mode.
31290         * genmodes.c (emit_mode_precision): Change the type of mode_precision
31291         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
31292         initializer.
31293         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31294         for GET_MODE_PRECISION.
31295         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31296         for GET_MODE_PRECISION.
31297         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
31298         as polynomial.
31299         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
31300         (expand_field_assignment, make_extraction): Likewise.
31301         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
31302         (get_last_value): Likewise.
31303         * convert.c (convert_to_integer_1): Likewise.
31304         * cse.c (cse_insn): Likewise.
31305         * expr.c (expand_expr_real_1): Likewise.
31306         * lra-constraints.c (simplify_operand_subreg): Likewise.
31307         * optabs-query.c (can_atomic_load_p): Likewise.
31308         * optabs.c (expand_atomic_load): Likewise.
31309         (expand_atomic_store): Likewise.
31310         * ree.c (combine_reaching_defs): Likewise.
31311         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
31312         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
31313         * tree.h (type_has_mode_precision_p): Likewise.
31314         * ubsan.c (instrument_si_overflow): Likewise.
31316 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31317             Alan Hayward  <alan.hayward@arm.com>
31318             David Sherwood  <david.sherwood@arm.com>
31320         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
31321         polynomial numbers of units.
31322         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
31323         (valid_vector_subparts_p): New function.
31324         (build_vector_type): Remove temporary shim and take the number
31325         of units as a poly_uint64 rather than an int.
31326         (build_opaque_vector_type): Take the number of units as a
31327         poly_uint64 rather than an int.
31328         * tree.c (build_vector_from_ctor): Handle polynomial
31329         TYPE_VECTOR_SUBPARTS.
31330         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
31331         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
31332         (build_vector_from_val): If the number of units is variable,
31333         use build_vec_duplicate_cst for constant operands and
31334         VEC_DUPLICATE_EXPR otherwise.
31335         (make_vector_type): Remove temporary is_constant ().
31336         (build_vector_type, build_opaque_vector_type): Take the number of
31337         units as a poly_uint64 rather than an int.
31338         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
31339         VECTOR_CST_NELTS.
31340         * cfgexpand.c (expand_debug_expr): Likewise.
31341         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
31342         (store_constructor, expand_expr_real_1): Likewise.
31343         (const_scalar_mask_from_tree): Likewise.
31344         * fold-const-call.c (fold_const_reduction): Likewise.
31345         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
31346         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
31347         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
31348         (fold_relational_const): Likewise.
31349         (native_interpret_vector): Likewise.  Change the size from an
31350         int to an unsigned int.
31351         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
31352         TYPE_VECTOR_SUBPARTS.
31353         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
31354         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
31355         duplicating a non-constant operand into a variable-length vector.
31356         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
31357         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
31358         * ipa-icf.c (sem_variable::equals): Likewise.
31359         * match.pd: Likewise.
31360         * omp-simd-clone.c (simd_clone_subparts): Likewise.
31361         * print-tree.c (print_node): Likewise.
31362         * stor-layout.c (layout_type): Likewise.
31363         * targhooks.c (default_builtin_vectorization_cost): Likewise.
31364         * tree-cfg.c (verify_gimple_comparison): Likewise.
31365         (verify_gimple_assign_binary): Likewise.
31366         (verify_gimple_assign_ternary): Likewise.
31367         (verify_gimple_assign_single): Likewise.
31368         * tree-pretty-print.c (dump_generic_node): Likewise.
31369         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31370         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
31371         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
31372         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
31373         (vect_shift_permute_load_chain): Likewise.
31374         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
31375         (expand_vector_condition, optimize_vector_constructor): Likewise.
31376         (lower_vec_perm, get_compute_type): Likewise.
31377         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31378         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
31379         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
31380         (vect_recog_mask_conversion_pattern): Likewise.
31381         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
31382         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
31383         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31384         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
31385         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
31386         (vectorizable_shift, vectorizable_operation, vectorizable_store)
31387         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
31388         (supportable_widening_operation): Likewise.
31389         (supportable_narrowing_operation): Likewise.
31390         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
31391         Likewise.
31392         * varasm.c (output_constant): Likewise.
31394 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31395             Alan Hayward  <alan.hayward@arm.com>
31396             David Sherwood  <david.sherwood@arm.com>
31398         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
31399         so that both the length == 3 and length != 3 cases set up their
31400         own permute vectors.  Add comments explaining why we know the
31401         number of elements is constant.
31402         (vect_permute_load_chain): Likewise.
31404 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31405             Alan Hayward  <alan.hayward@arm.com>
31406             David Sherwood  <david.sherwood@arm.com>
31408         * machmode.h (mode_nunits): Change from unsigned char to
31409         poly_uint16_pod.
31410         (ONLY_FIXED_SIZE_MODES): New macro.
31411         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
31412         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
31413         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
31414         New typedefs.
31415         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
31416         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
31417         or if measurement_type is not polynomial.
31418         * genmodes.c (ZERO_COEFFS): New macro.
31419         (emit_mode_nunits_inline): Make mode_nunits_inline return a
31420         poly_uint16.
31421         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
31422         Use ZERO_COEFFS when emitting initializers.
31423         * data-streamer.h (bp_pack_poly_value): New function.
31424         (bp_unpack_poly_value): Likewise.
31425         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31426         for GET_MODE_NUNITS.
31427         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31428         for GET_MODE_NUNITS.
31429         * tree.c (make_vector_type): Remove temporary shim and make
31430         the real function take the number of units as a poly_uint64
31431         rather than an int.
31432         (build_vector_type_for_mode): Handle polynomial nunits.
31433         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
31434         * emit-rtl.c (const_vec_series_p_1): Likewise.
31435         (gen_rtx_CONST_VECTOR): Likewise.
31436         * fold-const.c (test_vec_duplicate_folding): Likewise.
31437         * genrecog.c (validate_pattern): Likewise.
31438         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
31439         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31440         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
31441         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
31442         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
31443         * rtlanal.c (subreg_get_info): Likewise.
31444         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31445         (vect_grouped_load_supported): Likewise.
31446         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
31447         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
31448         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
31449         (simplify_const_unary_operation, simplify_binary_operation_1)
31450         (simplify_const_binary_operation, simplify_ternary_operation)
31451         (test_vector_ops_duplicate, test_vector_ops): Likewise.
31452         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
31453         instead of CONST_VECTOR_NUNITS.
31454         * varasm.c (output_constant_pool_2): Likewise.
31455         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
31456         explicit-encoded elements in the XVEC for variable-length vectors.
31458 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31460         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
31462 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31463             Alan Hayward  <alan.hayward@arm.com>
31464             David Sherwood  <david.sherwood@arm.com>
31466         * coretypes.h (fixed_size_mode): Declare.
31467         (fixed_size_mode_pod): New typedef.
31468         * builtins.h (target_builtins::x_apply_args_mode)
31469         (target_builtins::x_apply_result_mode): Change type to
31470         fixed_size_mode_pod.
31471         * builtins.c (apply_args_size, apply_result_size, result_vector)
31472         (expand_builtin_apply_args_1, expand_builtin_apply)
31473         (expand_builtin_return): Update accordingly.
31475 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31477         * cse.c (hash_rtx_cb): Hash only the encoded elements.
31478         * cselib.c (cselib_hash_rtx): Likewise.
31479         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
31480         CONST_VECTOR encoding.
31482 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31483             Jeff Law  <law@redhat.com>
31485         PR target/83641
31486         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
31487         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
31488         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
31489         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
31491         PR target/83641
31492         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
31493         explicitly probe *sp in a noreturn function if there were any callee
31494         register saves or frame pointer is needed.
31496 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31498         PR debug/83621
31499         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
31500         BLKmode for ternary, binary or unary expressions.
31502         PR debug/83645
31503         * var-tracking.c (delete_vta_debug_insn): New inline function.
31504         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
31505         insns from get_insns () to NULL instead of each bb separately.
31506         Use delete_vta_debug_insn.  No longer static.
31507         (vt_debug_insns_local, variable_tracking_main_1): Adjust
31508         delete_vta_debug_insns callers.
31509         * rtl.h (delete_vta_debug_insns): Declare.
31510         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
31511         instead of variable_tracking_main.
31513 2018-01-03  Martin Sebor  <msebor@redhat.com>
31515         PR tree-optimization/83603
31516         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
31517         arguments past the endof the argument list in functions declared
31518         without a prototype.
31519         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
31520         Avoid checking when arguments are null.
31522 2018-01-03  Martin Sebor  <msebor@redhat.com>
31524         PR c/83559
31525         * doc/extend.texi (attribute const): Fix a typo.
31526         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
31527         issuing -Wsuggest-attribute for void functions.
31529 2018-01-03  Martin Sebor  <msebor@redhat.com>
31531         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
31532         offset_int::from instead of wide_int::to_shwi.
31533         (maybe_diag_overlap): Remove assertion.
31534         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
31535         * gimple-ssa-sprintf.c (format_directive): Same.
31536         (parse_directive): Same.
31537         (sprintf_dom_walker::compute_format_length): Same.
31538         (try_substitute_return_value): Same.
31540 2018-01-03  Jeff Law  <law@redhat.com>
31542         PR middle-end/83654
31543         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
31544         non-constant residual for zero at runtime and avoid probing in
31545         that case.  Reorganize code for trailing problem to mirror handling
31546         of the residual.
31548 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
31550         PR tree-optimization/83501
31551         * tree-ssa-strlen.c (get_string_cst): New.
31552         (handle_char_store): Call get_string_cst.
31554 2018-01-03  Martin Liska  <mliska@suse.cz>
31556         PR tree-optimization/83593
31557         * tree-ssa-strlen.c: Include tree-cfg.h.
31558         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
31559         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
31560         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
31561         to false.
31562         (strlen_dom_walker::before_dom_children): Call
31563         gimple_purge_dead_eh_edges. Dump tranformation with details
31564         dump flags.
31565         (strlen_dom_walker::before_dom_children): Update call by adding
31566         new argument cleanup_eh.
31567         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
31569 2018-01-03  Martin Liska  <mliska@suse.cz>
31571         PR ipa/83549
31572         * cif-code.def (VARIADIC_THUNK): New enum value.
31573         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
31574         thunks.
31576 2018-01-03  Jan Beulich  <jbeulich@suse.com>
31578         * sse.md (mov<mode>_internal): Tighten condition for when to use
31579         vmovdqu<ssescalarsize> for TI and OI modes.
31581 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31583         Update copyright years.
31585 2018-01-03  Martin Liska  <mliska@suse.cz>
31587         PR ipa/83594
31588         * ipa-visibility.c (function_and_variable_visibility): Skip
31589         functions with noipa attribure.
31591 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31593         * gcc.c (process_command): Update copyright notice dates.
31594         * gcov-dump.c (print_version): Ditto.
31595         * gcov.c (print_version): Ditto.
31596         * gcov-tool.c (print_version): Ditto.
31597         * gengtype.c (create_file): Ditto.
31598         * doc/cpp.texi: Bump @copying's copyright year.
31599         * doc/cppinternals.texi: Ditto.
31600         * doc/gcc.texi: Ditto.
31601         * doc/gccint.texi: Ditto.
31602         * doc/gcov.texi: Ditto.
31603         * doc/install.texi: Ditto.
31604         * doc/invoke.texi: Ditto.
31606 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31608         * vector-builder.h (vector_builder::m_full_nelts): Change from
31609         unsigned int to poly_uint64.
31610         (vector_builder::full_nelts): Update prototype accordingly.
31611         (vector_builder::new_vector): Likewise.
31612         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
31613         (vector_builder::operator ==): Likewise.
31614         (vector_builder::finalize): Likewise.
31615         * int-vector-builder.h (int_vector_builder::int_vector_builder):
31616         Take the number of elements as a poly_uint64 rather than an
31617         unsigned int.
31618         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
31619         from unsigned int to poly_uint64.
31620         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
31621         (vec_perm_indices::new_vector): Likewise.
31622         (vec_perm_indices::length): Likewise.
31623         (vec_perm_indices::nelts_per_input): Likewise.
31624         (vec_perm_indices::input_nelts): Likewise.
31625         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
31626         number of elements per input as a poly_uint64 rather than an
31627         unsigned int.  Use the original encoding for variable-length
31628         vectors, rather than clamping each individual element.
31629         For the second and subsequent elements in each pattern,
31630         clamp the step and base before clamping their sum.
31631         (vec_perm_indices::series_p): Handle polynomial element counts.
31632         (vec_perm_indices::all_in_range_p): Likewise.
31633         (vec_perm_indices_to_tree): Likewise.
31634         (vec_perm_indices_to_rtx): Likewise.
31635         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
31636         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
31637         (tree_vector_builder::new_binary_operation): Handle polynomial
31638         element counts.  Return false if we need to know the number
31639         of elements at compile time.
31640         * fold-const.c (fold_vec_perm): Punt if the number of elements
31641         isn't known at compile time.
31643 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31645         * vec-perm-indices.h (vec_perm_builder): Change element type
31646         from HOST_WIDE_INT to poly_int64.
31647         (vec_perm_indices::element_type): Update accordingly.
31648         (vec_perm_indices::clamp): Handle polynomial element_types.
31649         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31650         (vec_perm_indices::all_in_range_p): Likewise.
31651         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
31652         than shwi trees.
31653         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
31654         polynomial vec_perm_indices element types.
31655         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
31656         * fold-const.c (fold_vec_perm): Likewise.
31657         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
31658         * tree-vect-generic.c (lower_vec_perm): Likewise.
31659         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31660         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
31661         element type to HOST_WIDE_INT.
31663 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31664             Alan Hayward  <alan.hayward@arm.com>
31665             David Sherwood  <david.sherwood@arm.com>
31667         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
31668         rather than an int.  Use plus_constant.
31669         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
31670         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
31672 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31673             Alan Hayward  <alan.hayward@arm.com>
31674             David Sherwood  <david.sherwood@arm.com>
31676         * calls.c (emit_call_1, expand_call): Change struct_value_size from
31677         a HOST_WIDE_INT to a poly_int64.
31679 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31680             Alan Hayward  <alan.hayward@arm.com>
31681             David Sherwood  <david.sherwood@arm.com>
31683         * calls.c (load_register_parameters): Cope with polynomial
31684         mode sizes.  Require a constant size for BLKmode parameters
31685         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
31686         forces a parameter to be padded at the lsb end in order to
31687         fill a complete number of words, require the parameter size
31688         to be ordered wrt UNITS_PER_WORD.
31690 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31691             Alan Hayward  <alan.hayward@arm.com>
31692             David Sherwood  <david.sherwood@arm.com>
31694         * reload1.c (spill_stack_slot_width): Change element type
31695         from unsigned int to poly_uint64_pod.
31696         (alter_reg): Treat mode sizes as polynomial.
31698 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31699             Alan Hayward  <alan.hayward@arm.com>
31700             David Sherwood  <david.sherwood@arm.com>
31702         * reload.c (complex_word_subreg_p): New function.
31703         (reload_inner_reg_of_subreg, push_reload): Use it.
31705 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31706             Alan Hayward  <alan.hayward@arm.com>
31707             David Sherwood  <david.sherwood@arm.com>
31709         * lra-constraints.c (process_alt_operands): Reject matched
31710         operands whose sizes aren't ordered.
31711         (match_reload): Refer to this check here.
31713 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31714             Alan Hayward  <alan.hayward@arm.com>
31715             David Sherwood  <david.sherwood@arm.com>
31717         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
31718         that the mode size is in the set {1, 2, 4, 8, 16}.
31720 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31721             Alan Hayward  <alan.hayward@arm.com>
31722             David Sherwood  <david.sherwood@arm.com>
31724         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
31725         Use plus_constant instead of gen_rtx_PLUS.
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         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
31732         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
31733         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
31734         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
31735         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
31736         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
31737         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
31738         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
31739         * config/i386/i386.c (ix86_push_rounding): ...this new function.
31740         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
31741         a poly_int64.
31742         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
31743         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
31744         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
31745         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
31746         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
31747         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
31748         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
31749         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
31750         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
31751         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
31752         function.
31753         * expr.c (emit_move_resolve_push): Treat the input and result
31754         of PUSH_ROUNDING as a poly_int64.
31755         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
31756         (emit_push_insn): Likewise.
31757         * lra-eliminations.c (mark_not_eliminable): Likewise.
31758         * recog.c (push_operand): Likewise.
31759         * reload1.c (elimination_effects): Likewise.
31760         * rtlanal.c (nonzero_bits1): Likewise.
31761         * calls.c (store_one_arg): Likewise.  Require the padding to be
31762         known at compile time.
31764 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31765             Alan Hayward  <alan.hayward@arm.com>
31766             David Sherwood  <david.sherwood@arm.com>
31768         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
31769         Use plus_constant instead of gen_rtx_PLUS.
31771 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31772             Alan Hayward  <alan.hayward@arm.com>
31773             David Sherwood  <david.sherwood@arm.com>
31775         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
31776         rather than an int.
31778 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31779             Alan Hayward  <alan.hayward@arm.com>
31780             David Sherwood  <david.sherwood@arm.com>
31782         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
31783         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
31784         via stack temporaries.  Treat the mode size as polynomial too.
31786 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31787             Alan Hayward  <alan.hayward@arm.com>
31788             David Sherwood  <david.sherwood@arm.com>
31790         * expr.c (expand_expr_real_2): When handling conversions involving
31791         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
31792         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
31793         as a poly_uint64 too.
31795 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31796             Alan Hayward  <alan.hayward@arm.com>
31797             David Sherwood  <david.sherwood@arm.com>
31799         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
31801 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31802             Alan Hayward  <alan.hayward@arm.com>
31803             David Sherwood  <david.sherwood@arm.com>
31805         * combine.c (can_change_dest_mode): Handle polynomial
31806         REGMODE_NATURAL_SIZE.
31807         * expmed.c (store_bit_field_1): Likewise.
31808         * expr.c (store_constructor): Likewise.
31809         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
31810         and polynomial REGMODE_NATURAL_SIZE.
31811         (gen_lowpart_common): Likewise.
31812         * reginfo.c (record_subregs_of_mode): Likewise.
31813         * rtlanal.c (read_modify_subreg_p): Likewise.
31815 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31816             Alan Hayward  <alan.hayward@arm.com>
31817             David Sherwood  <david.sherwood@arm.com>
31819         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
31820         numbers of elements.
31822 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31823             Alan Hayward  <alan.hayward@arm.com>
31824             David Sherwood  <david.sherwood@arm.com>
31826         * match.pd: Cope with polynomial numbers of vector elements.
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         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
31833         in a POINTER_PLUS_EXPR.
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         * omp-simd-clone.c (simd_clone_subparts): New function.
31840         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
31841         (ipa_simd_modify_function_body): Likewise.
31843 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31844             Alan Hayward  <alan.hayward@arm.com>
31845             David Sherwood  <david.sherwood@arm.com>
31847         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
31848         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
31849         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
31850         (expand_vector_condition, vector_element): Likewise.
31851         (subparts_gt): New function.
31852         (get_compute_type): Use subparts_gt.
31853         (count_type_subparts): Delete.
31854         (expand_vector_operations_1): Use subparts_gt instead of
31855         count_type_subparts.
31857 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31858             Alan Hayward  <alan.hayward@arm.com>
31859             David Sherwood  <david.sherwood@arm.com>
31861         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
31862         (vect_compile_time_alias): ...this new function.  Do the calculation
31863         on poly_ints rather than trees.
31864         (vect_prune_runtime_alias_test_list): Update call accordingly.
31866 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31867             Alan Hayward  <alan.hayward@arm.com>
31868             David Sherwood  <david.sherwood@arm.com>
31870         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
31871         numbers of units.
31872         (vect_schedule_slp_instance): Likewise.
31874 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31875             Alan Hayward  <alan.hayward@arm.com>
31876             David Sherwood  <david.sherwood@arm.com>
31878         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
31879         constant and extern definitions for variable-length vectors.
31880         (vect_get_constant_vectors): Note that the number of units
31881         is known to be constant.
31883 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31884             Alan Hayward  <alan.hayward@arm.com>
31885             David Sherwood  <david.sherwood@arm.com>
31887         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
31888         of units as polynomial.  Choose between WIDE and NARROW based
31889         on multiple_p.
31891 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31892             Alan Hayward  <alan.hayward@arm.com>
31893             David Sherwood  <david.sherwood@arm.com>
31895         * tree-vect-stmts.c (simd_clone_subparts): New function.
31896         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
31898 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31899             Alan Hayward  <alan.hayward@arm.com>
31900             David Sherwood  <david.sherwood@arm.com>
31902         * tree-vect-stmts.c (vectorizable_call): Treat the number of
31903         vectors as polynomial.  Use build_index_vector for
31904         IFN_GOMP_SIMD_LANE.
31906 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31907             Alan Hayward  <alan.hayward@arm.com>
31908             David Sherwood  <david.sherwood@arm.com>
31910         * tree-vect-stmts.c (get_load_store_type): Treat the number of
31911         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
31912         for variable-length vectors.
31913         (vectorizable_mask_load_store): Treat the number of units as
31914         polynomial, asserting that it is constant if the condition has
31915         already been enforced.
31916         (vectorizable_store, vectorizable_load): Likewise.
31918 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31919             Alan Hayward  <alan.hayward@arm.com>
31920             David Sherwood  <david.sherwood@arm.com>
31922         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
31923         of units as polynomial.  Punt if we can't tell at compile time
31924         which vector contains the final result.
31926 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31927             Alan Hayward  <alan.hayward@arm.com>
31928             David Sherwood  <david.sherwood@arm.com>
31930         * tree-vect-loop.c (vectorizable_induction): Treat the number
31931         of units as polynomial.  Punt on SLP inductions.  Use an integer
31932         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
31933         cast of such a series for variable-length floating-point
31934         reductions.
31936 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31937             Alan Hayward  <alan.hayward@arm.com>
31938             David Sherwood  <david.sherwood@arm.com>
31940         * tree.h (build_index_vector): Declare.
31941         * tree.c (build_index_vector): New function.
31942         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
31943         of units as polynomial, forcibly converting it to a constant if
31944         vectorizable_reduction has already enforced the condition.
31945         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
31946         to create a {1,2,3,...} vector.
31947         (vectorizable_reduction): Treat the number of units as polynomial.
31948         Choose vectype_in based on the largest scalar element size rather
31949         than the smallest number of units.  Enforce the restrictions
31950         relied on above.
31952 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31953             Alan Hayward  <alan.hayward@arm.com>
31954             David Sherwood  <david.sherwood@arm.com>
31956         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
31957         number of units as polynomial.
31959 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31960             Alan Hayward  <alan.hayward@arm.com>
31961             David Sherwood  <david.sherwood@arm.com>
31963         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
31964         * target.def (autovectorize_vector_sizes): Return the vector sizes
31965         by pointer, using vector_sizes rather than a bitmask.
31966         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
31967         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
31968         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
31969         Likewise.
31970         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
31971         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
31972         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
31973         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
31974         * omp-general.c (omp_max_vf): Likewise.
31975         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31976         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
31977         * tree-vect-loop.c (vect_analyze_loop): Likewise.
31978         * tree-vect-slp.c (vect_slp_bb): Likewise.
31979         * doc/tm.texi: Regenerate.
31980         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
31981         to a poly_uint64.
31982         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
31983         the vector size as a poly_uint64 rather than an unsigned int.
31984         (current_vector_size): Change from an unsigned int to a poly_uint64.
31985         (get_vectype_for_scalar_type): Update accordingly.
31986         * tree.h (build_truth_vector_type): Take the size and number of
31987         units as a poly_uint64 rather than an unsigned int.
31988         (build_vector_type): Add a temporary overload that takes
31989         the number of units as a poly_uint64 rather than an unsigned int.
31990         * tree.c (make_vector_type): Likewise.
31991         (build_truth_vector_type): Take the number of units as a poly_uint64
31992         rather than an unsigned int.
31994 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31995             Alan Hayward  <alan.hayward@arm.com>
31996             David Sherwood  <david.sherwood@arm.com>
31998         * target.def (get_mask_mode): Take the number of units and length
31999         as poly_uint64s rather than unsigned ints.
32000         * targhooks.h (default_get_mask_mode): Update accordingly.
32001         * targhooks.c (default_get_mask_mode): Likewise.
32002         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
32003         * doc/tm.texi: Regenerate.
32005 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32006             Alan Hayward  <alan.hayward@arm.com>
32007             David Sherwood  <david.sherwood@arm.com>
32009         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
32010         * omp-general.c (omp_max_vf): Likewise.
32011         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
32012         (expand_omp_simd): Handle polynomial safelen.
32013         * omp-low.c (omplow_simd_context): Add a default constructor.
32014         (omplow_simd_context::max_vf): Change from int to poly_uint64.
32015         (lower_rec_simd_input_clauses): Update accordingly.
32016         (lower_rec_input_clauses): Likewise.
32018 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32019             Alan Hayward  <alan.hayward@arm.com>
32020             David Sherwood  <david.sherwood@arm.com>
32022         * tree-vectorizer.h (vect_nunits_for_cost): New function.
32023         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
32024         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
32025         (vect_analyze_slp_cost): Likewise.
32026         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
32027         (vect_model_load_cost): Likewise.
32029 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32030             Alan Hayward  <alan.hayward@arm.com>
32031             David Sherwood  <david.sherwood@arm.com>
32033         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
32034         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
32035         from an unsigned int * to a poly_uint64_pod *.
32036         (calculate_unrolling_factor): New function.
32037         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
32039 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32040             Alan Hayward  <alan.hayward@arm.com>
32041             David Sherwood  <david.sherwood@arm.com>
32043         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
32044         from an unsigned int to a poly_uint64.
32045         (_loop_vec_info::slp_unrolling_factor): Likewise.
32046         (_loop_vec_info::vectorization_factor): Change from an int
32047         to a poly_uint64.
32048         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
32049         (vect_get_num_vectors): New function.
32050         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
32051         (vect_get_num_copies): Use vect_get_num_vectors.
32052         (vect_analyze_data_ref_dependences): Change max_vf from an int *
32053         to an unsigned int *.
32054         (vect_analyze_data_refs): Change min_vf from an int * to a
32055         poly_uint64 *.
32056         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
32057         than an unsigned HOST_WIDE_INT.
32058         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
32059         (vect_analyze_data_ref_dependence): Change max_vf from an int *
32060         to an unsigned int *.
32061         (vect_analyze_data_ref_dependences): Likewise.
32062         (vect_compute_data_ref_alignment): Handle polynomial vf.
32063         (vect_enhance_data_refs_alignment): Likewise.
32064         (vect_prune_runtime_alias_test_list): Likewise.
32065         (vect_shift_permute_load_chain): Likewise.
32066         (vect_supportable_dr_alignment): Likewise.
32067         (dependence_distance_ge_vf): Take the vectorization factor as a
32068         poly_uint64 rather than an unsigned HOST_WIDE_INT.
32069         (vect_analyze_data_refs): Change min_vf from an int * to a
32070         poly_uint64 *.
32071         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
32072         vfm1 as a poly_uint64 rather than an int.  Make the same change
32073         for the returned bound_scalar.
32074         (vect_gen_vector_loop_niters): Handle polynomial vf.
32075         (vect_do_peeling): Likewise.  Update call to
32076         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
32077         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
32078         be constant.
32079         * tree-vect-loop.c (vect_determine_vectorization_factor)
32080         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
32081         (vect_get_known_peeling_cost): Likewise.
32082         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
32083         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
32084         (vect_transform_loop): Likewise.  Use the lowest possible VF when
32085         updating the upper bounds of the loop.
32086         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
32087         rather than an int.
32088         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
32089         polynomial unroll factors.
32090         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
32091         (vect_make_slp_decision): Likewise.
32092         (vect_supported_load_permutation_p): Likewise, and polynomial
32093         vf too.
32094         (vect_analyze_slp_cost): Handle polynomial vf.
32095         (vect_slp_analyze_node_operations): Likewise.
32096         (vect_slp_analyze_bb_1): Likewise.
32097         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
32098         than an unsigned HOST_WIDE_INT.
32099         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
32100         (vectorizable_load): Handle polynomial vf.
32101         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
32102         a poly_uint64.
32103         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
32105 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32106             Alan Hayward  <alan.hayward@arm.com>
32107             David Sherwood  <david.sherwood@arm.com>
32109         * match.pd: Handle bit operations involving three constants
32110         and try to fold one pair.
32112 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32114         * tree-vect-loop-manip.c: Include gimple-fold.h.
32115         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
32116         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
32117         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
32118         Add a path that uses a step of VF instead of 1, but disable it
32119         for now.
32120         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
32121         and niters_no_overflow parameters.  Update calls to
32122         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
32123         Create a new SSA name if the latter choses to use a ste other
32124         than zero, and return it via niters_vector_mult_vf_var.
32125         * tree-vect-loop.c (vect_transform_loop): Update calls to
32126         vect_do_peeling, vect_gen_vector_loop_niters and
32127         slpeel_make_loop_iterate_ntimes.
32128         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
32129         (vect_gen_vector_loop_niters): Update declarations after above changes.
32131 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
32133         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
32134         128-bit round to integer instructions.
32135         (ceil<mode>2): Likewise.
32136         (btrunc<mode>2): Likewise.
32137         (round<mode>2): Likewise.
32139 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32141         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
32142         unaligned VSX load/store on P8/P9.
32143         (expand_block_clear): Allow the use of unaligned VSX
32144         load/store on P8/P9.
32146 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32148         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
32149         New function.
32150         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
32151         swap associated with both a load and a store.
32153 2018-01-02  Andrew Waterman  <andrew@sifive.com>
32155         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
32156         * config/riscv/riscv.md (clear_cache): Use it.
32158 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
32160         * web.c: Remove out-of-date comment.
32162 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32164         * expr.c (fixup_args_size_notes): Check that any existing
32165         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
32166         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
32167         (emit_single_push_insn): ...here.
32169 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32171         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
32172         (const_vector_encoded_nelts): New function.
32173         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
32174         (const_vector_int_elt, const_vector_elt): Declare.
32175         * emit-rtl.c (const_vector_int_elt_1): New function.
32176         (const_vector_elt): Likewise.
32177         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
32178         of CONST_VECTOR_ELT.
32180 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32182         * expr.c: Include rtx-vector-builder.h.
32183         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
32184         directly on the tree encoding.
32185         (const_vector_from_tree): Likewise.
32186         * optabs.c: Include rtx-vector-builder.h.
32187         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
32188         sequence of "u" values.
32189         * vec-perm-indices.c: Include rtx-vector-builder.h.
32190         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
32191         directly on the vec_perm_indices encoding.
32193 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32195         * doc/rtl.texi (const_vector): Describe new encoding scheme.
32196         * Makefile.in (OBJS): Add rtx-vector-builder.o.
32197         * rtx-vector-builder.h: New file.
32198         * rtx-vector-builder.c: Likewise.
32199         * rtl.h (rtx_def::u2): Add a const_vector field.
32200         (CONST_VECTOR_NPATTERNS): New macro.
32201         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
32202         (CONST_VECTOR_DUPLICATE_P): Likewise.
32203         (CONST_VECTOR_STEPPED_P): Likewise.
32204         (CONST_VECTOR_ENCODED_ELT): Likewise.
32205         (const_vec_duplicate_p): Check for a duplicated vector encoding.
32206         (unwrap_const_vec_duplicate): Likewise.
32207         (const_vec_series_p): Check for a non-duplicated vector encoding.
32208         Say that the function only returns true for integer vectors.
32209         * emit-rtl.c: Include rtx-vector-builder.h.
32210         (gen_const_vec_duplicate_1): Delete.
32211         (gen_const_vector): Call gen_const_vec_duplicate instead of
32212         gen_const_vec_duplicate_1.
32213         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
32214         (gen_const_vec_duplicate): Use rtx_vector_builder.
32215         (gen_const_vec_series): Likewise.
32216         (gen_rtx_CONST_VECTOR): Likewise.
32217         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
32218         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32219         Build a new vector rather than modifying a CONST_VECTOR in-place.
32220         (handle_special_swappables): Update call accordingly.
32221         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
32222         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32223         Build a new vector rather than modifying a CONST_VECTOR in-place.
32224         (handle_special_swappables): Update call accordingly.
32226 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32228         * simplify-rtx.c (simplify_const_binary_operation): Use
32229         CONST_VECTOR_ELT instead of XVECEXP.
32231 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32233         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
32234         the selector elements to be different from the data elements
32235         if the selector is a VECTOR_CST.
32236         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
32237         ssizetype for the selector.
32239 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32241         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
32242         before testing each element individually.
32243         * tree-vect-generic.c (lower_vec_perm): Likewise.
32245 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32247         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
32248         * selftest-run-tests.c (selftest::run_tests): Call it.
32249         * vector-builder.h (vector_builder::operator ==): New function.
32250         (vector_builder::operator !=): Likewise.
32251         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
32252         (vec_perm_indices::all_from_input_p): New function.
32253         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
32254         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
32255         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
32256         instead of reading the VECTOR_CST directly.  Detect whether both
32257         vector inputs are the same before constructing the vec_perm_indices,
32258         and update the number of inputs argument accordingly.  Use the
32259         utility functions added above.  Only construct sel2 if we need to.
32261 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32263         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
32264         the broadcast of the low byte.
32265         (expand_mult_highpart): Use an explicit encoding for the permutes.
32266         * optabs-query.c (can_mult_highpart_p): Likewise.
32267         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
32268         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32269         (vectorizable_bswap): Likewise.
32270         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
32271         explicit encoding for the power-of-2 permutes.
32272         (vect_permute_store_chain): Likewise.
32273         (vect_grouped_load_supported): Likewise.
32274         (vect_permute_load_chain): Likewise.
32276 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32278         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
32279         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
32280         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
32281         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
32282         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
32283         (vect_gen_perm_mask_any): Likewise.
32285 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32287         * int-vector-builder.h: New file.
32288         * vec-perm-indices.h: Include int-vector-builder.h.
32289         (vec_perm_indices): Redefine as an int_vector_builder.
32290         (auto_vec_perm_indices): Delete.
32291         (vec_perm_builder): Redefine as a stand-alone class.
32292         (vec_perm_indices::vec_perm_indices): New function.
32293         (vec_perm_indices::clamp): Likewise.
32294         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
32295         (vec_perm_indices::new_vector): New function.
32296         (vec_perm_indices::new_expanded_vector): Update for new
32297         vec_perm_indices class.
32298         (vec_perm_indices::rotate_inputs): New function.
32299         (vec_perm_indices::all_in_range_p): Operate directly on the
32300         encoded form, without computing elided elements.
32301         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
32302         encoding.  Update for new vec_perm_indices class.
32303         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
32304         the given vec_perm_builder.
32305         (expand_vec_perm_var): Update vec_perm_builder constructor.
32306         (expand_mult_highpart): Use vec_perm_builder instead of
32307         auto_vec_perm_indices.
32308         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
32309         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
32310         or double series encoding as appropriate.
32311         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
32312         vec_perm_indices instead of auto_vec_perm_indices.
32313         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32314         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32315         (vect_permute_store_chain): Likewise.
32316         (vect_grouped_load_supported): Likewise.
32317         (vect_permute_load_chain): Likewise.
32318         (vect_shift_permute_load_chain): Likewise.
32319         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32320         (vect_transform_slp_perm_load): Likewise.
32321         (vect_schedule_slp_instance): Likewise.
32322         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32323         (vectorizable_mask_load_store): Likewise.
32324         (vectorizable_bswap): Likewise.
32325         (vectorizable_store): Likewise.
32326         (vectorizable_load): Likewise.
32327         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
32328         vec_perm_indices instead of auto_vec_perm_indices.  Use
32329         tree_to_vec_perm_builder to read the vector from a tree.
32330         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
32331         vec_perm_builder instead of a vec_perm_indices.
32332         (have_whole_vector_shift): Use vec_perm_builder and
32333         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
32334         truncation to calc_vec_perm_mask_for_shift.
32335         (vect_create_epilog_for_reduction): Likewise.
32336         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
32337         from auto_vec_perm_indices to vec_perm_indices.
32338         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32339         instead of changing individual elements.
32340         (aarch64_vectorize_vec_perm_const): Use new_vector to install
32341         the vector in d.perm.
32342         * config/arm/arm.c (expand_vec_perm_d::perm): Change
32343         from auto_vec_perm_indices to vec_perm_indices.
32344         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32345         instead of changing individual elements.
32346         (arm_vectorize_vec_perm_const): Use new_vector to install
32347         the vector in d.perm.
32348         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
32349         Update vec_perm_builder constructor.
32350         (rs6000_expand_interleave): Likewise.
32351         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
32352         (rs6000_expand_interleave): Likewise.
32354 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32356         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
32357         to qimode could truncate the indices.
32358         * optabs.c (expand_vec_perm_var): Likewise.
32360 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32362         * Makefile.in (OBJS): Add vec-perm-indices.o.
32363         * vec-perm-indices.h: New file.
32364         * vec-perm-indices.c: Likewise.
32365         * target.h (vec_perm_indices): Replace with a forward class
32366         declaration.
32367         (auto_vec_perm_indices): Move to vec-perm-indices.h.
32368         * optabs.h: Include vec-perm-indices.h.
32369         (expand_vec_perm): Delete.
32370         (selector_fits_mode_p, expand_vec_perm_var): Declare.
32371         (expand_vec_perm_const): Declare.
32372         * target.def (vec_perm_const_ok): Replace with...
32373         (vec_perm_const): ...this new hook.
32374         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
32375         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
32376         * doc/tm.texi: Regenerate.
32377         * optabs.def (vec_perm_const): Delete.
32378         * doc/md.texi (vec_perm_const): Likewise.
32379         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
32380         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
32381         expand_vec_perm for constant permutation vectors.  Assert that
32382         the mode of variable permutation vectors is the integer equivalent
32383         of the mode that is being permuted.
32384         * optabs-query.h (selector_fits_mode_p): Declare.
32385         * optabs-query.c: Include vec-perm-indices.h.
32386         (selector_fits_mode_p): New function.
32387         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
32388         is defined, instead of checking whether the vec_perm_const_optab
32389         exists.  Use targetm.vectorize.vec_perm_const instead of
32390         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
32391         fit in the vector mode before using a variable permute.
32392         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
32393         vec_perm_indices instead of an rtx.
32394         (expand_vec_perm): Replace with...
32395         (expand_vec_perm_const): ...this new function.  Take the selector
32396         as a vec_perm_indices rather than an rtx.  Also take the mode of
32397         the selector.  Update call to shift_amt_for_vec_perm_mask.
32398         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
32399         Use vec_perm_indices::new_expanded_vector to expand the original
32400         selector into bytes.  Check whether the indices fit in the vector
32401         mode before using a variable permute.
32402         (expand_vec_perm_var): Make global.
32403         (expand_mult_highpart): Use expand_vec_perm_const.
32404         * fold-const.c: Includes vec-perm-indices.h.
32405         * tree-ssa-forwprop.c: Likewise.
32406         * tree-vect-data-refs.c: Likewise.
32407         * tree-vect-generic.c: Likewise.
32408         * tree-vect-loop.c: Likewise.
32409         * tree-vect-slp.c: Likewise.
32410         * tree-vect-stmts.c: Likewise.
32411         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
32412         Delete.
32413         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
32414         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
32415         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
32416         (aarch64_vectorize_vec_perm_const): ...this new function.
32417         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32418         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32419         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
32420         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
32421         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32422         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32423         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
32424         into...
32425         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
32426         check for NEON modes.
32427         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
32428         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
32429         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
32430         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
32431         into...
32432         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
32433         the old VEC_PERM_CONST conditions.
32434         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
32435         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
32436         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
32437         (ia64_vectorize_vec_perm_const_ok): Merge into...
32438         (ia64_vectorize_vec_perm_const): ...this new function.
32439         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
32440         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
32441         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
32442         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
32443         * config/mips/mips.c (mips_expand_vec_perm_const)
32444         (mips_vectorize_vec_perm_const_ok): Merge into...
32445         (mips_vectorize_vec_perm_const): ...this new function.
32446         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
32447         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
32448         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
32449         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
32450         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
32451         (rs6000_expand_vec_perm_const): Delete.
32452         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
32453         Delete.
32454         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32455         (altivec_expand_vec_perm_const_le): Take each operand individually.
32456         Operate on constant selectors rather than rtxes.
32457         (altivec_expand_vec_perm_const): Likewise.  Update call to
32458         altivec_expand_vec_perm_const_le.
32459         (rs6000_expand_vec_perm_const): Delete.
32460         (rs6000_vectorize_vec_perm_const_ok): Delete.
32461         (rs6000_vectorize_vec_perm_const): New function.
32462         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32463         an element count and rtx array.
32464         (rs6000_expand_extract_even): Update call accordingly.
32465         (rs6000_expand_interleave): Likewise.
32466         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
32467         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
32468         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
32469         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
32470         (rs6000_expand_vec_perm_const): Delete.
32471         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32472         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32473         (altivec_expand_vec_perm_const_le): Take each operand individually.
32474         Operate on constant selectors rather than rtxes.
32475         (altivec_expand_vec_perm_const): Likewise.  Update call to
32476         altivec_expand_vec_perm_const_le.
32477         (rs6000_expand_vec_perm_const): Delete.
32478         (rs6000_vectorize_vec_perm_const_ok): Delete.
32479         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
32480         reference to the SPE evmerge intructions.
32481         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32482         an element count and rtx array.
32483         (rs6000_expand_extract_even): Update call accordingly.
32484         (rs6000_expand_interleave): Likewise.
32485         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
32486         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
32487         new function.
32488         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32490 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32492         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
32493         vector mode and that that mode matches the mode of the data
32494         being permuted.
32495         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
32496         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
32497         directly using expand_vec_perm_1 when forcing selectors into
32498         registers.
32499         (expand_vec_perm_var): New function, split out from expand_vec_perm.
32501 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32503         * optabs-query.h (can_vec_perm_p): Delete.
32504         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
32505         * optabs-query.c (can_vec_perm_p): Split into...
32506         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
32507         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
32508         particular selector is valid.
32509         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32510         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32511         (vect_grouped_load_supported): Likewise.
32512         (vect_shift_permute_load_chain): Likewise.
32513         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32514         (vect_transform_slp_perm_load): Likewise.
32515         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32516         (vectorizable_bswap): Likewise.
32517         (vect_gen_perm_mask_checked): Likewise.
32518         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
32519         implementations of variable permutation vectors into account
32520         when deciding which selector to use.
32521         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
32522         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
32523         with a false third argument.
32524         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
32525         to test whether the constant selector is valid and can_vec_perm_var_p
32526         to test whether a variable selector is valid.
32528 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32530         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
32531         * optabs-query.c (can_vec_perm_p): Likewise.
32532         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
32533         instead of vec_perm_indices.
32534         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
32535         (vect_gen_perm_mask_checked): Likewise,
32536         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
32537         (vect_gen_perm_mask_checked): Likewise,
32539 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32541         * optabs-query.h (qimode_for_vec_perm): Declare.
32542         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
32543         (qimode_for_vec_perm): ...this new function.
32544         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
32546 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32548         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
32549         does not have a conditional at the top.
32551 2018-01-02  Richard Biener  <rguenther@suse.de>
32553         * ipa-inline.c (big_speedup_p): Fix expression.
32555 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32557         PR target/81616
32558         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
32559         for generic 4->6.
32561 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32563         PR target/81616
32564         Generic tuning.
32565         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
32566         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
32567         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
32568         cond_taken_branch_cost 3->4.
32570 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
32572         PR tree-optimization/83581
32573         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
32574         TODO_cleanup_cfg if any changes have been made.
32576         PR middle-end/83608
32577         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
32578         convert_modes if target mode has the right side, but different mode
32579         class.
32581         PR middle-end/83609
32582         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
32583         last argument when extracting from CONCAT.  If either from_real or
32584         from_imag is NULL, use expansion through memory.  If result is not
32585         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
32586         the parts directly to inner mode, if even that fails, use expansion
32587         through memory.
32589         PR middle-end/83623
32590         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
32591         check for bswap in mode rather than HImode and use that in expand_unop
32592         too.
32594 Copyright (C) 2018 Free Software Foundation, Inc.
32596 Copying and distribution of this file, with or without modification,
32597 are permitted in any medium without royalty provided the copyright
32598 notice and this notice are preserved.