Allow target to override gnu-user.h crti and crtn
[official-gcc.git] / gcc / ChangeLog
blob62dd38eef7b70ba70732a38dc7ba0abcd12007a7
1 2018-11-18  Alan Modra  <amodra@gmail.com>
3         * config/gnu-user.h (GNU_USER_TARGET_CRTI): Define.
4         (GNU_USER_TARGET_STARTFILE_SPEC): Use it here.
5         (GNU_USER_TARGET_CRTN): Define.
6         (GNU_USER_TARGET_ENDFILE_SPEC): Use it here.
7         (GNU_USER_TARGET_LINK_GCC_C_SEQUENCE_SPEC): Define.
9 2018-11-18  Alan Modra  <amodra@gmail.com>
11         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Delete
12         !HAVE_LD_PIE variant.
13         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
15 2018-11-17  Nick Clifton  <nickc@redhat.com>
16             Sandra Loosemore  <sandra@codesourcery.com>
18         PR driver/31357
19         * doc/invoke.texi (Overall Options): Document that --help and
20         --help=<value> options cannot be combined.
22 2018-11-17  Jim Wilson  <jimw@sifive.com>
24         * config/riscv/riscv.c (epilogue_cfa_sp_offset): New.
25         (riscv_restore_reg): If restoring HARD_FRAME_POINTER_REGNUM, and
26         epilogue_cfa_sp_offset set, then add REG_CFA_DEF_CFA regnote.
27         (riscv_expand_epilogue): Initialize epilogue_cfa_sp_offset.  Set it
28         to step2 if frame_pointer_needed and step1 is 0.
30 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
32         PR c++/4225
33         * doc/cpp.texi (System Headers): Add note about implicit
34         extern "C" block on targets that define SYSTEM_IMPLICIT_EXTERN_C.
36 2018-11-17  Sandra Loosemore  <sandra@codesourcery.com>
38         PR c++/4025
39         * doc/invoke.texi (C++ Dialect Options): Clarify usage of
40         -fno-implicit-templates.
42 2018-11-17  Jakub Jelinek  <jakub@redhat.com>
44         PR tree-optimization/87546
45         * tree-vect-patterns.c (vect_look_through_possible_promotion): Add
46         min_precision variable, initially set it to orig_precision, only does
47         something if op_type's precision is <= min_precision and update
48         min_precision whenever calling set_op.
50 2018-11-16  Jan Hubicka  <hubicka@ucw.cz>
52         PR ipa/87957
53         * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL.
55 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
57         * doc/invoke.texi (Option Summary): Fix whitespace and line
58         breaks in @gccoptlist environments.
59         (Warning Options): Likewise.
60         (Optimize Options): Likewise.
61         (PowerPC SPE Options): Likewise.
62         (RS/6000 and PowerPC Options): Likewise.
64 2018-11-16  Jeff Law  <law@redhat.com>
66         * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus
67         gcc_assert.
69 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
71         PR middle-end/23197
72         * doc/invoke.texi (Optimize Options): Update options enabled by
73         fprofile-generate, -fprofile-use, and -fauto-profile.
75 2018-11-16  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
77         PR target/87927
78         * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
79         Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
80         * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
81         the byte_op hook.
82         * target.h: Add psi, pdi, pti to struct asm_int_op definition.
83         * targhooks.c (default_print_patchable_function_entry): Assert
84         asm_int_op does not return a NULL string.
85         * varasm.c (integer_asm_op): Return the op for a partial int type
86         when the requested size does not correspond to an integer type.
87         * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
88         * doc/tm.texi: Regenerate.
90 2018-11-12  Jason Merrill  <jason@redhat.com>
92         * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
94 2018-11-16  Michael Meissner  <meissner@linux.ibm.com>
96         * config/rs6000/constraints.md (wF constraint): Remove power9
97         fusion documentation.  Just document wF's use for power8 fusion.
98         * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
99         (fusion_gpr_addis): Delete power9 fusion support.  Change power8
100         fusion support to require the upper 12 bits to be all 0's or all
101         1's.
102         (fusion_gpr_mem_load): Add comment.
103         (fusion_addis_mem_combo_load): Remove power9 fusion support.
104         (fusion_addis_mem_combo_store): Delete.
105         (fusion_offsettable_mem_operand): Delete.
106         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
107         power8 fusion here.
108         (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
109         (POWERPC_MASKS): Delete power9 fusion.
110         * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
111         (fusion_p9_p): Delete.
112         (expand_fusion_p9_load): Delete.
113         (expand_fusion_p9_store): Delete.
114         (emit_fusion_p9_load): Delete.
115         (emit_fusion_p9_store): Delete.
116         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
117         fusion support.
118         (rs6000_option_override_internal): Set power8 fusion based on
119         whether we are tuning for power8.  Delete power9 fusion support.
120         (rs6000_opt_masks): Delete -mpower9-fusion switch.
121         (emit_fusion_load): Rename emit_fusion_load_store to
122         emit_fusion_load, and drop fusion store support.  Update callers.
123         (emit_fusion_load_store): Likewise.
124         (emit_fusion_gpr_load): Likewise.
125         (fusion_p9_p): Delete.
126         (expand_fusion_p9_load): Delete.
127         (expand_fusion_p9_store): Delete.
128         (emit_fusion_p9_load): Delete.
129         (emit_fusion_p9_store): Delete.
130         * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete.
131         (GPR_FUSION): Delete.
132         (FPR_FUSION): Delete.
133         (power9 fusion peephole2s): Delete.
134         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
135         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
136         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
137         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
138         (fusion_p9_<mode>_constant): Delete.
139         * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
140         power9 fusion switch.
141         * doc/md.texi (PowerPC constraints): Update wF constraint
142         documentation for power8 fusion only.
144 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
146         PR rtl-optimization/87475
147         * cfgrtl.c (patch_jump_insn): Allow redirection failure for
148         CROSSING_JUMP_P insns.
149         (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
151 2018-11-16  Uros Bizjak  <ubizjak@gmail.com>
153         PR target/88051
154         * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
155         * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
156         (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
158 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
160         PR middle-end/88032
161         * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
162         is VOIDmode, in which case use int_mode.  Similarly for op1_mode.
164 2018-11-16  Eric Botcazou  <ebotcazou@adacore.com>
166         * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
167         location_t parameter.
168         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
169         (switch_decision_tree::do_jump_if_equal): Likewise.
170         * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
171         of switch statement to emit_case_nodes.
172         (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
173         set it on the newly built GIMPLE comparison statement.
174         (switch_decision_tree::do_jump_if_equal): Likewise.
175         (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
176         in calls to do_jump_if_equal as well as recursive calls.
178 2018-11-16  Andi Kleen  <ak@linux.intel.com>
180         * config/i386/i386.md: Allow memory operands to ptwrite.
182 2018-11-16  Michael Matz  <matz@suse.de>
184         * system.h (PRsa): New macro.
185         (SIZE_AMOUNT): Cast number to uint64_t.
186         * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
187         (pool_usage::dump_footer): Likewise and also use PRIu64.
188         * bitmap.h (bitmap_usage::dump): Likewise.
189         * ggc-common.c (ggc_usage::dump): Likewise.
190         * ggc-page.c (ggc_print_statistics): Likewise.
191         * input.c (dump_line_table_statistics): Likewise.
192         * mem-stats.h (mem_usage::dump): Likewise.
193         (mem_usage::dump_footer): Likewise.
194         * rtl.c (dump_rtx_statistics): Likewise.
195         * tree-cfg.c (dump_cfg_stats): Likewise.
196         * tree-dfa.c (dump_dfa_stats): Likewise.
197         * tree-phinodes.c (phinodes_print_statistics): Likewise.
198         * tree-ssanames (ssanames_print_statistics): Likewise.
199         * vec.c (vec_usage::dump): Likewise.
200         (vec_usage::dump_footer): Likewise.
202 2018-11-16  Richard Biener  <rguenther@suse.de>
204         PR tree-optimization/88011
205         * tree-vrp.c (extract_range_from_binary_expr): Fix error in
206         replacing set_value_range_to_undefined and
207         set_value_range_to_varying with method calls.
209 2018-11-16  Ilya Leoshkevich  <iii@linux.ibm.com>
211         * config/s390/s390.md
212         (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
213         assembler.
214         (*r<noxa>sbg_<mode>_sll): Likewise.
215         (*r<noxa>sbg_<mode>_srl): Likewise.
216         (*r<noxa>sbg_sidi_srl): New pattern.
218 2018-11-16  Jerome Lambourg  <lambourg@adacore.com>
220         * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
221         Default to TARGET_VXWORKS_RTP.
222         (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
223         of TARGET_VXWORKS_RTP.
224         * config/vxworksae.h: Also define TARGET_VXWORKS_HAVE_CTORS_DTORS.
225         * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
226         of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
228 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
230         PR c++/25759
231         * doc/extend.texi (Common Type Attributes): Make it explicit
232         that attribute "packed" can apply to C++ classes.
234 2018-11-15  Martin Sebor  <msebor@redhat.com>
236         PR c++/87541
237         PR c++/87542
238         * tree.c (type_argument_type): New function.
239         * tree.h (type_argument_type): Declare it.
240         * gcc/doc/extend.texi (alloc_align): Update and clarify.
241         (alloc_size, nonnull, sentinel): Same.
243 2018-11-15  Andrew Stubbs  <ams@codesourcery.com>
244             Kwok Cheung Yeung  <kcy@codesourcery.com>
246         * tree-vect-stmts.c (vectorizable_store): Don't ICE when
247         int_mode_for_size fails.
248         (vectorizable_load): Likewise.
250 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
252         * doc/ux.texi (Group logically-related diagnostics): Move
253         discussion of auto_diagnostic_group into this new subsection.
254         Give an example of where this grouping is used.
256 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
258         PR other/19165
259         * Makefile.in (OBJS): Move json.o to...
260         (OBJS-libcommon): ...here and add diagnostic-format-json.o.
261         * common.opt (fdiagnostics-format=): New option.
262         (diagnostics_output_format): New enum.
263         * diagnostic-format-json.cc: New file.
264         * diagnostic.c (default_diagnostic_final_cb): New function, taken
265         from start of diagnostic_finish.
266         (diagnostic_initialize): Initialize final_cb to
267         default_diagnostic_final_cb.
268         (diagnostic_finish): Move "being treated as errors" messages to
269         default_diagnostic_final_cb.  Call any final_cb.
270         (default_diagnostic_finalizer): Add diagnostic_t param.
271         (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
272         diagnostic_finalizer callback.
273         * diagnostic.h (enum diagnostics_output_format): New enum.
274         (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
275         (struct diagnostic_context): Add "final_cb".
276         (default_diagnostic_finalizer): Add diagnostic_t param.
277         (diagnostic_output_format_init): New decl.
278         * doc/invoke.texi (-fdiagnostics-format): New option.
279         * dwarf2out.c (gen_producer_string): Ignore
280         OPT_fdiagnostics_format_.
281         * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
282         * lto-wrapper.c (append_diag_options): Ignore it.
283         * opts.c (common_handle_option): Handle it.
285 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
287         PR tree-optimization/88015
288         * graphite-isl-ast-to-gimple.c
289         (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
290         for dump_enabled_p.
291         * graphite-sese-to-poly.c (build_poly_scop): Likewise.
293 2018-11-15  Richard Biener  <rguenther@suse.de>
295         PR middle-end/88029
296         * gimple.c (gimple_call_flags): Union flags from decl, type
297         and call fntype.
298         * trans-mem.c (is_tm_pure_call): Simplify.
300 2018-11-15  Richard Biener  <rguenther@suse.de>
302         PR tree-optimization/88031
303         * tree-vect-loop.c (vectorizable_reduction): Move check
304         for multiple types earlier so we get the expected dump.
305         Simplify calls to vectorizable_condition.
306         * tree-vect-stmts.h (vectorizable_condition): Update prototype.
307         * tree-vect-stmts.c (vectorizable_condition): Instead of
308         reduc_def and reduc_index take just a flag.  Simplify
309         code-generation now that we can rely on the defs being set up.
310         (vectorizable_comparison): Remove unused argument.
312 2018-11-15  Richard Biener  <rguenther@suse.de>
314         PR tree-optimization/88030
315         * tree-complex.c (need_eh_cleanup): New global.
316         (update_complex_assignment): Mark blocks that need EH update.
317         (expand_complex_comparison): Likewise.
318         (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
319         perform EH cleanup and schedule CFG cleanup if that did anything.
321 2018-11-15  Jakub Jelinek  <jakub@redhat.com>
323         PR rtl-optimization/88018
324         * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
325         on the presence of fallthru edge, rather than if it is a USE or not.
327 2018-11-15  Richard Biener  <rguenther@suse.de>
329         PR middle-end/87917
330         * tree-data-ref.c (analyze_miv_subscript): Guard calls to
331         analyze_subscript_affine_affine properly.
333 2018-11-15  Xianmiao Qu  <xianmiao_qu@c-sky.com>
335         * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
336         (GLIBC_DYNAMIC_LINKER): Define.
337         (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
339 2018-11-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
341         PR tree-optimization/84648
342         * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
343         (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
344         by calling adjust_cond_for_loop_until_wrap.
346 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
348         PR other/56334
349         * doc/extend.texi (Common Function Attributes): Clarify linker
350         restrictions on "aligned" attribute.
351         (Common Variable Attributes): Likewise.  Mention that linker
352         restrictions don't apply to stack-allocated variables.
354 2018-11-15  Joern Rennecke  <joern.rennecke@riscy-ip.com>
356         * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
357         as an lvalue.
359 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
361         PR lto/55102
362         PR lto/56700
363         * doc/invoke.texi (Optimize Options): Remove bad example about
364         interaction between -flto and -O.  Replace it with a note that
365         you need to compile with -O and not just link.  Copy-edit -flto
366         discussion to reduce verbiage and improve flow.
368 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
370         PR middle-end/59658
371         * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
372         also suppress many optimizations.  Alphabetize option lists for
373         -O1, -O2, and -Os.  Add list of options disabled with -Og, and
374         correct documentation for those options to say that.
375         * opts.c (default_options_table): Sort table by level and option
376         name, to make it easier to correlate to the manual.
378 2018-11-14  Uros Bizjak  <ubizjak@gmail.com>
380         * config/i386/i386.c (ix86_print_operand_address_as): Simplify
381         printing of the names of segment registers.
383 2018-11-14  Nathan Sidwell  <nathan@acm.org>
385         PR debug/88006
386         PR debug/87462
387         * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
388         type list.
390 2018-11-14  David Malcolm  <dmalcolm@redhat.com>
392         * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
394 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
396         P1236R1 - Signed integers are two's complement
397         * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
399         PR bootstrap/86739
400         * hash-map.h (hash_map::iterator::reference_pair): New class.
401         (hash_map::iterator::operator*): Return it rather than std::pair.
403 2018-11-14  Jeff Law  <law@redhat.com>
405         * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
406         iff the operand is a constant.
408 2018-11-14  Aldy Hernandez  <aldyh@redhat.com>
410         * gimple-ssa-evrp-analyze.c
411         (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
412         ignore_equivs_equal_p to equal_p.
413         * ipa-cp.c (meet_with_1): Use equal_p instead of
414         ignore_equivs_equal_p.
415         * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
416         * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
417         (value_range::operator==): Remove.
418         (value_range::operator!=): Remove.
419         (vrp_prop::visit_stmt): Use equal_p.
420         * tree-vrp.h (value_range): Remove operator==, operator!=,
421         ignore_equivs_equal_p.
422         * vr-values.c (update_value_range): Use equal_p.
424 2018-11-14  Michael Matz  <matz@suse.de>
426         PR middle-end/86575
427         * gimplify.c (collect_fallthrough_labels): Add new argument,
428         return location via that, don't modify statements.
429         (warn_implicit_fallthrough_r): Adjust call, don't use
430         statement location directly.
432 2018-11-14  Richard Biener  <rguenther@suse.de>
434         PR middle-end/87985
435         * tree-data-ref.c (split_constant_offset): Add wrapper
436         allocating a cache hash-map.
437         (split_constant_offset_1): Cache results of expanding
438         expressions from SSA def stmts.
440 2018-11-14  Richard Biener  <rguenther@suse.de>
442         PR middle-end/88021
443         * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
444         to lambda_int.
445         (lambda_vector_mult_const): Likewise.
446         (lambda_matrix_right_hermite): Use lambda_int temporaries.
448 2018-11-14  Wilco Dijkstra  <wdijkstr@arm.com>
449             Jackson Woodruff  <jackson.woodruff@arm.com>
451         PR 71026/tree-optimization
452         * match.pd: Simplify floating point comparisons.
454 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
456         PR rtl-optimization/87817
457         * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
458         last argument has low 8 bits clear, fold to 0.
460 2018-11-14  Iain Sandoe  <iain@sandoe.co.uk>
462         * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
463         MAX_OFILE_ALIGNMENT as needed.
465 2018-11-14  Richard Biener  <rguenther@suse.de>
467         PR tree-optimization/88019
468         * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
469         COND_EXPR arguments to non-trapping overflow.
471 2018-11-14  Claudiu Zissulescu  <claziss@synopsys.com>
473         * config/arc/arc.h (reg_class): Reorder registers classes, remove
474         unused register classes.
475         (REG_CLASS_NAMES): Likewise.
476         (REG_CLASS_CONTENTS): Likewise.
477         (FIXED_REGISTERS): Make lp_count fixed.
478         (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
479         (PROGRAM_COUNTER_REGNO): Remove.
480         * config/arc/arc.c (arc_conditional_register_usage): Remove unused
481         register classes, use constants for register numbers, remove
482         reg_class_contents references.
483         (arc_process_double_reg_moves): Add asserts.
484         (arc_secondary_reload): Remove LPCOUNT_REG reference, use
485         lra_in_progress predicate.
486         (arc_init_reg_tables): Remove unused register classes.
487         (arc_register_move_cost): Likewise.
488         (arc_preferred_reload_class): Likewise.
489         (hwloop_optimize): Update rtx patterns involving lp_count register.
490         (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
491         * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
492         CHEAP_CORE_REGS.  Former one will be used for LRA.
493         ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS.  Former
494         one will be used for LRA.
495         ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS.  Former
496         one will be used for LRA.
497         ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
498         Former one will be used for LRA.
499         ("f"): Delete constraint.
500         ("k"): Likewise.
501         ("e"): Likewise.
502         ("l"): Change it from register constraint to constraint.
503         * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
504         (movhi_insn): Likewise.
505         (movsi_insn): Update pattern.
506         (arc_lp): Likewise.
507         (dbnz): Likewise.
508         (stack_tie): Remove 'b' constraint letter.
509         (R4_REG): Define.
510         (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
511         (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
512         (R57_REG, R59_REG, PCL_REG): Likewise.
513         (ILINK1_REGNUM): Renamed to ILINK1_REG.
514         (ILINK2_REGNUM): Renamed to ILINK2_REG.
515         (Rgp): Remove.
516         (SP_REGS): Likewise.
517         (Rcw): Remove unused reg classes.
518         * config/arc/predicates.md (dest_reg_operand): Just default on
519         register_operand predicate.
520         (mpy_dest_reg_operand): Likewise.
521         (move_dest_operand): Use macros instead of constants.
522         * config/arc/arc.opt (mlra): Switch to lra as default.
524 2018-11-14  Richard Biener  <rguenther@suse.de>
526         PR tree-optimization/87974
527         * tree-vect-loop.c (vectorizable_reduction): When computing
528         the vectorized reduction PHI vector type ignore constant
529         and external defs.
531 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
533         PR tree-optimization/87977
534         * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
535         stmt, build a new one and replace the old one with it.  Formatting fix.
536         Call release_ssa_name (x) if !has_other_use and !delete_div.
537         (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
538         verify lhs of stmt is still def.
540 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
542         PR rtl-optimization/87507
543         * lower-subreg.c (operand_for_swap_move_operator): New function.
544         (simple_move): Strip simple operators.
545         (find_pseudo_copy): Likewise.
546         (resolve_operand_for_swap_move_operator): New function.
547         (resolve_simple_move): Strip simple operators and swap operands.
549 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
551         PR tree-optimization/87898
552         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
553         (ipa_simd_modify_function_body): Remove debug stmts where the first
554         argument was changed into a non-decl.
556 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
558         PR rtl-optimization/87899
559         * lra-lives.c (start_living): Update white space in comment.
560         (enum point_type): New.
561         (sparseset_contains_pseudos_p): New function.
562         (update_pseudo_point): Likewise.
563         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
564         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
565         handling.  Move early exit after adding conflicts.
566         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
567         if regno is already live.  Remove all handling of program points.
568         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
569         after adding conflicts.  Remove all handling of program points and
570         ignore_reg_for_conflicts.
571         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
572         and do not guard call to mark_pseudo_live.
573         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
574         and do not guard call to mark_pseudo_dead.
575         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
576         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
577         Use new function update_pseudo_point.  Handle register copies by
578         removing the source register from the live set.  Handle INOUT operands.
579         Update to the next program point using the unused_set, dead_set and
580         start_dying sets.
581         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
583 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
585         * builtins.c: Replace "source_location" with "location_t".
586         * diagnostic-show-locus.c: Likewise.
587         * diagnostic.c: Likewise.
588         * dumpfile.c: Likewise.
589         * gcc-rich-location.h: Likewise.
590         * genmatch.c: Likewise.
591         * gimple.h: Likewise.
592         * gimplify.c: Likewise.
593         * input.c: Likewise.
594         * input.h: Likewise.  Eliminate the typedef.
595         * omp-expand.c: Likewise.
596         * selftest.h: Likewise.
597         * substring-locations.h (get_source_location_for_substring):
598         Rename to...
599         (get_location_within_string): ...this.
600         * tree-cfg.c: Replace "source_location" with "location_t".
601         * tree-cfgcleanup.c: Likewise.
602         * tree-diagnostic.c: Likewise.
603         * tree-into-ssa.c: Likewise.
604         * tree-outof-ssa.c: Likewise.
605         * tree-parloops.c: Likewise.
606         * tree-phinodes.c: Likewise.
607         * tree-phinodes.h: Likewise.
608         * tree-ssa-loop-ivopts.c: Likewise.
609         * tree-ssa-loop-manip.c: Likewise.
610         * tree-ssa-phiopt.c: Likewise.
611         * tree-ssa-phiprop.c: Likewise.
612         * tree-ssa-threadupdate.c: Likewise.
613         * tree-ssa.c: Likewise.
614         * tree-ssa.h: Likewise.
615         * tree-vect-loop-manip.c: Likewise.
617 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
619         * doc/extend.texi: Fix typo in the weakref description.
621 2018-11-13  Richard Biener  <rguenther@suse.de>
623         PR tree-optimization/86991
624         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
625         group building until we have successfully detected the SLP
626         reduction.
627         (vect_is_simple_reduction): Remove fixup code here.
629 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
631         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
632         (dump_gimple_stmt): Use it.
633         (dump_gimple_stmt_loc): Likewise.
634         (dump_gimple_expr): Likewise.
635         (dump_gimple_expr_loc): Likewise.
636         (dump_generic_expr): Likewise.
637         (dump_generic_expr_loc): Likewise.
638         (dump_printf): Likewise.
639         (dump_printf_loc): Likewise.
640         (dump_dec): Likewise.
641         (dump_dec): Likewise.
642         (dump_hex): Likewise.
643         (dump_symtab_node): Likewise.
644         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
645         Guard dump call with dump_enabled_p.
646         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
647         * graphite-optimize-isl.c (optimize_isl): Likewise.
648         * graphite.c (graphite_transform_loops): Likewise.
649         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
650         * tree-parloops.c (parallelize_loops): Likewise.
651         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
652         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
653         (vect_prune_runtime_alias_test_list): Likewise.
654         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
655         (vect_estimate_min_profitable_iters): Likewise.
656         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
657         (vect_build_slp_tree_2): Likewise.
658         (vect_supported_load_permutation_p): Likewise.
659         (vect_slp_analyze_operations): Likewise.
660         (vect_slp_analyze_bb_1): Likewise.
661         (vect_slp_bb): Likewise.
662         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
663         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
664         (pass_slp_vectorize::execute): Likewise.
665         (increase_alignment): Likewise.
667 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
669         PR ipa/87955
670         * ipa-inline.c (report_inline_failed_reason): Guard calls to
671         cl_target_option_print_diff and cl_optimization_print_diff with
672         if (dump_file).
674 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
676         * doc/invoke.texi (-fsave-optimization-record): Note that the
677         output is compressed.
678         * optinfo-emit-json.cc: Include <zlib.h>.
679         (optrecord_json_writer::write): Compress the output.
681 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
683         * tree-vrp.c (value_range_base::dump): Dump type.
684         Do not use INF nomenclature for 1-bit types.
685         (dump_value_range): Group all variants to common dumping code.
686         (debug): New overloaded functions for value_ranges.
687         (value_range_base::dump): Remove no argument version.
688         (value_range::dump): Same.
690 2018-11-13  Richard Biener  <rguenther@suse.de>
692         PR tree-optimization/87931
693         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
694         nested cycles we support to latch computations vectorizable_reduction
695         handles.
697 2018-11-13  Martin Liska  <mliska@suse.cz>
699         PR tree-optimization/87885
700         * cfghooks.c (account_profile_record): Rename
701         to ...
702         (profile_record_check_consistency): ... this.
703         Calculate missing num_mismatched_freq_in.
704         (profile_record_account_profile): New function
705         that calculates time and size of a function.
706         * cfghooks.h (struct profile_record): Remove
707         all tuples.
708         (struct cfg_hooks): Remove after_pass flag.
709         (account_profile_record): Rename to ...
710         (profile_record_check_consistency): ... this.
711         (profile_record_account_profile): New.
712         * cfgrtl.c (rtl_account_profile_record): Remove
713         after_pass flag.
714         * passes.c (check_profile_consistency): Do only
715         checking.
716         (account_profile): Calculate size and time of
717         function only.
718         (pass_manager::dump_profile_report): Reformat
719         output.
720         (execute_one_ipa_transform_pass): Call
721         consistency check before clean upand call account_profile
722         after a clean up is done.
723         (execute_one_pass): Call check_profile_consistency and
724         account_profile instead of using after_pass flag..
725         * tree-cfg.c (gimple_account_profile_record): Likewise.
727 2018-11-13  Martin Liska  <mliska@suse.cz>
729         PR sanitizer/87930
730         * config/i386/i386.c (ix86_option_override_internal): Error
731         about usage -mabi=ms and -fsanitize={,kernel-}address.
733 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
735         * config/aarch64/aarch64.c
736         (aarch64_vectorize_preferred_vector_alignment): Change return type to
737         poly_uint64.
738         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
739         alignment being a poly int.
740         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
741         return type to poly_uint64.
742         * target.def (default_preferred_vector_alignment): Likewise.
743         * targhooks.c (default_preferred_vector_alignment): Likewise.
744         * targhooks.h (default_preferred_vector_alignment): Likewise.
745         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
746         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
747         poly int.
748         (vect_update_misalignment_for_peel): Likewise.
749         (vect_enhance_data_refs_alignment): Likewise.
750         (vect_find_same_alignment_drs): Likewise.
751         (vect_duplicate_ssa_name_ptr_info): Likewise.
752         (vect_setup_realignment): Likewise.
753         (vect_can_force_dr_alignment_p): Change alignment parameter type to
754         poly_uint64.
755         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
756         mask with a compile time variable vector alignment.
757         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
758         int.
759         (vect_do_peeling): Exit early if vector alignment is not constant.
760         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
761         a poly int.
762         (vectorizable_store): Likewise.
763         (vectorizable_load): Likweise.
764         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
765         poly_uint64.
766         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
767         poly int.
768         (vect_can_force_dr_alignment_p): Change alignment parameter type to
769         poly_uint64.
771 2018-11-13  Richard Biener  <rguenther@suse.de>
773         PR tree-optimization/87962
774         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
775         detect outer reduction for disqualifying in-loop uses.
777 2018-11-13  Richard Biener  <rguenther@suse.de>
779         PR tree-optimization/87967
780         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
781         for constants for the scalar loop.
783 2018-11-13  Alan Modra  <amodra@gmail.com>
785         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
786         integer constants when -mcmodel=medium.
788 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
790         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
791         * config/arc/elf.h (PROFILE_HOOK): Define.
792         * config/arc/linux.h (PROFILE_HOOK): Likewise.
794 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
796         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
797         jump table data in the text section.
799 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
801         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
802         to fit the eh_return pattern.
803         * config/arc/arc.md (eh_return): Define.
804         (VUNSPEC_ARC_EH_RETURN): Likewise.
805         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
806         new implementation.
807         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
810 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
812         * common/config/arc/arc-common.c (arc_option_optimization_table):
813         Millicode optimization is default on for size optimizations.
814         * config/arc/arc-protos.h (arc_check_multi): New function.
815         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
816         (ENTER_LEAVE_START_REG): Define.
817         (ENTER_LEAVE_END_REG): Likewise.
818         (arc_override_options): Disable millicode when long calls option
819         is on.
820         (arc_frame_info): Change it from int to bool.
821         (arc_compute_frame_size): Clean up.
822         (arc_save_restore): Remove.
823         (frame_save_reg): New function.
824         (frame_restore_reg): Likewise.
825         (arc_enter_leave_p): Likewise.
826         (arc_save_callee_saves): Likewise.
827         (arc_restore_callee_saves): Likewise.
828         (arc_save_callee_enter): Likewise.
829         (arc_restore_callee_leave): Likewise.
830         (arc_save_callee_milli): Likewise.
831         (arc_restore_callee_milli): Likewise.
832         (arc_expand_prologue): Reimplement to emit enter/leave
833         instructions.
834         (arc_expand_epilogue): Likewise.
835         (arc_check_multi): New function.
836         * config/arc/arc.md (push_multi_fp): New pattern.
837         (push_multi_fp_blink): Likewise.
838         (pop_multi_fp): Likewise.
839         (pop_multi_fp_blink): Likewise.
840         (pop_multi_fp_ret): Likewise.
841         (pop_multi_fp_blink_ret): Likewise.
842         * config/arc/arc.opt (mmillicode): Update option.
843         (mcode-density-frame): New option.
844         * config/arc/predicates.md (push_multi_operand): New predicate.
845         (pop_multi_operand): Likewise.
846         * doc/invoke.texi (ARC): Update ARC options information.
849 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
851         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
852         * config/arc/arc.c (operands_ok_ldd_std): New function.
853         (mem_ok_for_ldd_std): Likewise.
854         (gen_operands_ldd_std): Likewise.
855         * config/arc/arc.md: Add peephole2 rules for std/ldd.
857 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
859         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
860         test on stack_usage_file.
861         (lang_dependent_init): Do not open the .su file if generating LTO.
863 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
865         PR rtl-optimization/87918
866         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
867         simplify_gen_relational rather than simplify_gen_binary.
869 2018-11-13  Richard Biener  <rguenther@suse.de>
871         * tree-ssanames.h (set_range_info): Use value_range_base.
872         (get_range_info): Likewise.
873         * tree-ssanames.c (set_range_info): Likewise.
874         (get_range_info): Likewise.
875         * tree-vrp.c (value_range_base::union_helper): Split
876         out common parts of value_range[_base]::union_.
877         (value_range_base::union_): Update.
878         (value_range::union_): Likewise.
879         (determine_value_range_1): Use value_range_base.
880         (determine_value_range): Likewise.
881         * tree-vrp.h (value_range_base::union_helper): Move ...
882         (value_range::union_helper): ... from here.
884 2018-11-13  Alan Modra  <amodra@gmail.com>
886         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
887         offset for PRE_DEC.
888         (rs6000_secondary_reload_gpr): Don't call find_replacement.
890 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
892         PR middle-end/59634
893         * doc/invoke.texi (Optimize Options): Clarify that the
894         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
895         apply to data cache size.
897 2018-11-13  Alan Modra  <amodra@gmail.com>
899         * config/rs6000/predicates.md (logical_const_operand),
900         (logical_operand): Correct comment.
901         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
903 2018-11-13  Alan Modra  <amodra@gmail.com>
905         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
907 2018-11-13  Alan Modra  <amodra@gmail.com>
909         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
910         description.
911         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
912         covered by alternative.
913         (movcc_internal1): Ignore h for register preference.
914         (mov<mode>_hardfloat64): Likewise.
915         (mov<mode>_softfloat): Ignore c, l, h for register preference.
917 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
919         PR preprocessor/47823
920         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
921         documentation to...
922         (Pragmas): ...here.
923         * doc/extend.texi (Pragmas): Note additional pragmas documented
924         in the CPP manual.
926 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
928         PR middle-end/86677
929         PR middle-end/87528
930         * tree-scalar-evolution.c (expression_expensive_p): Make
931         BUILTIN POPCOUNT as expensive when backend does not define it.
933 2018-11-12  Fredrik Noring  <noring@nocrew.org>
935         * config.gcc: Update with-llsc defaults for MIPS r5900.
937 2018-11-12  Martin Liska  <mliska@suse.cz>
939         PR gcov-profile/87442
940         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
941         options.
942         * doc/invoke.texi: Document them.
943         * tree-profile.c (parse_profile_filter): New.
944         (parse_profile_file_filtering): Likewise.
945         (release_profile_file_filtering): Likewise.
946         (include_source_file_for_profile): Likewise.
947         (tree_profiling): Filter source files based on the
948         newly added options.
950 2018-11-12  Martin Liska  <mliska@suse.cz>
952         PR target/87903
953         * doc/extend.texi: Add missing values for __builtin_cpu_is and
954         __builtin_cpu_supports for x86 target.
956 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
958         PR middle-end/21110
959         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
960         have to be Pmode.
961         (Arithmetic): Likewise for "lo_sum".
963 2018-11-12  Renlin Li  <renlin.li@arm.com>
965         PR target/87815
966         * dse.c (get_stored_val): Add check for compile-time constantness
967         of gap.
969 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
971         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
972         (ARMv8_5a): New fgroup.
973         (armv8.5-a): New arch.
974         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
975         sb and predres.
976         * config/arm/arm-tables.opt: Regenerate.
977         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
978         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
979         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
980         Add matching rules for -march=armv8.5-a and extensions.
981         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
982         Add sb and predres to all armv8-a except armv8.5-a.
984 2018-11-12  Richard Biener  <rguenther@suse.de>
986         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
987         overload for single value.
988         (value_range[_base]::set_nonnull): New.
989         (value_range[_base]::set_null): Likewise.
990         (value_range): Document bitmap copying behavior, mark
991         copy constructor and assignment operator deleted.
992         (value_range::move): New.
993         (value_range::set_and_canonicalize): Default bitmap to zero.
994         (set_value_range_to_nonnull): Remove.
995         (set_value_range_to_null): Likewise.
996         (set_value_range): Likewise.
997         (set_value_range_to_value): Likewise.
998         (extract_range_from_unary_expr): Work on value_range_base.
999         (extract_range_from_binary_expr_1): Likewise.  Rename to...
1000         (extract_range_from_binary_expr): ... this.
1001         * tree-vrp.c (value_range::update): Clear equiv bitmap
1002         if required.
1003         (value_range::move): New, move equiv bitmap.
1004         (value_range_base::set_undefined): Avoid assignment.
1005         (value_range::set_undefined): Likewise.
1006         (value_range_base::set_varying): Likewise.
1007         (value_range::set_varying): Likewise.
1008         (set_value_range): Remove.
1009         (value_range_base::set): New overload for value.
1010         (value_range::set): Likewise.
1011         (set_value_range_to_nonnull): Remove.
1012         (value_range_base::set_nonnull): New.
1013         (value_range::set_nonnull): Likewise.
1014         (set_value_range_to_null): Remove.
1015         (value_range_base::set_null): New.
1016         (value_range::set_null): Likewise.
1017         (range_is_null): Work on value_range_base.
1018         (range_is_nonnull): Likewise.
1019         (ranges_from_anti_range): Likewise.
1020         (extract_range_into_wide_ints): Likewise.
1021         (extract_range_from_multiplicative_op): Likewise.
1022         (extract_range_from_binary_expr): Likewise.  Update for API changes.
1023         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
1024         handling.
1025         (value_range::intersect_helper): Avoid copy and assignment.
1026         (value_range::union_helper): Likewise.
1027         (determine_value_range_1): Adjust.
1028         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
1029         Avoid assignment by using move.
1030         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
1031         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
1032         Likewise.
1033         * tree-ssanames.c (get_range_info): Likewise.
1034         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
1035         * vr-values.c (vr_values::get_value_range): Adjust.
1036         (vr_values::update_value_range): Likewise.
1037         (symbolic_range_based_on_p): Work on value_range_base.
1038         (vr_values::extract_range_from_binary_expr): Use value_range_base.
1039         (vr_values::extract_range_from_unary_expr): Likewise.
1040         (vr_values::extract_range_from_cond_expr): Avoid assignment.
1041         (vr_values::extract_range_from_comparison): Adjust.
1042         (vr_values::check_for_binary_op_overflow): Use value_range_base.
1043         (vr_values::extract_range_basic): Adjust.
1044         (vr_values::adjust_range_with_scev): Likewise.
1045         (vr_values::vrp_visit_assignment_or_call): Likewise.
1046         (vr_values::get_vr_for_comparison): Change API to avoid
1047         assignment and copy construction.
1048         (vr_values::compare_name_with_value): Adjust accordingly.
1049         (vr_values::compare_names): Likewise.
1050         (vr_values::extract_range_from_phi_node): Avoid assignment and
1051         bogus in-place modify of equiv bitmap.
1052         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
1053         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
1054         for extract_range_from_unary_expr API change.
1055         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
1057 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
1059         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
1061 2018-11-12  Richard Biener  <rguenther@suse.de>
1063         * tree-vrp.h (value_range_base::symbolic_p,
1064         value_range_base::constant_p, value_range_base::zero_p,
1065         value_range_base::singleton_p): Move from value_range.
1066         (value_range::dump): Add.
1067         * gimple-ssa-evrp-analyze.c
1068         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
1069         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
1070         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
1071         Use set_varying.
1072         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
1073         (value_range::constant_p): Likewise.
1074         (value_range::singleton_p): Likewise.
1075         (value_range_base::dump): Add.
1076         (set_value_range_to_undefined): Remove.
1077         (set_value_range_to_varying): Likewise.
1078         (range_int_cst_p): Take value_range_base argument.
1079         (range_int_cst_singleton_p): Likewise.
1080         (value_range_constant_singleton): Likewise.
1081         (vrp_set_zero_nonzero_bits): Likewise.
1082         (extract_range_from_multiplicative_op): Use set_varying.
1083         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
1084         (extract_range_from_unary_expr): Likewise.
1085         (dump_value_range_base): Change to overload of dump_value_range.
1086         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
1087         (vrp_prop::visit_stmt): Likewise.
1088         (value_range::intersect_helper): Likewise.
1089         (value_range::union_helper): Likewise.
1090         (determine_value_range_1): Likewise.
1092 2018-11-12  Richard Biener  <rguenther@suse.de>
1094         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
1095         (set_value_range_to_null): Likewise.
1096         * vr-values.c (vr_values::extract_range_from_comparison):
1097         Clear equiv for constant singleton ranges.
1099 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
1101         * config/i386/sse.md: Combine VFIXUPIMM* patterns
1102         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1103         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1104         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
1105         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1106         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1107         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
1109 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1111         PR c/69502
1112         * doc/extend.texi (Common Type Attributes): For the align type
1113         attribute, copy language about decreasing alignment from the
1114         corresponding variable attribute.
1116 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1118         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
1119         -frounding-math.
1121 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1123         PR c++/43105
1124         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
1125         -frtti and -fno-rtti code.
1127 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1129         PR c/26366
1130         * doc/extend.texi (Other Builtins): Document probability associated
1131         with __builtin_expect.
1133 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
1135         PR target/87928
1136         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
1137         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
1138         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
1139         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
1141 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1143         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
1145 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1147         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
1149 2018-11-11  Richard Biener  <rguenther@suse.de>
1151         * tree-vrp.h (class value_range_base): New base class for
1152         value_range containing all but the m_equiv member.
1153         (dump_value_range_base): Add.
1154         (range_includes_zero_p): Work on value_range_base.
1155         * tree-vrp.c (value_range_base::set): Split out base handling
1156         from...
1157         (value_range::set): this.
1158         (value_range::set_equiv): New.
1159         (value_range_base::value_range_base): New constructors.
1160         (value_range_base::check): Split out base handling from...
1161         (value_range::check): this.
1162         (value_range::equal_p): Refactor in terms of
1163         ignore_equivs_equal_p which is now member of the base.
1164         (value_range_base::set_undefined): New.
1165         (value_range_base::set_varying): Likewise.
1166         (value_range_base::dump):Split out base handling from...
1167         (value_range::dump): this.
1168         (value_range_base::set_and_canonicalize): Split out base handling
1169         from...
1170         (value_range::set_and_canonicalize): this.
1171         (value_range_base::union_): New.
1172         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
1173         for m_vr.
1174         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
1175         instead of value_range everywhere.
1176         (ipcp_vr_lattice::print): Use dump_value_range_base.
1177         (ipcp_vr_lattice::meet_with): Adjust.
1178         (ipcp_vr_lattice::meet_with_1): Likewise.
1179         (ipa_vr_operation_and_type_effects): Likewise.
1180         (propagate_vr_across_jump_function): Likewise.
1181         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
1182         (ipa_get_value_range): Likewise.
1183         (ipa_set_jfunc_vr): Likewise.
1184         (ipa_compute_jump_functions_for_edge): Likewise.
1186 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
1188         PR middle-end/65703
1189         * doc/invoke.texi (Optimize Options): Add @opindex entries
1190         for the positive forms of -fno-xxx and -mno-xxx options
1191         that were lacking them.
1193 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
1195         * combine.c (make_more_copies): Only make an intermediate copy if the
1196         dest of a move is a pseudo.
1198 2018-11-09  Maya Rashish  <coypu@sdf.org>
1200         PR target/87221
1201         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
1202         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
1204 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
1206         PR driver/41179
1207         PR middle-end/65703
1208         * doc/invoke.texi (Optimize Options): Clarify default behavior
1209         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
1211 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1213         PR tree-optimization/87940
1214         * expr.c (string_constant): Don't strip NOPS in subexpressions.
1215         Fold PLUS_EXPR correctly.
1217 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
1219         PR target/87762
1220         * config/s390/s390.c (s390_safe_relative_long_p): New function.
1221         (annotate_constant_pool_refs): Skip insns which support
1222         relative addressing.
1223         (annotate_constant_pool_refs_1): New helper function.
1224         (find_constant_pool_ref): Skip insns which support relative
1225         addression.
1226         (find_constant_pool_ref_1): New helper function.
1227         (replace_constant_pool_ref): Skip insns which support
1228         relative addressing.
1229         (replace_constant_pool_ref_1): New helper function.
1230         (s390_mainpool_start): Adapt to the new signature.
1231         (s390_mainpool_finish): Likewise.
1232         (s390_chunkify_start): Likewise.
1233         (s390_chunkify_finish): Likewise.
1234         (pass_s390_early_mach::execute): Likewise.
1235         (s390_prologue_plus_offset): Likewise.
1236         (s390_emit_prologue): Likewise.
1237         (s390_emit_epilogue): Likewise.
1239 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
1241         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
1242         but unsupported lastprivate with conditional modifier.
1244 2018-11-09  Jeff Law  <law@redhat.com>
1246         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
1247         unused argument better.  Add gcc_unreachable to silence warning.
1249 2018-11-09  Martin Sebor  <msebor@redhat.com>
1251         PR middle-end/81824
1252         * attribs.c (has_attribute): New helper function.
1253         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
1254         * attribs.h (decls_mismatched_attributes): Declare.
1255         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
1256         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
1257         * common.opt (-Wattribute-alias): Take an argument.
1258         (-Wno-attribute-alias): New option.
1259         * doc/extend.texi (Common Function Attributes): Document copy.
1260         (Common Variable Attributes): Same.
1261         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
1262         (-Wattribute-alias): Document new option argument.
1264 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
1266         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
1267         matching criteria.  Remove unused array initializer.
1269 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
1270             Jinsong Ji  <jji@us.ibm.com>
1272         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
1273         constraints by introducing a new temporary.
1274         (_mm_cvtss_si64): Likewise.
1276 2018-11-09  Martin Liska  <mliska@suse.cz>
1278         * common.opt: Add -fipa-stack-alignment flag.
1279         * doc/invoke.texi: Document it.
1280         * final.c (rest_of_clean_state): Guard stack
1281         shrinking with flag.
1283 2018-11-09  Martin Liska  <mliska@suse.cz>
1285         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
1286         to ...
1287         (ipa_discover_variable_flags): ... this.
1288         * common.opt: Come up with new flag -fipa-reference-addressable.
1289         * doc/invoke.texi: Document it.
1290         * ipa-reference.c (propagate): Call the renamed fn.
1291         * ipa-visibility.c (whole_program_function_and_variable_visibility):
1292         Likewise.
1293         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
1294         ...
1295         (ipa_discover_variable_flags): ... this.  Discover
1296         non-addressable variables only with the newly added flag.
1297         * opts.c: Enable the newly added flag with -O1 and higher
1298         optimization level.
1300 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
1302         * json.cc (selftest::test_writing_literals): Fix comment.
1304 2018-11-09  Martin Liska  <mliska@suse.cz>
1306         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
1307         string to a stack buffer.
1308         (aarch64_parse_cpu): Likewise.
1309         (aarch64_parse_tune): Likewise.
1311 2018-11-09  Richard Biener  <rguenther@suse.de>
1313         PR tree-optimization/87953
1314         * tree-vect-loop.c (vectorizable_reduction): For analysis
1315         always pass ops[0] to vectorizable_condition.
1317 2018-11-09  Stafford Horne  <shorne@gmail.com>
1318             Richard Henderson  <rth@twiddle.net>
1319             Joel Sherrill  <joel@rtems.org>
1321         * common/config/or1k/or1k-common.c: New file.
1322         * config/or1k/*: New.
1323         * config.gcc (or1k*-*-*): New.
1324         * configure.ac (or1k*-*-*): New test for openrisc tls.
1325         * configure: Regenerated.
1326         * doc/install.texi: Document OpenRISC triplets.
1327         * doc/invoke.texi: Document OpenRISC arguments.
1328         * doc/md.texi: Document OpenRISC.
1330 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
1332         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
1333         (arm7tdmi-s): Delete CPU.
1334         (arm710t): Add aliases for arm720t and arm740t.
1335         (arm720t, arm740t): Delete CPUs.
1336         (arm920t): Add aliases for arm920, arm922t and arm940t.
1337         (arm920, arm922t, arm940t): Delete CPUs.
1338         (arm10tdmi): Add alias for arm1020t.
1339         (arm1020t): Delete CPU.
1340         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
1341         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
1342         (arm10e): Add aliases for arm1020e and arm1022e.
1343         (arm1020e, arm1022e): Delete CPU.
1344         * config/arm/arm.md (generic_sched): Remove entries that are now
1345         handled by aliases.
1346         (generic_vfp): Likewise.
1347         * config/arm/arm1020e.md: Simplify tuning selection based on alias
1348         changes.
1349         * config/arm/arm-tune.md: Regenerated.
1350         * config/arm/arm-tables.opt: Regenerated.
1352 2018-11-09  Richard Biener  <rguenther@suse.de>
1354         PR tree-optimization/87621
1355         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
1356         op with only phi inputs.
1357         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
1358         (ch_base::copy_headers): Run CSE on copied loop headers.
1359         (pass_ch_vect::process_loop_p): Simplify.
1361 2018-11-09  Alexandre Oliva <oliva@adacore.com>
1363         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
1364         for non-w64 x86_64 biarch.
1366 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
1368         PR rtl-optimization/86438
1369         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
1370         of in_b for the compare if in_b is SET_DEST.
1372         PR target/87793
1373         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
1374         non-toplevel UNSPEC.
1376 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1378         * tree-vrp.c (value_range::check): Do not access internals
1379         directly.
1380         (value_range::singleton_p): Same.
1381         (value_range::type): Same.
1382         (vrp_finalize): Use value_range API.
1384 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1386         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
1388 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1390         * vr-values.c (vr_values::get_value_range): Use value_range API
1391         instead of piecing together ranges.
1392         (vr_values::update_value_range): Same.
1394 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1396         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
1397         of performing ad-hoc calculations.
1398         * tree-ssanames.c (set_range_info): New overloaded function
1399         accepting value_range &.
1400         (get_range_info): Same.
1401         * tree-ssanames.h (set_range_info_raw): Remove.
1402         (set_range_info): New prototype.
1403         (get_range_info): Same.
1404         * tree-vrp.h (value_range::null_p): Rename to zero_p.
1405         * tree-vrp.c (value_range::null_p): Same.
1407 2018-11-09  Jan Hubicka  <jh@suse.cz>
1409         * tree.c (fld_type_variant_equal_p): Test user align flag.
1410         (flt_type_variant): Copy user align flag.
1411         (fld_incomplete_type_of): Clear it.
1413 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1415         * config/arm/neon.md (div<mode>3): New pattern.
1417 2018-11-08  Andi Kleen  <ak@linux.intel.com>
1419         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
1420         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
1421         (ix86_handle_option): Handle OPT_mptwrite.
1422         * config/i386/cpuid.h (bit_PTWRITE): Add.
1423         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
1424         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
1425         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
1426         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
1427         (ix86_option_override_internal): Handle PTA_PTWRITE.
1428         (ix86_valid_target_attribute_inner_p): Define ptwrite.
1429         (def_builtin2): Force UINT64 to be 64bit only.
1430         * config/i386/i386.h (TARGET_PTWRITE): Add.
1431         (TARGET_PTWRITE_P): Add.
1432         (PTA_PTWRITE): Add.
1433         * config/i386/i386.md: Define ptwrite.
1434         * config/i386/i386.opt: Add -mptwrite.
1435         * config/i386/immintrin.h (_ptwrite64): Add.
1436         (_ptwrite32): Add
1437         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
1438         * doc/invoke.texi: Document -mptwrite.
1440 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
1442         PR rtl-optimization/87600
1443         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
1444         * lra-constraints.c (process_alt_operands): Skip illegal hard
1445         register usage.  Prefer reloading non hard register operands.
1447 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
1449         PR other/36572
1450         * doc/invoke.texi (Optimize Options): Clarify default behavior
1451         for -fno-sched-interblock and -fno-sched-spec.
1453 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
1455         * collect2.c (linker_select):  Add USE_LLD_LD.
1456         (ld_suffixes): Add ld.lld.
1457         (main): Handle -fuse-ld=lld.
1458         * common.opt (-fuse-ld=lld): New option.
1459         * doc/invoke.texi (-fuse-ld=lld): Document.
1460         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
1462 2018-11-08  Paul Koning  <ni1d@arrl.net>
1464         * config/pdp11/constraints.md: Add "Z" series constraints for use
1465         with pre-dec and post-inc addressing.
1466         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
1467         (pdp11_expand_operands): Add int argument (word count).
1468         (pdp11_sp_frame_offset): Delete.
1469         (pdp11_cmp_length): New function.
1470         (pushpop_regeq): New function.
1471         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
1472         Add hook.
1473         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
1474         frame layout.
1475         (pdp11_initial_elimination_offset): Ditto.
1476         (pdp11_expand_operands): Add word count argument.  Bugfixes.
1477         (output_move_multiple): Change how pointer adjustment is done.
1478         (pdp11_gen_int_label): Correct format.
1479         (output_ascii): Ditto.
1480         (pdp11_asm_output_var): Add code for DEC assembler case.
1481         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
1482         value.
1483         (legitimate_const_double_p): Ditto.
1484         (pdp11_register_move_cost): Adjust for new register classes.
1485         (pdp11_regno_reg_class): Ditto.
1486         (expand_block_move): Delete.
1487         (pushpop_regeq): New function.
1488         (pdp11_legitimate_address_p): Bugfix in check for constant
1489         offset.
1490         (pdp11_sp_frame_offset): Delete.
1491         (pdp11_reg_save_size): New helper function for new frame layout.
1492         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
1493         (pdp11_expand_shift): Bugfix in check for constant shift count.
1494         (pdp11_shift_length): Ditto.
1495         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
1496         (pdp11_cmp_length): New function.
1497         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
1498         some compile options.
1499         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
1500         (CALL_USED_REGISTERS): Ditto.
1501         (ELIMINABLE_REGS): Ditto.
1502         (REGISTER_NAMES): Ditto.
1503         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
1504         constraints.
1505         (REG_CLASS_NAMES): Ditto.
1506         (REG_CLASS_CONTENTS): Ditto.  Also remove
1507         HARD_FRAME_POINTER_REGNUM.
1508         (CPU_REG_CLASS): New macro.
1509         (CLASS_MAX_NREGS): Adjust for new register classes.
1510         (FUNCTION_PROFILER): Make no-op.
1511         (may_call_alloca): Remove unused declaration.
1512         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
1513         (ASM_OUTPUT_SKIP): Fix format.
1514         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
1515         (HARD_FRAME_POINTER_REGNUM): Remove.
1516         (return): Delete.
1517         (*rts): Rename.  Remove epilogue related checks.
1518         (cmpsi, cmpdi): New insn.
1519         (cbranch<mode>4): Change to apply to SI and DI modes as well.
1520         (mov<mode>): Change constraints to enforce that push/pop
1521         destination cannot use the same register as source.
1522         (*mov<mode><cc_cc>): Ditto.
1523         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
1524         at assembly output rather than as RTL expander.
1525         (zero_extendqihi2): Bugfix in check for same registers.
1526         (adddi3_nocc): Bugfix in check for constant operand.
1527         (addsi3_nocc): Ditto.
1528         (subdi3_nocc): Ditto.
1529         (subsi3_nocc): Ditto.
1530         (negdi2_nocc): Copy input to pdp11_expand_operands.
1531         (negsi2_nocc): Ditto.
1532         (bswap2_nocc): Ditto.
1533         * config/pdp11/pdp11.opt (mlra): Fix documentation.
1534         * config/pdp11/t-pdp11: Use -Os.
1536 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
1538         * config/arm/parsecpu.awk (/alias/): New parsing rule.
1539         (/begin cpu/): Check that the cpu name hasn't been previously defined.
1540         (gen_comm_data): Print out CPU alias tables.
1541         (check_cpu): Match aliases when checking the CPU name.
1542         * config/arm/arm-protos.h (cpu_alias): New structure.
1543         (cpu_option): Add entry for aliases.
1544         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
1545         strongarm1100 and strongarm1110.
1546         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
1547         (config/arm/arm-generic.md): Remove redundant references to
1548         strongarm110, strongarm1100 and strongarm1110.
1549         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1550         Scan aliases for additional hints.
1551         (arm_parse_cpu_option_name): Also match a cpu name against the list
1552         of aliases.
1553         * config/arm/arm-tables.opt: Regenerated.
1554         * config/arm/arm-tune.md: Regenerated.
1556 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
1558         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
1559         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
1560         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1561         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
1562         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1563         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
1564         * gengtype.c (open_base_files): Add omp-general.h.
1565         * gimple.c (gimple_build_omp_critical):
1566         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
1567         gimple_omp_taskgroup_set_clauses.
1568         (gimple_build_omp_atomic_load): Add mo argument, call
1569         gimple_omp_atomic_set_memory_order.
1570         (gimple_build_omp_atomic_store): Likewise.
1571         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
1572         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
1573         instead of GSS_OMP.
1574         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
1575         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
1576         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
1577         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
1578         different value for GF_OMP_ATOMIC_NEED_VALUE.
1579         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
1580         comments.
1581         (struct gimple_statement_omp_single_layout): And remove here.
1582         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
1583         than gimple_statement_omp_single_layout.
1584         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
1585         GIMPLE_OMP_TEAMS.
1586         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
1587         (gimple_omp_subcode): Formatting fix.
1588         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
1589         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
1590         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
1591         gimple_omp_teams_host, gimple_omp_teams_set_host,
1592         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
1593         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
1594         gimple_omp_taskgroup_set_clauses): New inline functions.
1595         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
1596         (gimple_build_omp_atomic_store): Likewise.
1597         (gimple_omp_atomic_seq_cst_p): Remove.
1598         (gimple_omp_atomic_memory_order): New function.
1599         (gimple_omp_atomic_set_seq_cst): Remove.
1600         (gimple_omp_atomic_set_memory_order): New function.
1601         (gimple_build_omp_taskgroup): Add clauses argument.
1602         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
1603         (dump_gimple_omp_task): Print taskwait with depend clauses.
1604         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
1605         dump_omp_atomic_memory_order.
1606         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
1607         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
1608         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
1609         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
1610         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
1611         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
1612         ORT_UNTIED_TASKLOOP enumerators.
1613         (enum gimplify_defaultmap_kind): New.
1614         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
1615         target_map_pointers_as_0len_arrays members, add defaultmap.
1616         (new_omp_context): Initialize defaultmap member.
1617         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1618         (maybe_fold_stmt): Don't fold even in host teams regions.
1619         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
1620         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
1621         ctx->omp_firstprivatize_variable.
1622         (omp_add_variable): Don't add private/firstprivate for VLAs in
1623         ORT_TASKGROUP.
1624         (omp_default_clause): Print "taskloop" rather than "task" if
1625         ORT_*TASKLOOP.
1626         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1627         Handle new defaultmap clause kinds.
1628         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
1629         iterator to be lastprivate or private.  Fix up diagnostics if linear
1630         is used on collapse>1 simd iterator.
1631         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1632         (gimplify_omp_depend): New function.
1633         (gimplify_scan_omp_clauses): Add shared clause on parallel for
1634         combined parallel master taskloop{, simd} if taskloop has
1635         firstprivate, lastprivate or reduction clause.  Handle
1636         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
1637         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
1638         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
1639         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
1640         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
1641         lastprivate.
1642         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
1643         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.
1644         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
1645         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1646         (gimplify_omp_task): Handle taskwait with depend clauses.
1647         (gimplify_omp_for): Add shared clause on parallel for combined
1648         parallel master taskloop{, simd} if taskloop has firstprivate,
1649         lastprivate or reduction clause.  Use ORT_TASKLOOP or
1650         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
1651         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
1652         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
1653         __for_end and __for_range temporaries on OMP_PARALLEL for
1654         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
1655         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
1656         sandwiched in between two taskloops.
1657         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
1658         instead of ctx->omp_firstprivatize_variable.
1659         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
1660         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
1661         host teams, use gimplify_and_return_first etc. for body like
1662         for target or target data constructs, and at the end call
1663         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
1664         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
1665         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
1666         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
1667         gimple_omp_atomic_set_seq_cst calls.
1668         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
1669         case, handle taskgroup clauses.
1670         * lto-streamer-out.c (hash_tree): Handle
1671         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1672         * Makefile.in (GTFILES): Add omp-general.h.
1673         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
1674         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
1675         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
1676         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
1677         BUILT_IN_GOMP_LOOP_DOACROSS_START,
1678         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
1679         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1680         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
1681         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
1682         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
1683         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
1684         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
1685         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1686         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
1687         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
1688         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
1689         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
1690         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
1691         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
1692         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
1693         * omp-expand.c (workshare_safe_to_combine_p): Return false for
1694         non-worksharing loops.
1695         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
1696         (determine_parallel_type): Don't combine parallel with worksharing
1697         which has _reductemp_ clause.
1698         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
1699         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
1700         if there is nonmonotonic modifier or if there is no modifier and no
1701         ordered clause.  For dynamic and guided schedule without monotonic
1702         and nonmonotonic modifier, default to nonmonotonic.
1703         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
1704         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
1705         task reductions.
1706         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
1707         there are any reduction clauses.
1708         (expand_taskwait_call): New function.
1709         (expand_teams_call): New function.
1710         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
1711         expand_teams_call for it.  Formatting fix.  Handle taskwait with
1712         depend clauses.
1713         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
1714         of worksharing loops with task reductions.
1715         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
1716         expansion of worksharing loops with task reductions.
1717         (expand_omp_sections): Handle expansion of sections with task
1718         reductions.
1719         (expand_omp_synch): For host teams call expand_omp_taskreg.
1720         (omp_memory_order_to_memmodel): New function.
1721         (expand_omp_atomic_load, expand_omp_atomic_store,
1722         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
1723         instead of gimple_omp_atomic_seq_cst_p.
1724         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
1725         depend clauses as a standalone directive.
1726         * omp-general.c (enum omp_requires): New variable.
1727         (omp_extract_for_data): Initialize have_reductemp member.  Allow
1728         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
1729         GT_EXPR loops depending on incr sign.  Formatting fixes.
1730         * omp-general.h (struct omp_for_data): Add have_reductemp member.
1731         (enum omp_requires): New enum.
1732         (omp_requires_mask): Declare.
1733         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
1734         Fix comment typos.
1735         * omp-low.c (struct omp_context): Add task_reductions and
1736         task_reduction_map fields.
1737         (is_host_teams_ctx): New function.
1738         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
1739         (use_pointer_for_field): Use is_global_var instead of
1740         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
1741         in outer contexts.
1742         (build_outer_var_ref): Ignore taskgroup outer contexts.
1743         (delete_omp_context): Release task_reductions and task_reduction_map.
1744         (scan_sharing_clauses): Don't add any fields for reduction clause on
1745         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
1746         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
1747         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
1748         Handle OMP_CLAUSE_NONTEMPORAL.
1749         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
1750         needed.
1751         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
1752         clauses with task modifier.
1753         (scan_omp_task): Handle taskwait with depend clauses.
1754         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
1755         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
1756         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
1757         (scan_omp_for): Fix comment formatting.
1758         (scan_omp_teams): Handle host teams constructs.
1759         (check_omp_nesting_restrictions): Allow teams with no outer
1760         OpenMP context.  Adjust diagnostics for teams strictly nested into
1761         some explicit OpenMP construct other than target.  Allow OpenMP atomics
1762         inside of simd regions.
1763         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
1764         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
1765         taskreg_nesting_level while scanning host teams construct.
1766         (task_reduction_read): New function.
1767         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
1768         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
1769         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
1770         second argument create_tmp_var if it is NULL.  Don't ignore shared
1771         clauses in is_host_teams_ctx contexts.  Handle
1772         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
1773         clauses.
1774         (lower_reduction_clauses): Ignore reduction clauses with task
1775         modifier.  Remove second argument create_tmp_var if it is NULL.
1776         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
1777         (lower_send_clauses): Ignore reduction clauses with task modifier.
1778         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
1779         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
1780         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
1781         rely that it is the last stmt in body so far.  Ignore outer taskgroup
1782         contexts.
1783         (omp_task_reductions_find_first, omp_task_reduction_iterate,
1784         lower_omp_task_reductions): New functions.
1785         (lower_omp_sections): Handle reduction clauses with taskgroup
1786         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
1787         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
1788         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
1789         modifiers.
1790         (lower_omp_taskgroup): Handle taskgroup reductions.
1791         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
1792         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
1793         (lower_depend_clauses): If there are any
1794         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1795         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
1796         seen, assume lowering is done already and return early.  Set kind
1797         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
1798         (lower_omp_taskreg): Handle reduction clauses with task modifier on
1799         parallel construct.  Handle reduction clause on taskloop construct.
1800         Handle taskwait with depend clauses.
1801         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
1802         for host teams constructs.
1803         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
1804         nontemporal and _reductemp_ clause entries.
1805         (omp_clause_code_name): Likewise.
1806         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
1807         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
1808         * tree-core.h (enum omp_clause_code): Add
1809         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
1810         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
1811         (struct tree_base): Add omp_atomic_memory_order field into union.
1812         Remove OMP_ATOMIC_SEQ_CST comment.
1813         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1814         and OMP_CLAUSE_DEPEND_DEPOBJ.
1815         (struct tree_omp_clause): Add subcode.defaultmap_kind.
1816         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
1817         OpenMP constructs with body and clauses operands.
1818         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
1819         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
1820         (OMP_TASKGROUP_CLAUSES): Define.
1821         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
1822         OMP_CLAUSE__LOOPTEMP_.
1823         (OMP_ATOMIC_SEQ_CST): Remove.
1824         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
1825         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
1826         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
1827         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
1828         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
1829         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
1830         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
1831         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
1832         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
1833         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
1834         Define.
1835         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
1836         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1837         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
1838         (convert_local_omp_clauses): Likewise.  Remove useless test.
1839         * tree-parloops.c (create_call_for_reduction_1): Pass
1840         OMP_MEMORY_ORDER_RELAXED as new argument to
1841         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
1842         * tree-pretty-print.c (dump_omp_iterators): New function.
1843         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
1844         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
1845         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
1846         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
1847         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
1848         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
1849         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
1850         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
1851         (dump_omp_atomic_memory_order): New function.
1852         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
1853         taskwait with depend clauses.
1854         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
1855         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
1856         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
1857         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
1858         write_ts_omp_clause_tree_pointers): Likewise.
1860 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1862         PR ipa/86395
1863         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
1864         "-missed", "-note", and "-all" sub-options.
1865         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
1866         API.
1867         (can_early_inline_edge_p): Likewise.
1868         (want_early_inline_function_p): Likewise.
1869         (want_inline_self_recursive_call_p): Likewise.
1870         (recursive_inlining): Likewise.
1871         (inline_small_functions): Likewise.
1872         (flatten_function): Likewise.
1873         (ipa_inline): Likewise.
1874         (inline_always_inline_functions): Likewise.
1875         (early_inline_small_functions): Likewise.
1876         (early_inliner): Likewise.
1877         * tree-inline.c (expand_call_inline): Likewise.
1879 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1881         * pretty-print.c (pp_format): Handle %f.
1882         (selftest::test_pp_format): Add test of %f.
1883         * pretty-print.h (pp_double): New macro.
1885 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1887         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
1888         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
1889         (dump_pretty_printer::decode_format): Implement "%C" for
1890         cgraph_node *.
1891         (selftest::test_capture_of_dump_calls): Rename "where" to
1892         "stmt_loc".  Convert test_decl to a function decl and set its
1893         location.  Add a symbol_table_test RAII instance and a
1894         cgraph_node, using it to test "%C" and dump_symtab_node.
1896 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
1898         PR middle-end/87916
1899         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
1901 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1903         * cgraph.c: Include "selftest.h".
1904         (saved_symtab): New variable.
1905         (selftest::symbol_table_test::symbol_table_test): New ctor.
1906         (selftest::symbol_table_test::~symbol_table_test): New dtor.
1907         (selftest::test_symbol_table_test): New test.
1908         (selftest::cgraph_c_tests): New.
1909         * cgraph.h (saved_symtab): New decl.
1910         (selftest::symbol_table_test): New class.
1911         * selftest-run-tests.c (selftest::run_tests): Call
1912         selftest::cgraph_c_tests.
1913         * selftest.h (selftest::cgraph_c_tests): New decl.
1915 2018-11-08  Richard Biener  <rguenther@suse.de>
1917         * tree-data-ref.h (lambda_int): New typedef.
1918         (lambda_vector_gcd): Adjust.
1919         (lambda_vector_new): Likewise.
1920         (lambda_matrix_new): Likewise.
1921         * tree-data-ref.c  (print_lambda_vector): Adjust.
1923 2018-11-08  Richard Biener  <rguenther@suse.de>
1925         PR tree-optimization/87929
1926         * tree-complex.c (expand_complex_comparison): Clean EH.
1928 2018-11-08  Martin Liska  <mliska@suse.cz>
1930         * doc/extend.texi: Reword.
1931         * predict.c (expr_expected_value_1): Likewise.
1933 2018-11-08  Richard Biener  <rguenther@suse.de>
1935         PR tree-optimization/87913
1936         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
1937         of extreme values to ordered comparisons.
1939 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1941         PR middle-end/42726
1942         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
1944 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1946         * doc/invoke.texi: Remove leading dash from @opindex entries
1947         throughout the file.
1949 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1951         PR driver/80828
1952         * doc/invoke.texi (Option Summary): Add -e and --entry.
1953         (Link Options): Likewise.
1955 2018-11-07  Nathan Sidwell  <nathan@acm.org>
1957         PR 87926
1958         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
1959         --disable-checking bootstrap.
1961 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
1963         * configure: Regenerated.
1965 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1967         PR c/87691
1968         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
1969         to the mode of the widest field iff the widest field has mode class
1970         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
1971         reference.
1973 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
1975         * common.opt: Add -Wattribute-warning.
1976         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
1977         * expr.c (expand_expr_real_1): Add new attribute to warning_at
1978         call to allow user configure behavior of "warning" attribute.
1980 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
1982         * target.def: Put @: after every vs., e.g., and i.e. where it is
1983         followed by whitespace.
1984         * doc/extend.texi: Ditto.
1985         * doc/fragments.texi: Ditto.
1986         * doc/gimple.texi: Ditto.
1987         * doc/implement-c.texi: Ditto.
1988         * doc/install.texi: Ditto.
1989         * doc/invoke.texi: Ditto.
1990         * doc/md.texi: Ditto.
1991         * doc/plugins.texi: Ditto.
1992         * doc/rtl.texi: Ditto.
1993         * doc/sourcebuild.texi: Ditto.
1994         * doc/tm.texi.in: Ditto.
1995         * doc/ux.texi: Ditto.
1996         * doc/tm.texi: Regenerate.
1998 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2000         * config/arm/arm-cpus.in (ares): New entry.
2001         * config/arm/arm-tables.opt: Regenerate.
2002         * config/arm/arm-tune.md: Likewise.
2003         * doc/invoke.texi (ARM Options): Document ares.
2005 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2007         * config/aarch64/aarch64-cores.def (ares): Define.
2008         * config/aarch64/aarch64-tune.md: Regenerate.
2009         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
2011 2018-11-07  Jan Hubicka  <jh@suse.cz>
2013         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
2014         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
2015         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
2016         TYPE_DECL.
2018 2018-11-07  Richard Biener  <rguenther@suse.de>
2020         PR tree-optimization/87914
2021         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
2022         of nested cycles.
2023         (vectorizable_reduction): Handle shifts and rotates by dispatching
2024         to vectorizable_shift.
2025         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
2026         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
2027         def cases.
2028         (vectorizable_shift): Export and handle being called as
2029         vect_nested_cycle.
2030         (vect_analyze_stmt): Call vectorizable_shift after
2031         vectorizable_reduction.
2032         * tree-vectorizer.h (vectorizable_shift): Declare.
2034 2018-11-07  Jan Hubicka  <jh@suse.cz>
2036         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
2037         than const decls in TREE_VALUE of enum.
2038         (dump_type_inheritance_graph): Improve duplicate dumping.
2039         (free_enum_values): New.
2040         (build_type_inheritance_graph): Use it.
2041         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
2042         which are not main variants or not ODR types.
2043         (verify_type_variant): Expect variants to have no TYPE_VALUES.
2045 2018-11-07  Richard Biener  <rguenther@suse.de>
2047         * ipa-inline.c (want_inline_small_function_p): Compute
2048         big_speedup_p lazily and last.
2050 2018-11-07  Jan Hubicka  <jh@suse.cz>
2052         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
2053         building incomplete variant of complete type.
2054         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
2055         variant of complete type.
2057 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2059         * config/mips/mips.c: Fix typo in documentation of
2060         mips_loongson_ext2_prefetch_cookie.
2061         (mips_option_override): fix brain twister logical.
2062         * config/mips/mips.h: Fix typo in documentation of
2063         ISA_HAS_CTZ_CTO and define pattern.
2064         * config/mips/mips.md (prefetch): Hoist EXT2 above
2065         the 2EF/EXT block.
2066         (prefetch_indexed): Hoist EXT2 above the EXT block.
2068 2018-11-07  Jan Hubicka  <jh@suse.cz>
2070         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
2071         return and parameter types of function and method types.
2072         (free_lang_data_in_cgraph): Update.
2074 2018-11-07  Martin Liska  <mliska@suse.cz>
2076         PR rtl-optimization/87868
2077         * postreload-gcse.c (eliminate_partially_redundant_load): Set
2078         threshold to max_count if we would overflow.
2079         * profile-count.h: Make max_count a public constant.
2081 2018-11-07  Martin Liska  <mliska@suse.cz>
2083         * mem-stats.h: Fix GNU coding style.
2085 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2087         * config/mips/gs264e.md: New.
2088         * config/mips/mips-cpus.def: Define gs264e.
2089         * config/mips/mips-tables.opt: Regenerate.
2090         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
2091         gs264e.
2092         (mips_issue_rate): Add support for gs264e.
2093         (mips_multipass_dfa_lookahead): Likewise.
2094         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
2095         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
2096         (MIPS_ASE_MSA_SPEC): New.
2097         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
2098         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
2099         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
2100         * config/mips/mips.md: Include gs264e.md.
2101         (processor): Add gs264e.
2102         * config/mips/mips.opt (MSA): Use Mask instead of Var.
2103         * doc/invoke.texi: Add gs264e to supported architectures.
2105 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2107         * config/mips/gs464e.md: New.
2108         * config/mips/mips-cpus.def: Define gs464e.
2109         * config/mips/mips-tables.opt: Regenerate.
2110         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
2111         gs464e.
2112         (mips_issue_rate): Add support for gs464e.
2113         (mips_multipass_dfa_lookahead): Likewise.
2114         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
2115         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
2116         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
2117         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
2118         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
2119         * config/mips/mips.md: Include gs464e.md.
2120         (processor): Add gs464e.
2121         * doc/invoke.texi: Add gs464e to supported architectures.
2123 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2125         * config/mips/loongson3a.md: Rename to ...
2126         * config/mips/gs464.md: ... here.
2127         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
2128         as an alias of gs464 processor.
2129         * config/mips/mips-tables.opt: Regenerate.
2130         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
2131         instead of PROCESSOR_LOONGSON_3A.
2132         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
2133         TUNE_LOONGSON_3A.
2134         (mips_option_override): Enable MMI and EXT for gs464.
2135         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
2136         Rename TUNE_LOONGSON_3A to TUNE_GS464.
2137         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
2138         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
2139         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
2140         TARGET_LOONGSON_3A.
2141         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
2142         (processor): Add gs464;
2143         * doc/invoke.texi: Add gs464 to supported architectures.
2145 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2147         * config/mips/mips-protos.h
2148         (mips_loongson_ext2_prefetch_cookie): New prototype.
2149         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
2150         (mips_option_override): Enable TARGET_LOONGSON_EXT when
2151         TARGET_LOONGSON_EXT2 is true.
2152         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
2153         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
2154         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
2155         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
2156         TARGET_LOONGSON_EXT2.
2157         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
2158         (define_insn "ctz<mode>2"): New insn pattern.
2159         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
2160         (define_insn "prefetch_indexed_<mode>"): Include
2161         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
2162         * config/mips/mips.opt (-mloongson-ext2): Add option.
2163         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
2165 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2167         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
2168         __mips_loongson_ext.
2169         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
2170         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
2171         -mloongson-ext.
2172         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
2173         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
2174         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
2175         instead of TARGET_LOONGSON_3A.
2176         * config/mips/mips.opt (-mloongson-ext): Add option.
2177         * gcc/doc/invoke.texi (-mloongson-ext): Document.
2179 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2181         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
2182         * config/mips/loongson.md: Move to ...
2183         * config/mips/loongson-mmi.md: here; Adjustment.
2184         * config/mips/loongson.h: Move to ...
2185         State as deprecated. Include loongson-mmiintrin.h for back
2186         compatibility and warning.
2187         * config/mips/loongson-mmiintrin.h: ... here.
2188         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
2189         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
2190         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
2191         (mips_option_override): Make sure MMI use hard float;
2192         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
2193         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
2194         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
2195         TARGET_LOONGSON_VECTORS.
2196         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
2197         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
2198         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
2199         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
2200         -mloongson-mmi.
2201         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
2202         TARGET_LOONGSON_VECTORS.
2203         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
2204         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
2205         (Loongson MMI patterns): Include loongson-mmi.md instead of
2206         loongson.md.
2207         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
2208         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
2210 2018-11-07  Richard Biener  <rguenther@suse.de>
2212         PR lto/87906
2213         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
2214         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
2216 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
2218         PR rtl-optimization/87874
2219         * lra.c (lra_substitute_pseudo): Do not create a subreg for
2220         const wide ints.
2222 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
2224         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
2225         if not in indexed or indirect form.
2226         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
2227         (bswap<mode>2_store): Ditto.
2229 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
2231         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
2232         the UNSPEC.
2234 2018-11-06  Richard Biener  <rguenther@suse.de>
2236         PR tree-optimization/86850
2237         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
2238         instead of src.m_vec.
2240 2018-11-06  Jan Hubicka  <jh@suse.cz>
2242         * tree.c (fld_simplified_type_name): Break out form ...
2243         (free_lang_data_in_type): ... here.
2244         (fld_type_variant_equal_p): Use it.
2246 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2248         * config/default-d.c: Include memmodel.h.
2250         * config/sol2-d.c: New file.
2251         * config/t-sol2 (sol2-d.o): New rule.
2252         * config.gcc <*-*-solaris2*>: Set d_target_objs,
2253         target_has_targetdm.
2255 2018-11-06  Jan Hubicka  <jh@suse.cz>
2257         * tree.c (fld_type_variant): Also copy alignment; be sure that
2258         new variant is equal.
2260 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2262         PR target/87762
2263         * config/s390/s390.md: Add relative_long attribute.
2265 2018-11-06  Jan Hubicka  <jh@suse.cz>
2267         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
2268         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
2269         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
2270         stream TYPE_NEEDS_CONSTRUCTING.
2271         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
2272         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
2274 2018-11-06  Richard Biener  <rguenther@suse.de>
2276         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
2277         dump-scope ...
2278         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
2280 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2282         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
2283         REG_CLASS_CONTENTS[GEN_REGS].
2284         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
2286 2018-11-06  Jan Hubicka  <jh@suse.cz>
2288          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
2290 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2292         PR sanitizer/80953
2293         * config/sol2.h (ASAN_CC1_SPEC): Define.
2294         (LD_WHOLE_ARCHIVE_OPTION): Define.
2295         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
2296         (ASAN_REJECT_SPEC): Provide default.
2297         (LIBASAN_EARLY_SPEC): Define.
2298         (LIBTSAN_EARLY_SPEC): Define.
2299         (LIBLSAN_EARLY_SPEC): Define.
2300         * config/i386/sol2.h (CC1_SPEC): Redefine.
2301         (ASAN_REJECT_SPEC): Define.
2303         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
2304         (TARGET_ASAN_SHADOW_OFFSET): Define.
2305         (sparc_asan_shadow_offset): New function.
2306         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
2307         (ASAN_REJECT_SPEC): Define.
2309 2018-11-06  Jan Hubicka  <jh@suse.cz>
2311         * tree.c (fld_type_variant): Copy canonical type.
2312         (fld_incomplete_type_of): Check that canonical types looks sane;
2313         copy canonical type.
2314         (verify_type): Accept when incomplete type has complete canonical type.
2316 2018-11-06  Jan Hubicka  <jh@suse.cz>
2318         * tree.c (free_lang_data): Reset overwite_assembler_name,
2319         print_xnode, print_decl, print_type and print_identifier of
2320         langhooks.
2322 2018-11-06  Richard Biener  <rguenther@suse.de>
2324         PR tree-optimization/87889
2325         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2326         Do nothing if old and new arg are the same
2328 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
2330         PR target/87723
2331         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
2332         attributes for operands 3 and 4.
2334 2018-11-06  Richard Biener  <rguenther@suse.de>
2336         PR middle-end/18041
2337         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
2338         matching bitfield insertion.
2340 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
2342         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
2343         (reg_next_debug_use): New.
2344         (attempt_change): Propagate adjusted expression into affected
2345         debug insns.
2346         (merge_in_block): Track uses in debug insns.
2347         (pass_inc_dec::execute): Allocate and release
2348         reg_next_debug_use.
2350 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
2352         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
2353         (_mm512_fixupimm_round_pd): Update parameters and builtin.
2354         (_mm512_maskz_fixupimm_round_pd): Ditto.
2355         (_mm512_fixupimm_round_ps): Ditto.
2356         (_mm512_maskz_fixupimm_round_ps): Ditto.
2357         (_mm_fixupimm_round_sd): Ditto.
2358         (_mm_maskz_fixupimm_round_sd): Ditto.
2359         (_mm_fixupimm_round_ss): Ditto.
2360         (_mm_maskz_fixupimm_round_ss): Ditto.
2361         (_mm512_fixupimm_pd): Ditto.
2362         (_mm512_maskz_fixupimm_pd): Ditto.
2363         (_mm512_fixupimm_ps): Ditto.
2364         (_mm512_maskz_fixupimm_ps): Ditto.
2365         (_mm_fixupimm_sd): Ditto.
2366         (_mm_maskz_fixupimm_sd): Ditto.
2367         (_mm_fixupimm_ss): Ditto.
2368         (_mm_maskz_fixupimm_ss): Ditto.
2369         (_mm512_mask_fixupimm_round_pd): Update builtin.
2370         (_mm512_mask_fixupimm_round_ps): Ditto.
2371         (_mm_mask_fixupimm_round_sd): Ditto.
2372         (_mm_mask_fixupimm_round_ss): Ditto.
2373         (_mm512_mask_fixupimm_pd): Ditto.
2374         (_mm512_mask_fixupimm_ps): Ditto.
2375         (_mm_mask_fixupimm_sd): Ditto.
2376         (_mm_mask_fixupimm_ss): Ditto.
2377         * config/i386/avx512vlintrin.h:
2378         (_mm256_fixupimm_pd): Update parameters and builtin.
2379         (_mm256_maskz_fixupimm_pd): Ditto.
2380         (_mm256_fixupimm_ps): Ditto.
2381         (_mm256_maskz_fixupimm_ps): Ditto.
2382         (_mm_fixupimm_pd): Ditto.
2383         (_mm_maskz_fixupimm_pd): Ditto.
2384         (_mm_fixupimm_ps): Ditto.
2385         (_mm_maskz_fixupimm_ps): Ditto.
2386         (_mm256_mask_fixupimm_pd): Update builtin.
2387         (_mm256_mask_fixupimm_ps): Ditto.
2388         (_mm_mask_fixupimm_pd): Ditto.
2389         (_mm_mask_fixupimm_ps): Ditto.
2390         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
2391         * config/i386/i386-builtin.def: Update builtin definitions.
2392         * config/i386/i386.c: Handle new builtin types and remove useless ones.
2393         * config/i386/sse.md: Update VFIXUPIMM* patterns.
2394         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2395         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2396         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
2397         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2398         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2399         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
2400         * config/i386/subst.md:
2401         (round_saeonly_sd_mask_operand4): Add new subst_attr.
2402         (round_saeonly_sd_mask_op4): Ditto.
2403         (round_saeonly_expand_operand5): Ditto.
2404         (round_saeonly_expand): Update.
2406 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
2408         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
2410 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
2412         PR rtl-optimization/87871
2413         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
2415 2018-11-05  Paul Koning  <ni1d@arrl.net>
2417         * doc/sourcebuild.texi (target attributes): Document new "inf"
2418         effective target keyword.
2420 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2422         * config/s390/s390.c (s390_register_move_cost): Increase costs for
2423         moves involving the CC reg.
2425 2018-11-05  Richard Biener  <rguenther@suse.de>
2427         PR tree-optimization/87873
2428         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
2429         argument.
2430         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
2431         * tree-vect-loop.c (vect_transform_loop): When splitting the
2432         loop exit also create forwarder PHIs for constants.
2433         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2434         Handle constant to_arg, add extra checking we match up the correct
2435         PHIs.
2437 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2439         * config/s390/s390.md: QImode and HImode for load on condition.
2441 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2443         * config/s390/predicates.md: Fix typo.
2444         * config/s390/s390.md: Allow immediates for load on condition.
2446 2018-11-05  Martin Liska  <mliska@suse.cz>
2448         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
2449         * bitmap.h (struct bitmap_usage): Likewise.
2450         * ggc-common.c (SCALE): Remove.
2451         (LABEL): Likewise.
2452         (struct ggc_usage): Use SIZE_AMOUNT. And update
2453         compare method.
2454         * ggc-page.c (SCALE): Remove.
2455         (STAT_LABEL): Remove.
2456         (ggc_print_statistics): Use SIZE_AMOUNT.
2457         * gimple.h (SCALE): Remove.
2458         (LABEL): Likewise.
2459         * input.c (ONE_K): Remove.
2460         (ONE_M): Likewise.
2461         (SCALE): Likewise.
2462         (STAT_LABEL): Likewise.
2463         (FORMAT_AMOUNT): Likewise.
2464         (dump_line_table_statistics): Use SIZE_AMOUNT.
2465         * mem-stats.h (struct mem_usage): Likewise.
2466         * rtl.c (dump_rtx_statistics): Likewise.
2467         (rtx_alloc_counts): Change type to size_t.
2468         (rtx_alloc_sizes): Likewise.
2469         (rtx_count_cmp): New.
2470         (dump_rtx_statistics): Sort first based on counts.
2471         * tree.c (tree_nodes_cmp): New.
2472         (tree_codes_cmp): New.
2473         (dump_tree_statistics): Sort first based on counts.
2474         * system.h (ONE_K): New.
2475         (ONE_M): Likewise.
2476         (SIZE_SCALE): Likewise.
2477         (SIZE_LABEL): Likewise.
2478         (SIZE_AMOUNT): Likewise.
2479         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
2480         * tree-dfa.c (dump_dfa_stats): Likewise.
2481         * tree-phinodes.c (phinodes_print_statistics): Likewise.
2482         * tree-ssanames.c (ssanames_print_statistics): Likewise.
2483         * tree.c (dump_tree_statistics): Likewise.
2484         * vec.c (struct vec_usage): Likewise.
2485         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
2486         trigger a -Werror=format-overflow with
2487         --enable-gather-detailed-stats.
2489 2018-11-05  Martin Liska  <mliska@suse.cz>
2491         * mem-stats.h (mem_alloc_description::release_instance_overhead):
2492         Return T *.
2493         * vec.c (struct vec_usage): Register m_element_size.
2494         (vec_prefix::register_overhead): New arguments: elements and
2495         element_size.
2496         (vec_prefix::release_overhead): Subtract elements.
2497         * vec.h (struct vec_prefix): Change signature.
2498         (va_heap::reserve): Pass proper arguments.
2499         (va_heap::release): Likewise.
2501 2018-11-05  Martin Liska  <mliska@suse.cz>
2503         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
2504         style.
2505         * vec.c: Likewise.
2507 2018-11-05  Richard Biener  <rguenther@suse.de>
2509         * tree-scalar-evolution.h (final_value_replacement_loop): Update
2510         prototype.
2511         * tree-scalar-evolution.c (final_value_replacement_loop): Return
2512         whether anything was done.
2513         (scev_const_prop): Remove constant propagation part, fold
2514         remains into ...
2515         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
2516         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
2517         conditionally.
2519 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
2521         PR tree-optimization/87859
2522         * gimple-ssa-store-merging.c (struct merged_store_group): Add
2523         only_constants and first_nonmergeable_order members.
2524         (merged_store_group::merged_store_group): Initialize them.
2525         (merged_store_group::do_merge): Clear only_constants member if
2526         adding something other than INTEGER_CST store.
2527         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
2528         stores with order >= first_nonmergeable_order.  Use
2529         merged_store->only_constants instead of always recomputing it.
2530         Set merged_store->first_nonmergeable_order if we've skipped any
2531         stores.  Attempt to merge overlapping INTEGER_CST stores that
2532         we would otherwise skip.
2534         PR sanitizer/87837
2535         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
2537 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
2539         PR target/87853
2540         * config/i386/emmintrin.h (__v16qs): New to cope with option
2541         -funsigned-char.
2542         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
2543         (_mm_cmplt_epi8): Likewise.
2544         (_mm_cmpgt_epi8): Likewise.
2546 2018-11-05  Richard Biener  <rguenther@suse.de>
2548         PR rtl-optimization/87852
2549         * fwprop.c (use_killed_between): Only consider single-defs of the
2550         use whose definition statement dominates the use.
2552 2018-11-05  Martin Liska  <mliska@suse.cz>
2554         PR web/87829
2555         * doc/invoke.texi: Remove options that are
2556         not disabled with -Os.
2558 2018-11-05  Martin Liska  <mliska@suse.cz>
2560         PR c/87811
2561         * doc/extend.texi: Update constrain about the last argument
2562         of __builtin_expect_with_probability.
2564 2018-11-05  Martin Liska  <mliska@suse.cz>
2566         PR c/87811
2567         * predict.c (expr_expected_value_1): Verify
2568         that last argument is a real constants and emit
2569         error.
2571 2018-11-05  Martin Liska  <mliska@suse.cz>
2573         PR gcov-profile/77698
2574         * ipa-profile.c (ipa_profile): Adjust hotness threshold
2575         only in LTO mode.
2577 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2579         PR tree-optimization/86572
2580         * builtins.c (c_strlen): Handle negative offsets in a safe way.
2582 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2584         PR tree-optimization/87672
2585         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
2586         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
2588 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
2590         PR middle-end/58372
2591         * cfgexpand.c (pass_expand::execute): Move the call to
2592         finish_eh_generation in front of the call to expand_stack_alignment.
2594 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
2596         * common/config/i386/i386-common.c (processor_alias_table): Add
2597         znver2 entry.
2598         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
2599         (case ${target}): Add znver2.
2600         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
2601         -march=native recognize znver2 processors.
2602         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
2603         * config/i386/i386.c (m_znver2): New definition.
2604         (m_ZNVER): New definition.
2605         (m_AMD_MULTIPLE): Includes m_znver2.
2606         (processor_cost_table): Add znver2 entry.
2607         (processor_target_table): Add znver2 entry.
2608         (get_builtin_code_for_version): Set priority for
2609         PROCESSOR_ZNVER2.
2610         (processor_model): Add M_AMDFAM17H_ZNVER2.
2611         (arch_names_table): Ditto.
2612         (ix86_reassociation_width): Include znver2.
2613         * config/i386/i386.h (TARGET_znver2): New definition.
2614         (struct ix86_size_cost): Add TARGET_ZNVER2.
2615         (enum processor_type): Add PROCESSOR_ZNVER2.
2616         * config/i386/i386.md (define_attr "cpu"): Add znver2.
2617         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
2618         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
2619         (ix86_adjust_cost): Add znver2.
2620         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
2621         * gcc/doc/extend.texi: Add details about znver2.
2622         * gcc/doc/invoke.texi: Add details about znver2.
2624 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
2626         PR target/87079
2628         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
2629         pattern.
2631 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
2633         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
2634         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
2636 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
2638         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
2639         shrink-wrapping when -mtrack-speculation.
2641 2018-11-02  Richard Biener  <rguenther@suse.de>
2643         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
2644         (pop_cost_one_pair): Do not free pair.
2645         (pop_best_coalesce): Likewise.
2646         (create_coalesce_list): Initialize obstack.
2647         (delete_coalesce_list): Free obstack.
2648         (find_coalesce_pair): Obstack-allocate coalesce pairs.
2649         (add_cost_one_coalesce): Likewise.
2650         (struct live_track): Remove bitmap pointer indirections.
2651         (new_live_track): Adjust.
2652         (delete_live_track): Likewise.
2653         (live_track_remove_partition): Likewise.
2654         (live_track_add_partition): Likewise.
2655         (live_track_live_p): Likewise.
2656         (live_track_process_def): Likewise.
2657         (live_track_clear_base_vars): Likewise.
2659 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2661         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
2662         * configure: Regenerate.
2663         * config.in: Regenerate.
2664         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
2665         HAVE_AS_REGISTER_PSEUDO_OP guard.
2666         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
2667         !HAVE_AS_REGISTER_PSEUDO_OP support.
2668         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
2669         guard.
2671 2018-11-02  Richard Biener  <rguenther@suse.de>
2673         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
2674         with is_type_die.
2676 2018-11-02  Richard Biener  <rguenther@suse.de>
2678         PR tree-optimization/87776
2679         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
2680         executable when iterating but running into rpo-vn-max-loop-depth
2681         and not eliding the iteration.
2683 2018-11-30  Jan Hubicka  <jh@suse.cz>
2685         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
2686         TYPE_DECL.
2688 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
2690         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
2691         prototype.
2692         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
2693         Combine with rs6000_address_for_fpconvert.
2694         (rs6000_address_for_fpconvert): Combine with
2695         rs6000_force_indexed_or_indirect_mem.
2696         (rs6000_expand_vector_init): Change function call from
2697         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2698         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
2699         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2700         (floatsi<mode>2_lfiwax_mem): Ditto.
2701         (floatunssi<mode>2_lfiwzx): Ditto.
2702         (floatunssi<mode>2_lfiwzx_mem): Ditto.
2703         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
2704         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
2705         (fix_trunc<mode>si2_stfiwx): Ditto.
2706         (fixuns_trunc<mode>si2_stfiwx): Ditto.
2707         (float_<mode>si2_hw): Ditto.
2708         (floatuns_<mode>si2_hw): Ditto.
2709         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
2710         (vsx_splat_<mode>): Ditto.
2712 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
2714         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
2715         at end of assembler input text.
2716         * configure: Regenerate.
2718 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
2720         PR tree-optimization/87826
2721         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
2722         negative or larger or equal to type's precision.
2724 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
2726         * opts.c (default_options_table): Do not enable
2727         OPT_fdelayed_branch at -Og.
2728         * doc/invoke.texi (-fdelayed-branch): Document it.
2730 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2732         * optabs-libfuncs.c (build_libfunc_function_visibility):
2733         New, split out from...
2734         (build_libfunc_function): ... here.
2735         (init_one_libfunc_visibility): New, split out from ...
2736         (init_one_libfunc): ... here.
2738         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
2739         scratch register need not be early-clobber.  Document the reason
2740         why we cannot use ST<OP>.
2742 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
2744         PR bootstrap/82856
2745         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
2746         line for second argument of AC_DEFINE_UNQUOTED.
2747         * doc/install.texi (Tools/packages necessary for modifying GCC):
2748         Update to autoconf 2.69 and automake 1.15.1.
2749         * aclocal.m4, config.in, configure: Regenerate.
2751 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
2753         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
2754         initialization.
2756 2018-10-31  Martin Liska  <mliska@suse.cz>
2758         PR driver/83193
2759         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
2760         Add new argument invalid_extension.
2761         (aarch64_get_all_extension_candidates): New function.
2762         (aarch64_rewrite_selected_cpu): Add NULL to function call.
2763         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
2764         new argument.
2765         (aarch64_get_all_extension_candidates): New function.
2766         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
2767         argument invalid_extension.
2768         (aarch64_parse_cpu): Likewise.
2769         (aarch64_print_hint_for_extensions): New function.
2770         (aarch64_validate_mcpu): Provide hint about invalid extension.
2771         (aarch64_validate_march): Likewise.
2772         (aarch64_handle_attr_arch): Pass new argument.
2773         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
2774         (aarch64_handle_attr_isa_flags): Likewise.
2776 2018-10-31  Richard Biener  <rguenther@suse.de>
2778         PR middle-end/70359
2779         PR middle-end/86270
2780         * tree-outof-ssa.c (insert_backedge_copies): Restrict
2781         copy generation to useful cases.  Place the copy before
2782         the definition of the backedge value when possible.
2784 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2786         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
2787         * config/arc/arc.c (arc_active_insn): New function.
2788         (check_store_cacheline_hazard): Likewise.
2789         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
2790         (arc_override_options): Disable delay slot scheduler for older
2791         A7.
2792         (arc_store_addr_hazard_p): New implementation, old one renamed to
2793         ...
2794         (arc_store_addr_hazard_internal_p): Renamed.
2795         (arc_reorg): Don't combine into brcc instructions which are part
2796         of hardware hazard solution.
2797         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
2798         (tune_arc700): Likewise.
2799         * config/arc/arc.opt (arc7xx): New tune value.
2800         * config/arc/arc700.md: Improve A7 scheduler.
2802 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2804         * config/arc/arc.c (arc_override_options): Remove
2805         TARGET_COMPACT_CASESI.
2806         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
2807         (CASE_VECTOR_MODE): Likewise.
2808         (CASE_VECTOR_PC_RELATIVE): Likewise.
2809         (CASE_VECTOR_SHORTEN_MODE): Likewise.
2810         (CASE_VECTOR_SHORTEN_MODE1): Delete.
2811         (ADDR_VEC_ALIGN): Update.
2812         (ASM_OUTPUT_CASE_LABEL): Undefine.
2813         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
2814         (TARGET_BI_BIH): Define.
2815         (DEFAULT_BRANCH_INDEX): Likewise.
2816         * config/arc/arc.md (casesi): Rework to accept BI/BIH
2817         instructions, remove compact_casesi use case.
2818         (casesi_compact_jump): Remove.
2819         (casesi_dispatch): New pattern.
2820         * config/arc/arc.opt: Add mbranch-index option. Deprecate
2821         compact_casesi option.
2822         * doc/invoke.texi: Document mbranch-index option.
2824 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2826         * config/arc/arc.c (arc_get_tp): Remove function.
2827         (arc_emit_call_tls_get_addr): Likewise.
2828         (arc_call_tls_get_addr): New function.
2829         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
2830         * config/arc/arc.md (tls_load_tp_soft): Remove.
2831         (tls_gd_get_addr): Likewise.
2833 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2835         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
2836         (mulsi3_600_lib): Remove pattern.
2837         (umulsi3_highpart_600_lib_le): Likewise.
2838         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
2839         (umulsidi3): Remove call to umulsidi3_600_lib.
2840         (umulsidi3_600_lib): Remove pattern.
2841         (peephole2): Remove peephole using the above deprecated patterns.
2843 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
2845         PR target/87374
2846         * config/arm/arm.c (arm_option_check_internal): Disable the combined
2847         use of -mslow-flash-data and -mword-relocations.
2848         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
2849         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
2850         flag_pic.
2851         * doc/invoke.texi (-mword-relocations): Mention conflict with
2852         -mslow-flash-data.
2853         (-mslow-flash-data): Reciprocally.
2855 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2857         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
2858         16-byte modes held in GP registers to use an even regno.
2860         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
2861         (aarch64_atomic_ldop_supported_p): Remove.
2862         (aarch64_gen_atomic_ldop): Remove.
2863         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
2864         Fully expand LSE operations here.
2865         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
2866         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
2867         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
2868         and use ATOMIC_LDOP instead; use register_operand for the input;
2869         drop the split and emit insns directly.
2870         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
2871         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
2872         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
2874         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
2875         (aarch64_gen_atomic_ldop): Don't call it.
2876         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
2877         Use aarch64_reg_or_zero.
2878         (aarch64_atomic_exchange<ALLI>): Likewise.
2879         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
2880         operand 0; use aarch64_reg_or_zero for input; merge ...
2881         (@aarch64_atomic_swp<ALLI>): ... this and remove.
2883         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
2884         (aarch64_split_compare_and_swap): Use it.
2885         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
2886         test oldval against the proper predicate.
2887         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
2888         Use nonmemory_operand for expected.
2889         (cas_short_expected_pred): New.
2890         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
2891         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
2892         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
2893         (aarch64_plushi_operand): New.
2895         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
2896         Force oldval into the rval register for TARGET_LSE; emit the compare
2897         during initial expansion so that it may be deleted if unused.
2898         (aarch64_gen_atomic_cas): Remove.
2899         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
2900         Change =&r to +r for operand 0; use match_dup for operand 2;
2901         remove is_weak and mod_f operands as unused.  Drop the split
2902         and merge with...
2903         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
2904         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
2905         (@aarch64_atomic_cas<GPI>): Similarly.
2907 2018-10-31  Richard Biener  <rguenther@suse.de>
2909         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
2910         using ABSU_EXPR.
2912 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2914         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
2915         * config/aarch64/aarch64.md: Include saphira.md
2916         * config/aarch64/saphira.md: New file for pipeline description.
2918 2018-10-30  Martin Sebor  <msebor@redhat.com>
2920         PR middle-end/87041
2921         * gimple-ssa-sprintf.c (format_directive): Use %G to include
2922         inlining context.
2923         (sprintf_dom_walker::compute_format_length):
2924         Avoid setting POSUNDER4K here.
2925         (get_destination_size): Handle null argument values.
2926         (get_user_idx_format): New function.
2927         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
2928         functions, including user-defined with attribute format printf.
2929         Use %G to include inlining context.
2930         Set POSUNDER4K here.
2932 2018-10-30  Jan Hubicka  <jh@suse.cz>
2934         * params.def (lto-partitions): Bump from 32 to 128.
2936 2018-10-30  Jan Hubicka  <jh@suse.cz>
2938         * tree.c
2939         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
2940         head in file.
2941         (free_lang_data_in_type): Forward declare.
2942         (fld_type_variant_equal_p): New function.
2943         (fld_type_variant): New function
2944         (fld_incomplete_types): New hash.
2945         (fld_incomplete_type_of): New function
2946         (fld_simplfied-type): New function.
2947         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
2948         (free_lang_data): Allocate and free fld_incomplete_type; update call
2949         of free_lang_data_in_decl.
2951 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
2953         * gcov.c (output_lines): Remove duplicate line.
2955 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
2957         * config/rs6000/rs6000.md (bswapdi2): Force address into register
2958         if not in indexed or indirect form.
2959         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
2960         (bswapdi2_store): Ditto.
2961         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
2962         helper function.
2963         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
2964         Prototype for helper function.
2966 2018-10-30  Martin Sebor  <msebor@redhat.com>
2968         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
2969         (target, pragma GCC optimize, pragma GCC target): Ditto.
2971 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
2973         * doc/extend.texi: Fix prototype and description of
2974         __builtin_expect_with_probability.
2976 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2978         * cgraph.h (clone_function_name_1): Replaced by new
2979           clone_function_name_numbered that takes name as string; for
2980           privatize_symbol_name_1 use only.
2981           (clone_function_name): Renamed to
2982           clone_function_name_numbered to be explicit about numbering.
2983           (clone_function_name): New two-argument function that does
2984           not number its output.
2985           (clone_function_name): New three-argument function that
2986           takes a number to append to its output.
2987         * cgraphclones.c (duplicate_thunk_for_node):
2988           (clone_function_name_1): Renamed.
2989           (clone_function_name_numbered): Two new functions.
2990           (clone_function_name): Improved documentation.
2991           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
2992         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
2993         * final.c (final_scan_insn_1): Use the new clone_function_name
2994           without numbering.
2995         * multiple_target.c (create_dispatcher_calls): Ditto.
2996           (create_target_clone): Ditto.
2997         * omp-expand.c (grid_expand_target_grid_body): Ditto.
2998         * omp-low.c (create_omp_child_function_name): Ditto.
2999         * omp-simd-clone.c (simd_clone_create): Ditto.
3000         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
3001           new clone_function_name without numbering.
3003 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
3005         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
3006         Assert that the allocation size is not zero.
3008 2018-10-30  Richard Biener  <rguenther@suse.de>
3010         PR tree-optimization/87800
3011         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
3012         non-induction or reduction PHIs.
3014 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
3016         * config/aarch64/falkor-tag-collision-avoidance.c
3017         (execute_tag_collision_avoidance): Call df_note_add_problem.
3019 2018-10-30  Martin Liska  <mliska@suse.cz>
3021         * doc/extend.texi: Fix typo in documentation
3022         of __builtin_expect_with_probability.
3024 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
3026         PR c++/87721
3027         * input.c (get_substring_ranges_for_loc): Detect if
3028         linemap_resolve_location gives us a NULL map, and reject
3029         this case.
3031 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
3033         * config.gcc (xstormy16-*-elf): Set tm_d_file.
3035 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
3036             Martin Sebor  <msebor@redhat.com>
3037             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
3039         * doc/ux.texi (Quoting): New subsection, adapted from material at
3040         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
3041         MartinSebor and ManuelLopezIbanez.
3042         (Fix-it hints): Note that fix-it hints shouldn't be marked for
3043         translation.
3045 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
3047         PR middle-end/87469
3048         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
3049         max value.
3051 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
3053         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
3055 2018-10-29  Paul Koning  <ni1d@arrl.net>
3057         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
3059 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
3061         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
3062         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
3063         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
3064         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
3065         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
3066         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
3067         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
3068         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
3069         Change 'vector' to '__vector'.
3070         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
3071         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
3072         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
3073         _mm_avg_pu16): Likewise.  And, whitespace corrections.
3075 2018-10-29  Richard Biener  <rguenther@suse.de>
3077         PR tree-optimization/87785
3078         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
3079         internal defs.
3081 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3083         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
3085 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3087         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
3088         ports configuration macro, defaults to "gnu".
3089         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
3090         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
3092 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3094         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
3096 2018-10-29  Richard Biener  <rguenther@suse.de>
3098         PR tree-optimization/87790
3099         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
3100         (vect_make_slp_decision): Adjust.
3101         (vect_slp_analyze_bb_1): Likewise.
3102         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
3103         edges.
3105 2018-10-29  Richard Biener  <rguenther@suse.de>
3107         PR tree-optimization/87785
3108         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
3109         and processing.
3110         (vect_build_slp_tree): Likewise.
3111         (vect_gather_slp_loads): New function.
3112         (vect_analyze_slp_instance): Gather loads separately from the
3113         SLP tree build.
3115 2018-10-29  Martin Liska  <mliska@suse.cz>
3117         * Makefile.in: Make dependency to json.o.
3118         * doc/gcov.texi: Document new JSON format, remove
3119         old intermediate format documentation.
3120         * gcov.c (struct function_info): Come up with m_name and
3121         m_demangled_name.
3122         (function_info::function_info): Initialize it.
3123         (function_info::~function_info): Release it.
3124         (main): Rename flag_intermediate_format to flag_json_format.
3125         (print_usage): Describe --json-format.
3126         (process_args): Set flag_json_format.
3127         (output_intermediate_line): Remove.
3128         (output_intermediate_json_line): Likewise.
3129         (get_gcov_intermediate_filename): Return new extension
3130         ".gcov.json.gz".
3131         (output_intermediate_file): Implement JSON emission.
3132         (output_json_intermediate_file): Implement JSON emission.
3133         (generate_results): Use ::get_name for function name.
3134         Handle JSON output file.
3135         (read_graph_file): Use ::get_name instead of cplus_demangle.
3136         (read_count_file): Likewise.
3137         (solve_flow_graph): Likewise.
3138         (add_line_counts): Likewise.
3139         (accumulate_line_counts): Use new flag_json_format.
3140         (output_function_details): Use ::get_name instead of cplus_demangle.
3141         (output_lines): Likewise.
3142         * json.cc (test_writing_literals): Add new tests.
3143         * json.h (class literal): Add new boolean constructor.
3145 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
3147         PR rtl-optimization/87701
3148         PR rtl-optimization/87780
3149         * combine.c (make_more_copies): Rewrite.
3151 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
3153         * doc/generic.texi (ABSU_EXPR): Document.
3154         * match.pd (absu(x)*absu(x) -> x*x): Handle.
3155         (absu(absu(X)) -> absu(X)): Likewise.
3156         (absu(-X) -> absu(X)): Likewise.
3157         (absu(X)  where X is nonnegative -> X): Likewise.
3159 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
3161         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
3162         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
3163         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
3164         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
3165         (s-tm-texi): Also check timestamp on d-target.def.
3166         (generated_files): Add TM_D_H and d-target-hooks-def.h.
3167         (build/genhooks.o): Also depend on D_TARGET_DEF.
3168         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
3169         variables.
3170         * config/aarch64/aarch64-d.c: New file.
3171         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
3172         Define.
3173         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
3174         prototype.
3175         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
3176         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
3177         * config/arm/arm-d.c: New file.
3178         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
3179         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
3180         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3181         * config/arm/t-arm (arm-d.o): New rule.
3182         * config/default-d.c: New file.
3183         * config/glibc-d.c: New file.
3184         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3185         * config/i386/i386-d.c: New file.
3186         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
3187         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
3188         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3189         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
3190         * config/i386/t-i386 (i386-d.o): New rule.
3191         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3192         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3193         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
3194         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3195         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3196         * config/mips/mips-d.c: New file.
3197         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
3198         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
3199         * config/mips/t-mips (mips-d.o): New rule.
3200         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3201         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3202         * config/powerpcspe/powerpcspe-d.c: New file.
3203         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
3204         New prototype.
3205         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3206         Support GNU D by using 0 as the language type.
3207         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
3208         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
3209         * config/riscv/riscv-d.c: New file.
3210         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
3211         prototype.
3212         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
3213         * config/riscv/t-riscv (riscv-d.o): New rule.
3214         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3215         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3216         * config/rs6000/rs6000-d.c: New file.
3217         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
3218         prototype.
3219         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3220         Support GNU D by using 0 as the language type.
3221         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
3222         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
3223         * config/s390/s390-d.c: New file.
3224         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
3225         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
3226         * config/s390/t-s390 (s390-d.o): New rule.
3227         * config/sparc/sparc-d.c: New file.
3228         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
3229         prototype.
3230         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
3231         * config/sparc/t-sparc (sparc-d.o): New rule.
3232         * config/t-glibc (glibc-d.o): New rule.
3233         * configure: Regenerated.
3234         * configure.ac (tm_d_file): New variable.
3235         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
3236         * doc/contrib.texi (Contributors): Add self for the D frontend.
3237         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
3238         * doc/install.texi (Configuration): Mention libphobos as an option for
3239         --enable-shared.  Mention d as an option for --enable-languages.
3240         (Testing): Mention check-d as a target.
3241         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
3242         name suffixes.  Mention d as a -x option.
3243         * doc/sourcebuild.texi (Top Level): Mention libphobos.
3244         * doc/standards.texi (Standards): Add section on D language.
3245         * doc/tm.texi: Regenerated.
3246         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
3247         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
3248         * dwarf2out.c (is_dlang): New function.
3249         (gen_compile_unit_die): Use DW_LANG_D for D.
3250         (declare_in_namespace): Return module die for D, instead of adding
3251         extra declarations into the namespace.
3252         (gen_namespace_die): Generate DW_TAG_module for D.
3253         (gen_decl_die): Handle CONST_DECLSs for D.
3254         (dwarf2out_decl): Likewise.
3255         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
3256         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
3257         kinds of aggregates.
3258         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
3259         * genhooks.c: Include d/d-target.def.
3261 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
3263         PR target/85669
3264         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
3265         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
3266         (STACK_DYNAMIC_OFFSET): Likewise.
3268 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
3270         PR target/80024
3271         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
3272         error message.
3274 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
3276         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
3277         return type and other typos.
3279 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
3281         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
3282         a shorter sequence with fewer branches.
3283         (emit_final_str_compare_gpr): Ditto.
3285 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3287         * config/rs6000/tmmintrin.h: New file.
3288         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
3290 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3292         * config/rs6000/mmintrin.h: Enable 32bit compilation.
3293         * config/rs6000/xmmintrin.h: Likewise.
3295 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3297         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
3299 2018-10-26  Richard Biener  <rguenther@suse.de>
3301         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
3302         and wrapper.
3303         (vect_mark_slp_stmts_relevant): Likewise.
3304         (vect_detect_hybrid_slp_stmts): Likewise.
3305         (vect_bb_slp_scalar_cost): Likewise.
3306         (vect_remove_slp_scalar_calls): Likewise.
3308 2018-10-26  Jan Hubicka  <jh@suse.cz>
3310         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
3311         (warn_types_mismatch): Fix walk of DECL_NAME.
3312         (odr_types_equivalent_p): Fix overactive assert.
3314 2018-10-26  Richard Biener  <rguenther@suse.de>
3316         PR tree-optimization/87105
3317         * tree-vectorizer.h (_slp_tree::refcnt): New member.
3318         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
3319         refcnt.
3320         (vect_create_new_slp_node): Initialize refcnt to one.
3321         (bst_traits): Move.
3322         (scalar_stmts_set_t, bst_fail): Remove.
3323         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
3324         (vect_build_slp_tree): Add bst_map argument and lookup
3325         already created SLP nodes.
3326         (vect_print_slp_tree): Handle a SLP graph, print SLP node
3327         addresses.
3328         (vect_slp_rearrange_stmts): Handle a SLP graph.
3329         (vect_analyze_slp_instance): Adjust and free SLP nodes from
3330         the CSE map.  Fix indenting.
3331         (vect_schedule_slp_instance): Add short-cut.
3333 2018-10-26  Martin Liska  <mliska@suse.cz>
3335         PR testsuite/86158
3336         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
3337         addr_expr and not with pointers.
3339 2018-10-26  Jan Hubicka  <jh@suse.cz>
3341         * tree.c (free_lang_data_in_type): Only check main variants.
3342         * ipa-devirt.c (warn_odr): Make static.
3343         (types_same_for_odr): Drop strict variant.
3344         (types_odr_comparable): Likewise.
3345         (odr_or_derived_type_p): Look for main variants.
3346         (odr_name_hasher::equal): Cleanup comment.
3347         (odr_subtypes_equivalent): Add warn and warned arguments; check main
3348         variants.
3349         (type_variants_equivalent_p): break out from ...
3350         (odr_types_equivalent): ... here; go for main variants where needed.
3351         (warn_odr): ... here; turn static.
3352         (warn_types_mismatch): Compare mangled names of main variants.
3353         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
3354         (type_with_linkage_p): Sanity check that we look at main variant.
3355         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
3356         * tree.h (types_same_for_odr): Drop strict argument.
3358 2018-10-26  Richard Biener  <rguenther@suse.de>
3360         PR tree-optimization/87746
3361         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
3362         Simplify and fix WRT strided store groups with size not
3363         equal to step in element count.
3364         (vect_analyze_group_access_1): Dump the whole group.
3366 2018-10-25  Carl Love  <cel@us.ibm.com>
3368         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
3369         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
3370         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
3371         precicion entry for each overloaded builtin.
3372         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
3373         VSCEDPUO): Rename overloaded name.
3374         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
3375         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
3376         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
3377         define_expand for xscmpexqp instruction.
3378         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
3380 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3381             Jinsong Ji <jji@us.ibm.com>
3383         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
3384         function with vec_sl.
3385         (_mm_slli_epi32): Likewise.
3386         (_mm_slli_epi64): Likewise.
3387         (_mm_srai_epi16): Replace deprecated function with vec_sra.
3388         (_mm_srai_epi32): Likewise.
3389         (_mm_srli_epi16): Replace deprecated function with vec_sr.
3390         (_mm_srli_epi32): Likewise.
3391         (_mm_srli_epi64): Likewise.
3392         (_mm_sll_epi16): Replace deprecated function with vec_sl.
3393         (_mm_sll_epi32): Likewise.
3394         (_mm_sll_epi64): Likewise.
3395         (_mm_sra_epi16): Replace deprecated function with vec_sra.
3396         (_mm_sra_epi32): Likewise.
3397         (_mm_srl_epi16): Replace deprecated function with vec_sr.
3398         (_mm_srl_epi32): Likewise.
3399         (_mm_srl_epi64): Likewise.
3401 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3402             Jinsong Ji <jji@us.ibm.com>
3404         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
3405         comparison operators with vec_cmp* for compatibility due to
3406         unfortunate history; clean up formatting and use types more
3407         appropriately.
3408         (_mm_sll_epi32): Likewise.
3409         (_mm_sll_epi64): Likewise.
3410         (_mm_srl_epi16): Likewise.
3411         (_mm_srl_epi32): Likewise.
3412         (_mm_srl_epi64): Likewise.
3414 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3415             Jinsong Ji <jji@us.ibm.com>
3417         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
3418         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
3419         __vector __bool int.  Use vec_cmpgt in preference to deprecated
3420         function vec_vcmpgtfp.
3421         (_mm_max_ps): Likewise.
3423 2018-10-25  Jeff Law  <law@redhat.com>
3425         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
3426         if returning false.
3428 2018-10-25  Martin Sebor  <msebor@redhat.com>
3430         * doc/extend.texi (aligned): Expand attribute description.
3431         (Alignment): Rename section.  Discuss function arguments.
3433 2018-10-25  Jan Hubicka  <jh@suse.cz>
3435         * ipa-devirt.c (main_odr_variant): Remove.
3436         (hash_odr_name, types_same_for_odr, types_odr_comparable,
3437         odr_name_hasher::equal, odr_subtypes_equivalent_p):
3438         Drop use of main_odr_variant.
3439         (add_type_duplicate): Silence confused warnings on integer types.
3440         (get_odr_type): Always look for main variant.
3441         (register_odr_type): Simplify.
3443 2018-10-25  Richard Biener  <rguenther@suse.de>
3445         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
3446         Initialize ng to silence error with release checking bootstrap.
3448 2018-10-25  Richard Biener  <rguenther@suse.de>
3450         * tree-if-conv.c: Include tree-ssa-sccvn.h.
3451         (tree_if_conversion): Run CSE on the if-converted loop body.
3453 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3455         * config/s390/constraints.md (ZL): New constraint.
3456         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
3457         operands.
3458         * config/s390/s390.md (movdi_larl): Remove.
3459         (movdi_64): Add the LARL alternative.
3461 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3463         PR bootstrap/87747
3464         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
3465         (RTX_CODE_HWINT_P): New macro.
3466         (rtx_code_size): Use RTX_CODE_HWINT_P ().
3468 2018-10-25  Jan Hubicka  <jh@suse.cz>
3470         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
3471         is anonymous.
3473 2018-10-25  Richard Biener  <rguenther@suse.de>
3475         PR tree-optimization/87665
3476         PR tree-optimization/87745
3477         * tree-vectorizer.h (get_earlier_stmt): Remove.
3478         (get_later_stmt): Pick up UID from the original non-pattern stmt.
3480 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
3482         * options.texi (Deprecated): Move list to Var section.
3484 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
3485             Jinsong Ji <jji@us.ibm.com>
3487         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
3488         __vector long to __vector long long.
3489         (_mm_cvtpd_ps): Likewise.
3490         (_mm_cvttpd_epi32): Likewise.
3491         (_mm_cvtpi32_pd): Likewise.
3492         (_mm_unpackhi_epi64): Likewise.
3493         (_mm_unpacklo_epi64): Likewise.
3495 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
3497         PR rtl-optimization/87720
3498         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
3500 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
3502         * gimple-ssa-isolate-paths.c
3503         (find_implicit_erroneous_behavior): Do not change code if the
3504         pass is running for warnings only.
3505         (find_explicit_erroneous_behavior): Likewise.
3507 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
3509         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
3510         Define as rs6000_mangle_decl_assembler_name.
3511         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
3512         long double to IEEE long double, switch the names of the long
3513         double built-in functions to be <func>f128 instead of <func>l.
3515 2018-10-24  Martin Sebor  <msebor@redhat.com>
3517         * doc/extend.texi (nonnull): List no-argument form.  Reference
3518         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
3520 2018-10-24  Richard Biener  <rguenther@suse.de>
3522         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
3524 2018-10-24  Martin Liska  <mliska@suse.cz>
3526         PR tree-optimization/84436
3527         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
3528         Remove.
3529         (switch_conversion::contains_linear_function_p): New.
3530         (switch_conversion::build_one_array): Support linear
3531         transformation on input.
3532         * tree-switch-conversion.h (struct switch_conversion): Add
3533         contains_linear_function_p declaration.
3535 2018-10-24  Richard Biener  <rguenther@suse.de>
3537         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
3538         if its argument is CONSTANT_CLASS_P.
3540 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
3542         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
3543         it is wrong for forward declarations.
3545 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3547         * config/s390/s390.c (s390_check_qrst_address): Add the missing
3548         SYMBOL_REF_P () check.
3550 2018-10-24  Richard Biener  <rguenther@suse.de>
3552         PR tree-optimization/87105
3553         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
3554         dump classification.
3555         (vect_analyze_data_ref_accesses): Handle duplicate loads and
3556         stores by splitting the affected group after the fact.
3557         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
3558         fail the SLP build because of size constraints.
3560 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3562         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
3563         * configure: Regenerate.
3564         * config.in: Regenerate.
3565         * varasm.c (mergeable_string_section): Use readonly_data_section
3566         if linker doesn't support SHF_MERGE with alignment > 8.
3567         (mergeable_constant_section): Likewise.
3569 2018-10-24  Richard Biener  <rguenther@suse.de>
3571         PR tree-optimization/84013
3572         * tree-ssa-structalias.c (struct msdi_data): New struct for
3573         marshalling data to walk_stmt_load_store_ops.
3574         (maybe_set_dependence_info): Refactor as callback for
3575         walk_stmt_load_store_ops.
3576         (compute_dependence_clique): Set restrict info on all stmt kinds.
3578 2018-10-24  Martin Liska  <mliska@suse.cz>
3580         * cgraph.c (cgraph_node::dump):
3581         Remove reduntant dumps and make tp_first_run dump more compact.
3583 2018-10-24  Richard Biener  <rguenther@suse.de>
3585         PR tree-optimization/87665
3586         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
3587         to reflect reality.
3589 2018-10-12  Jeff Law  <law@redhat.com>
3591         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
3592         for H8/S.
3594 2018-10-23  Richard Biener  <rguenther@suse.de>
3596         * tree-vrp.c (add_assert_info): Guard dump_printf with
3597         dump_enabled_p.
3598         * gimple-ssa-evrp-analyze.c
3599         (evrp_range_analyzer::record_ranges_from_incoming_edge):
3600         Use value_range::ignore_equivs_equal_p.
3602 2018-10-23  Richard Biener  <rguenther@suse.de>
3604         PR tree-optimization/87105
3605         PR tree-optimization/87608
3606         * passes.def (pass_all_early_optimizations): Add early phi-opt
3607         after dce.
3608         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
3609         addition to debug stmts.
3610         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
3611         and abs replacement early.
3612         * tree-cfg.c (gimple_empty_block_p): Likewise.
3614 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
3616         PR target/86383
3617         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
3618         specified to configure.
3619         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
3621 2018-10-23  Richard Biener  <rguenther@suse.de>
3623         PR tree-optimization/87700
3624         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
3626 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
3628         PR target/87674
3629         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
3630         second argument from __mmask16 to __mmask8.
3631         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
3632         _mm_mask_packs_epi32): Likewise.
3633         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
3634         Likewise.
3635         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
3637 2018-10-23  Richard Biener  <rguenther@suse.de>
3639         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
3641 2018-10-23  Richard Biener  <rguenther@suse.de>
3643         PR tree-optimization/86144
3644         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
3645         over simd attribute.
3647 2018-10-23  Richard Biener  <rguenther@suse.de>
3649         PR tree-optimization/87693
3650         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
3651         the case we do not find the taken edge.
3653 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
3654             Jinsong Ji  <jji@us.ibm.com>
3656         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
3657         (_mm_store_pd): Use unaligned vector type for pointer cast.
3658         (_mm_maskmoveu_si128): Likewise.
3659         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
3660         (_mm_store_ps): Use unaligned vector type for pointer cast.
3662 2018-10-22  Paul Koning  <ni1d@arrl.net>
3664         * symtab.c (symtab_node::increase_alignment): Correct max
3665         alignment check.
3667 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
3669         PR tree-optimization/87633
3670         * match.pd: Do not generate unordered integer comparisons.
3672 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3674         PR rtl-optimization/87600
3675         * combine.c: Add include of expr.h.
3676         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
3677         register to a pseudo.
3678         (make_more_copies): New function, add a copy to a new pseudo after
3679         the moves from hard registers into pseudos.
3680         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
3681         later.  Call make_more_copies.
3683 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
3685         * lra-constraints.c (process_alt_operands): New local array,
3686         matching_early_clobber.  Check matching_early_clobber before
3687         decrementing reject, and set matching_early_clobber after.
3689 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3691         PR target/87598
3692         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
3693         call output_addr_const and hope for the best.
3695 2018-10-22  Richard Biener  <rguenther@suse.de>
3697         * gimple-ssa-evrp-analyze.c
3698         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
3699         smarter about what ranges to use.
3700         * tree-vrp.c (add_assert_info): Dump here.
3701         (register_edge_assert_for_2): Instead of here at multiple but
3702         not all places.
3704         * gcc.dg/tree-ssa/evrp12.c: New testcase.
3705         * gcc.dg/predict-6.c: Adjust.
3706         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
3707         * gcc.dg/tree-ssa/vrp02.c: Likewise.
3708         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
3710 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
3711         Richard Biener  <rguenther@suse.de>
3713         * bitmap.h: Update data structure documentation, including a
3714         description of bitmap views as either linked-lists or splay trees.
3715         (struct bitmap_element_def): Update comments for splay tree bitmaps.
3716         (struct bitmap_head_def): Likewise.
3717         (bitmap_list_view, bitmap_tree_view): New prototypes.
3718         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
3719         tree_form fields.
3720         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
3721         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
3722         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
3723         released bitmap element here.
3724         (bitmap_element_free): Remove.
3725         (bitmap_elt_clear_from): Work on splay tree bitmaps.
3726         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
3727         this function similar ones such that linked-list bitmap implementation
3728         functions are grouped.
3729         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
3730         and moved for grouping.
3731         (bitmap_list_insert_element_after): Renamed from
3732         bitmap_elt_insert_after, and moved for grouping.
3733         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
3734         (bitmap_tree_link_left, bitmap_tree_link_right,
3735         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
3736         bitmap_tree_link_element, bitmap_tree_unlink_element,
3737         bitmap_tree_find_element): New functions for splay-tree bitmap
3738         implementation.
3739         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
3740         Renamed and moved, see above entries.
3741         (bitmap_tree_listify_from): New function to convert part of a splay
3742         tree bitmap to a linked-list bitmap.
3743         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
3744         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
3745         (bitmap_find_bit): Remove.
3746         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
3747         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
3748         Handle splay tree bitmaps.
3749         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
3750         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
3751         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
3752         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
3753         bitmap_intersect_compl_p, bitmap_ior_and_compl,
3754         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
3755         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
3756         corresponding changes to use linked-list specific bitmap_element
3757         manipulation functions as applicable for efficiency.
3758         (bitmap_tree_to_vec): New function.
3759         (debug_bitmap_elt_file): New function split out from ...
3760         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
3761         (bitmap_print): Likewise.
3763         PR tree-optimization/63155
3764         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
3765         SSA edge worklists.
3766         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
3767         in tree-view.
3769 2018-10-22  Martin Liska  <mliska@suse.cz>
3771         PR tree-optimization/87686
3772         Revert
3773         2018-08-29  Martin Liska  <mliska@suse.cz>
3775         * tree-switch-conversion.c (switch_conversion::expand):
3776         Strenghten assumption about gswitch statements.
3778 2018-10-22  Martin Liska  <mliska@suse.cz>
3780         * ipa-icf.c (sem_item::compare_attributes): Remove.
3781         (sem_item::compare_referenced_symbol_properties): Use
3782         attribute_list_equal instead.
3783         (sem_function::equals_wpa): Likewise.
3784         * ipa-icf.h: Remove compare_attributes.
3786 2018-10-22  Richard Biener  <rguenther@suse.de>
3788         PR middle-end/87682
3789         * mem-stats.h (mem_usage::operator==): Fix pasto.
3791 2018-10-22  Richard Biener  <rguenther@suse.de>
3793         PR tree-optimization/87640
3794         * tree-vrp.c (set_value_range_with_overflow): Decompose
3795         incomplete result.
3796         (extract_range_from_binary_expr_1): Adjust.
3798 2018-10-22  Martin Jambor  <mjambor@suse.cz>
3800         * tree-eh.h (stmt_could_throw_p): Add function parameter.
3801         (stmt_can_throw_external): Likewise.
3802         (stmt_can_throw_internal): Likewise.
3803         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
3804         (lower_eh_constructs_2): Likewise.
3805         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
3806         (stmt_can_throw_external): Likewise.
3807         (stmt_can_throw_internal): Likewise.
3808         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
3809         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
3810         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
3811         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
3812         (pass_lower_eh_dispatch::execute): Pass cfun to
3813         stmt_can_throw_external.
3814         (cleanup_empty_eh): Likewise.
3815         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
3816         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
3817         stmt_can_throw_external instead of pushing it to cfun.
3818         (symbol_table::create_edge): Likewise.
3819         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
3820         stmt_can_throw_internal.
3821         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
3822         to stmt_could_throw_p.
3823         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
3824         stmt_can_throw_internal.
3825         (pass_store_merging::execute): Likewise.
3826         * gimple-ssa-strength-reduction.c
3827         (find_candidates_dom_walker::before_dom_children): Pass cfun to
3828         stmt_could_throw_p.
3829         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
3830         stmt_can_throw_internal.
3831         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
3832         to stmt_can_throw_external.
3833         (check_stmt): Pass cfun to stmt_could_throw_p.
3834         (check_stmt): Pass cfun to stmt_can_throw_external.
3835         (pass_nothrow::execute): Likewise.
3836         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
3837         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
3838         stmt_can_throw_internal.
3839         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
3840         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
3841         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
3842         * tree-complex.c (expand_complex_libcall): Pass cfun to
3843         stmt_could_throw_p and to stmt_can_throw_internal.
3844         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
3845         * tree-inline.c (copy_edges_for_bb): Likewise.
3846         (maybe_move_debug_stmts_to_successors): Likewise.
3847         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
3848         stmt_could_throw_p.
3849         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
3850         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
3851         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
3852         stmt_can_throw_internal.
3853         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
3854         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
3855         stmt_could_throw_p.
3856         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
3857         stmt_can_throw_internal.
3858         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
3859         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
3860         stmt_could_throw_p.
3861         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
3862         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
3863         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3864         (convert_mult_to_fma_1): Likewise.
3865         (convert_to_divmod): Likewise.
3866         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
3867         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
3868         * tree-ssa-propagate.c
3869         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
3870         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
3871         (maybe_optimize_range_tests): Likewise.
3872         (linearize_expr_tree): Likewise.
3873         (reassociate_bb): Likewise.
3874         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3875         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
3876         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
3877         (handle_char_store): Likewise.
3878         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
3879         stmt_can_throw_internal.
3880         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
3881         stmt_could_throw_p.
3882         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3883         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
3884         (vectorizable_simd_clone_call): Likewise.
3885         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
3886         (gimple_stringop_fixed_value): Likewise.
3888 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
3890         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
3891         literal pool references.
3892         (s390_check_qrst_address): Adapt to the new behavior of
3893         s390_loadrelative_operand_p ().
3895 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3897         PR target/72782
3898         * config/i386/sse.md (*andnot<mode>3_bcst): New.
3900 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3902         PR target/72782
3903         * config/i386/sse.md (*<code><mode>3_bcst): New.
3905 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3907         PR target/72782
3908         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
3909         V4DI, V16SI and V8DI.
3910         (*sub<mode>3<mask_name>_bcst): New.
3911         (*add<mode>3<mask_name>_bcst): Likewise.
3913 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
3914             Jinsong Ji  <jji@us.ibm.com>
3916         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
3917         __m64 with __vector unsigned long long for compatibility.
3918         (_mm_movemask_epi8): Likewise.
3919         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
3920         (_mm_cvttps_pi32): Likewise.
3921         (_mm_cvtpi32_ps): Likewise.
3922         (_mm_cvtps_pi16): Likewise.
3923         (_mm_loadh_pi): Likewise.
3924         (_mm_storeh_pi): Likewise.
3925         (_mm_movehl_ps): Likewise.
3926         (_mm_movelh_ps): Likewise.
3927         (_mm_loadl_pi): Likewise.
3928         (_mm_storel_pi): Likewise.
3929         (_mm_movemask_ps): Likewise.
3930         (_mm_shuffle_pi16): Likewise.
3932 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3934         PR target/72782
3935         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
3936         __builtin_ia32_vfnmsubpd512_mask.
3937         (_mm512_mask_fnmsub_round_pd): Likewise.
3938         (_mm512_fnmsub_pd): Likewise.
3939         (_mm512_mask_fnmsub_pd): Likewise.
3940         (_mm512_maskz_fnmsub_round_pd): Use
3941         __builtin_ia32_vfnmsubpd512_maskz.
3942         (_mm512_maskz_fnmsub_pd): Likewise.
3943         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
3944         (_mm512_mask_fnmsub_round_ps): Likewise.
3945         (_mm512_fnmsub_ps): Likewise.
3946         (_mm512_mask_fnmsub_ps): Likewise.
3947         (_mm512_maskz_fnmsub_round_ps): Use
3948         __builtin_ia32_vfnmsubps512_maskz.
3949         (_mm512_maskz_fnmsub_ps): Likewise.
3950         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
3951         __builtin_ia32_vfnmsubpd256_mask.
3952         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
3953         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3954         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
3955         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3956         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3957         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
3958         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
3959         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
3960         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
3961         __builtin_ia32_vfnmsubpd.
3962         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
3963         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
3964         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
3965         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
3966         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
3967         * config/i386/i386-builtin.def: Add
3968         __builtin_ia32_vfnmsubpd256_mask,
3969         __builtin_ia32_vfnmsubpd256_maskz,
3970         __builtin_ia32_vfnmsubpd128_mask,
3971         __builtin_ia32_vfnmsubpd128_maskz,
3972         __builtin_ia32_vfnmsubps256_mask,
3973         __builtin_ia32_vfnmsubps256_maskz,
3974         __builtin_ia32_vfnmsubps128_mask,
3975         __builtin_ia32_vfnmsubps128_maskz,
3976         __builtin_ia32_vfnmsubpd512_mask,
3977         __builtin_ia32_vfnmsubpd512_maskz,
3978         __builtin_ia32_vfnmsubps512_mask,
3979         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
3980         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
3981         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
3982         __builtin_ia32_vfnmsubpd256.
3983         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
3984         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
3985         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
3986         Likewise.
3987         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
3988         Likewise.
3989         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
3990         Likewise.
3991         (fmai_vmfnmsub_<mode><round_name>): Likewise.
3993 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3995         PR target/72782
3996         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
3997         __builtin_ia32_vfnmaddpd512_mask.
3998         (_mm512_mask_fnmadd_round_pd): Likewise.
3999         (_mm512_fnmadd_pd): Likewise.
4000         (_mm512_mask_fnmadd_pd): Likewise.
4001         (_mm512_maskz_fnmadd_round_pd): Use
4002         __builtin_ia32_vfnmaddpd512_maskz.
4003         (_mm512_maskz_fnmadd_pd): Likewise.
4004         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
4005         (_mm512_mask_fnmadd_round_ps): Likewise.
4006         (_mm512_fnmadd_ps): Likewise.
4007         (_mm512_mask_fnmadd_ps): Likewise.
4008         (_mm512_maskz_fnmadd_round_ps): Use
4009         __builtin_ia32_vfnmaddps512_maskz.
4010         (_mm512_maskz_fnmadd_ps): Likewise.
4011         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
4012         __builtin_ia32_vfnmaddpd256_mask.
4013         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
4014         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
4015         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
4016         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
4017         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
4018         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
4019         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
4020         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
4021         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
4022         __builtin_ia32_vfnmaddpd.
4023         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
4024         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
4025         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
4026         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
4027         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
4028         * config/i386/i386-builtin.def: Add
4029         __builtin_ia32_vfnmaddpd256_mask,
4030         __builtin_ia32_vfnmaddpd256_maskz,
4031         __builtin_ia32_vfnmaddpd128_mask,
4032         __builtin_ia32_vfnmaddpd128_maskz,
4033         __builtin_ia32_vfnmaddps256_mask,
4034         __builtin_ia32_vfnmaddps256_maskz,
4035         __builtin_ia32_vfnmaddps128_mask,
4036         __builtin_ia32_vfnmaddps128_maskz,
4037         __builtin_ia32_vfnmaddpd512_mask,
4038         __builtin_ia32_vfnmaddpd512_maskz,
4039         __builtin_ia32_vfnmaddps512_mask,
4040         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
4041         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
4042         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
4043         __builtin_ia32_vfnmaddpd256.
4044         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
4045         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
4046         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
4047         Likewise.
4048         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
4049         Likewise.
4050         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
4051         Likewise.
4052         (fmai_vmfnmadd_<mode><round_name>): Likewise.
4054 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4056         PR target/72782
4057         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
4058         __builtin_ia32_vfmsubpd512_mask.
4059         (_mm512_mask_fmsub_round_pd): Likewise.
4060         (_mm512_fmsub_pd): Likewise.
4061         (_mm512_mask_fmsub_pd): Likewise.
4062         (_mm512_maskz_fmsub_round_pd): Use
4063         __builtin_ia32_vfmsubpd512_maskz.
4064         (_mm512_maskz_fmsub_pd): Likewise.
4065         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
4066         (_mm512_mask_fmsub_round_ps): Likewise.
4067         (_mm512_fmsub_ps): Likewise.
4068         (_mm512_mask_fmsub_ps): Likewise.
4069         (_mm512_maskz_fmsub_round_ps): Use
4070         __builtin_ia32_vfmsubps512_maskz.
4071         (_mm512_maskz_fmsub_ps): Likewise.
4072         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
4073         __builtin_ia32_vfmsubpd256_mask.
4074         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
4075         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
4076         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
4077         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
4078         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
4079         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
4080         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
4081         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
4082         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
4083         __builtin_ia32_vfmsubpd.
4084         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
4085         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
4086         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
4087         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
4088         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
4089         * config/i386/i386-builtin.def: Add
4090         __builtin_ia32_vfmsubpd256_mask,
4091         __builtin_ia32_vfmsubpd256_maskz,
4092         __builtin_ia32_vfmsubpd128_mask,
4093         __builtin_ia32_vfmsubpd128_maskz,
4094         __builtin_ia32_vfmsubps256_mask,
4095         __builtin_ia32_vfmsubps256_maskz,
4096         __builtin_ia32_vfmsubps128_mask,
4097         __builtin_ia32_vfmsubps128_maskz,
4098         __builtin_ia32_vfmsubpd512_mask,
4099         __builtin_ia32_vfmsubpd512_maskz,
4100         __builtin_ia32_vfmsubps512_mask,
4101         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
4102         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
4103         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
4104         __builtin_ia32_vfmsubpd256.
4105         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
4106         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
4107         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
4108         Likewise.
4109         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
4110         Likewise.
4111         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
4112         Likewise.
4113         (fmai_vmfmsub_<mode><round_name>): Likewise.
4115 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4117         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
4118         Remove plus.  Renamed to ...
4119         (*sub<mode>3<mask_name>_bcst): This.
4120         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
4121         (*add<mode>3<mask_name>_bcst): This.
4123 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4125         PR target/72782
4126         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
4128 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4130         PR target/87662
4131         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
4132         (_mm_or_epi32): Likewise.
4133         (_mm256_xor_epi32): Likewise.
4134         (_mm_xor_epi32): Likewise.
4135         (_mm256_or_epi64): Likewise.
4136         (_mm_or_epi64): Likewise.
4137         (_mm256_xor_epi64): Likewise.
4138         (_mm_xor_epi64): Likewise.
4140 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
4142         PR target/72782
4143         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
4145 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
4147         PR middle-end/87647
4148         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
4150 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
4152         * doc/ux.texi: Move @section directly after @node.
4154 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
4156         PR middle-end/85488
4157         PR middle-end/87649
4158         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
4159         depend closely nested inside of loop with ordered clause with
4160         a parameter.
4162 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
4164         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
4165         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
4166         * doc/ux.texi: New file.
4168 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
4170         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
4171         be the first CR field allocated.
4173 2018-10-19  Richard Biener  <rguenther@suse.de>
4175         PR target/87657
4176         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
4177         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
4179 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
4181         PR target/72782
4182         * config/i386/sse.md
4183         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
4184         (*add<mode>3<mask_name>_bcst_2): Likewise.
4186 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
4188         * config/i386/sse.md
4189         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
4190         Replace nonimmediate_operand with register_operand.
4191         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
4192         Likewise.
4193         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
4194         Likewise.
4196 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
4198         PR rtl-optimization/87596
4199         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
4200         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
4201         for instructions in FROM..TO range.
4203 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
4205         * cfgexpand.c (expand_one_var): Use specific wording in error message
4206         for non-local frame variables.
4207         * stor-layout.c (layout_decl): Do not issue a warning for them.
4209 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
4211         * haifa-sched.c (priority): Add force_recompute parameter.
4212         (apply_replacement): Call priority () with force_recompute = true.
4213         (restore_pattern): Likewise.
4215 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4217         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
4218         HOST_BITS_PER_WIDE_INT.
4219         (test_vector_ops_duplicate): Likewise.
4221 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4223         PR target/72782
4224         * config/i386/sse.md (VF_AVX512): New.
4225         (avx512bcst): Likewise.
4226         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
4227         Likewise.
4228         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
4229         Likewise.
4230         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
4231         Likewise.
4233 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
4235         * doc/invoke.texi (-dumpversion): Improve grammar.
4236         (-dumpfullversion): Make more consistent with -dumpversion.
4238 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
4240         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
4241         Set JUMP_LABEL to the jump insn.
4242         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
4243         Predict emitted jump and add label to jump insn.
4245 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
4247         PR tree-optimization/87562
4248         * input.c (get_substring_ranges_for_loc): Use
4249         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
4250         getting the linemap for the endpoint.  Verify that it's either
4251         in the same linemap as the start point's spelling location, or
4252         at least in the same file.
4254 2018-10-18  Richard Biener  <rguenther@suse.de>
4256         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
4257         feed width-specific load/store costs through ix86_vec_cost.
4258         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
4259         (k8_cost): Likewise.
4260         (bdver_cost): Likewise.
4261         (znver1_cost): Likewise.
4262         (btver1_cost): Likewise.
4263         (btver2_cost): Likewise.
4265 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4267         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
4268         to simplify subreg of vec_merge.
4270 2018-10-18  Richard Biener  <rguenther@suse.de>
4272         * config/i386/i386.c: Fix costing of vector FMA.
4274 2018-10-18  Richard Biener  <rguenther@suse.de>
4276         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
4277         and argument.
4278         (ix86_builtin_vectorization_cost): For vec_construct properly
4279         cost insertion into SSE regs.
4280         (...): Adjust calls to ix86_vec_cost.
4282 2018-10-18  Richard Biener  <rguenther@suse.de>
4284         PR middle-end/87087
4285         Revert
4286         2018-02-07  Richard Biener  <rguenther@suse.de>
4288         PR tree-optimization/84204
4289         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4290         this place.
4292 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4294         PR target/87537
4295         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
4296         of vec_duplicate.
4297         (test_vector_ops_duplicate): Add test for a scalar subreg of a
4298         VEC_MERGE of a VEC_DUPLICATE.
4300 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
4302         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
4303         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
4304         * doc/standards.texi (C Language): Document C2X.
4305         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
4306         (rl78_option_override): Handle "GNU C2X" language name.
4308 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
4310         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
4311         Document C17 as published in 2018.
4313 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
4315         PR middle-end/87623
4316         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
4317         bail out if both sides do not have the same storage order.
4319 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4321         * bitmap.c (bitmap_head::dump): New.
4322         * bitmap.h (bitmap_head): Add dump().
4323         * gimple-ssa-evrp-analyze.c
4324         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
4325         (evrp_range_analyzer::set_ssa_range_info): Same.
4326         (evrp_range_analyzer::record_ranges_from_phis): Same.
4327         (evrp_range_analyzer::record_ranges_from_stmt): Same.
4328         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
4329         * gimple-ssa-sprintf.c (get_int_range): Same.
4330         (format_integer): Same.
4331         (sprintf_dom_walker::handle_gimple_call): Same.
4332         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
4333         (ipcp_vr_lattice::top_p): Same.
4334         (ipcp_vr_lattice::bottom_p): Same.
4335         (ipcp_vr_lattice::set_to_bottom): Same.
4336         (ipa_vr_operation_and_type_effects): Same.
4337         (propagate_vr_across_jump_function): Same.
4338         (ipcp_store_vr_results): Same.
4339         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
4340         (ipa_print_node_jump_functions_for_edge): Same.
4341         (ipa_get_value_range): Same.
4342         (ipa_compute_jump_functions_for_edge): Same.
4343         (ipa_write_jump_function): Same.
4344         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
4345         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4346         Same.
4347         * vr-values.c (set_value_range_to_nonnegative): Same.
4348         (set_value_range_to_truthvalue): Same.
4349         (vr_values::get_value_range): Same.
4350         (vr_values::set_defs_to_varying): Same.
4351         (vr_values::update_value_range): Same.
4352         (symbolic_range_based_on_p): Same.
4353         (vr_values::op_with_boolean_value_range_p): Same.
4354         (vr_values::extract_range_for_var_from_comparison_expr): Same.
4355         (vr_values::extract_range_from_ssa_name): Same.
4356         (vr_values::extract_range_from_binary_expr): Same.
4357         (vr_values::extract_range_from_unary_expr): Same.
4358         (vr_values::extract_range_from_cond_expr): Same.
4359         (vr_values::extract_range_from_comparison): Same.
4360         (vr_values::check_for_binary_op_overflow): Same.
4361         (vr_values::extract_range_basic): Same.
4362         (vr_values::extract_range_from_assignment): Same.
4363         (compare_ranges): Same.
4364         (compare_range_with_value): Same.
4365         (vr_values::adjust_range_with_scev): Same.
4366         (vrp_valueize): Same.
4367         (vrp_valueize_1): Same.
4368         (vr_values::get_vr_for_comparison): Same.
4369         (vr_values::compare_name_with_value): Same.
4370         (vr_values::compare_names): Same.
4371         (vr_values::vrp_evaluate_conditional): Same.
4372         (find_case_label_ranges): Same.
4373         (vr_values::vrp_visit_switch_stmt): Same.
4374         (vr_values::extract_range_from_phi_node): Same.
4375         (vr_values::simplify_div_or_mod_using_ranges): Same.
4376         (vr_values::simplify_bit_ops_using_ranges): Same.
4377         (test_for_singularity): Same.
4378         (range_fits_type_p): Same.
4379         (vr_values::simplify_cond_using_ranges_1): Same.
4380         (vr_values::simplify_switch_using_ranges): Same.
4381         (vr_values::simplify_float_conversion_using_ranges): Same.
4382         (vr_values::two_valued_val_range_p): Same.
4383         (vr_values::add_equivalence): Move to value_range::equiv_add.
4384         * vr-values.h (vr_values::add_equivalence): Remove.
4385         (VR_INITIALIZER): Remove.
4386         * tree-vrp.c (value_range::set): New.
4387         (value_range::equiv_add): New.
4388         (value_range::value_range): New.
4389         (value_range::deep_copy): New.
4390         (value_range::check): New.
4391         (value_range::equal_p): New.
4392         (value_range::ignore_equivs_equal_p): New.
4393         (value_range::operator==): New.
4394         (value_range::operator!=): New.
4395         (value_range::symbolic_p): New.
4396         (value_range::numeric_p): New.
4397         (value_range::set_undefined): New.
4398         (value_range::set_varying): New.
4399         (value_range::may_contain_p): New.
4400         (value_range::equiv_clear): New.
4401         (value_range::singleton_p): New.
4402         (value_range::intersect): New.
4403         (value_range::dump): New.
4404         (value_range::set_and_canonicalize): New.
4405         (set_value_range): Adjust for value_range API.
4406         (set_value_range_to_undefined): Same.
4407         (set_value_range_to_varying): Same.
4408         (set_and_canonicalize_value_range): Same.
4409         (set_value_range_to_nonnull): Same.
4410         (set_value_range_to_null): Same.
4411         (range_is_null): Same.
4412         (range_is_nonnull): Same.
4413         (range_int_cst_p): Same.
4414         (range_int_cst_singleton_p): Same.
4415         (symbolic_range_p): Same.
4416         (range_includes_zero_p): Same.
4417         (value_range_constant_singleton): Same.
4418         (vrp_set_zero_nonzero_bits): Same.
4419         (ranges_from_anti_range): Same.
4420         (extract_range_into_wide_ints): Same.
4421         (extract_range_from_multiplicative_op): Same.
4422         (set_value_range_with_overflow): Same.
4423         (extract_range_from_binary_expr_1): Same.
4424         (extract_range_from_unary_expr): Same.
4425         (dump_value_range): Same.
4426         (debug_value_range): Same.
4427         (vrp_prop::check_array_ref): Same.
4428         (vrp_prop::check_mem_ref): Same.
4429         (vrp_prop::vrp_initialize): Same.
4430         (vrp_prop::visit_stmt): Same.
4431         (intersect_ranges): Same.
4432         (vrp_prop::visit_phi): Same.
4433         (vrp_prop::vrp_finalize): Same.
4434         (determine_value_range_1): Same.
4435         (determine_value_range): Same.
4436         (vrp_intersect_ranges_1): Rename to...
4437         (vrp_intersect_1): this.
4438         (vrp_intersect_ranges): Rename to...
4439         (value_range::intersect_helper): ...this.
4440         (vrp_meet_1): Rename to...
4441         (value_range::union_helper): ...this.
4442         (vrp_meet): Rename to...
4443         (value_range::union_): ...this.
4444         (copy_value_range): Remove.
4445         * tree-vrp.h (struct value_range): Rewrite into a proper class.
4446         (value_range::vrtype): New.
4447         (value_range::type): New.
4448         (value_range::equiv): New.
4449         (value_range::min): New.
4450         (value_range::max): New.
4451         (value_range::varying_p): New.
4452         (value_range::undefined_p): New.
4453         (value_range::null_p): New.
4454         (value_range::equiv_add): New.
4455         (copy_value_range): Remove.
4457 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
4459         * Makefile.in (SELFTEST_TARGETS): New.
4460         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
4461         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
4462         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
4463         c/Make-lang.in.
4464         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
4465         (selftest-c++-gdb, selftest-c++-valgrind): Move to
4466         cp/Make-lang.in.
4467         * configure: Regenerate.
4468         * configure.ac (selftest_languages): New.
4470 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4472         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
4473         overflow wraps argument.
4474         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
4475         wide_int_range_multiplicative_op.
4476         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
4477         overflow wraps argument.
4478         (wide_int_range_multiplicative_op): Same.
4479         (wide_int_range_lshift): Same.
4480         (wide_int_range_div): Same.
4481         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
4482         (wide_int_range_lshift): Same.
4483         (wide_int_range_div): Same.
4485 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4487         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
4488         use sign as argument.
4489         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
4490         wide_int_range_shift_undefined_p.
4492 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4494         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
4495         Rename to...
4496         (@despeculate_copy<ALLI_TI:mode>): ... This.
4497         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
4498         switch statement.
4500 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4502         * config.gcc: Obsolete *-*-solaris2.10*.
4503         * doc/install.texi (Specific, *-*-solaris2*): Document it.
4505 2018-10-12  Jeff Law  <law@redhat.com>
4507         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
4508         reg + sym +- const_int addressing modes.
4510 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4512         * common.opt (fdiagnostics-minimum-margin-width=): New option.
4513         * diagnostic-show-locus.c (layout::layout): Apply the minimum
4514         margin width.
4515         (layout::start_annotation_line): Only print up to 3 of the
4516         margin character, to avoid touching the left-hand side.
4517         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
4518         minimum margin width, as set by test_diagnostic_context's ctor.
4519         (selftest::test_fixit_insert_containing_newline): Likewise.
4520         (selftest::test_fixit_insert_containing_newline_2): Likewise.
4521         (selftest::test_line_numbers_multiline_range): Clear
4522         dc.min_margin_width.
4523         * diagnostic.c (diagnostic_initialize): Initialize
4524         min_margin_width.
4525         * diagnostic.h (struct diagnostic_context): Add field
4526         "min_margin_width".
4527         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
4528         * opts.c (common_handle_option): Handle
4529         OPT_fdiagnostics_minimum_margin_width_.
4530         * selftest-diagnostic.c
4531         (selftest::test_diagnostic_context::test_diagnostic_context):
4532         Initialize min_margin_width to 6.
4533         * toplev.c (general_init): Initialize global_dc->min_margin_width.
4535 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4537         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4538         Fix usage of "error_at_rich_loc" in the comment.
4540 2018-10-15  Renlin Li  <renlin.li@arm.com>
4542         PR target/87563
4543         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
4544         if-conversioned loop when it contains ifn with types not
4545         supported by backend.
4546         * internal-fn.c (expand_direct_optab_fn): Add an assert.
4547         (direct_internal_fn_supported_p): New helper function.
4548         * internal-fn.h (direct_internal_fn_supported_p): Declare.
4550 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
4552         PR target/87572
4553         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4554         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4555         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4557 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
4559         PR tree-optimization/87022
4560         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
4561         bits in dist vector rather than the first one.
4563 2018-10-15  Richard Biener  <rguenther@suse.de>
4565         PR middle-end/87610
4566         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
4567         (visit_loadstore): When a used restrict tag escaped verify that
4568         the points-to solution of "other" pointers do not include
4569         escaped.
4570         (compute_dependence_clique): If a used restrict tag escaped
4571         communicated that down to visit_loadstore.
4573 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
4575         * config/s390/s390.c (s390_expand_vec_init): Force vector element
4576         into reg if it isn't a general operand.
4578 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4580         PR target/87599
4581         * config/i386/sse.md (*vec_dupv2di): Add register source to
4582         movddup.
4584 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4586         PR target/87572
4587         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4588         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
4589         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4590         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4592 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
4594         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
4595         (notice_args_size): Set it in the current trace if no insn that can
4596         throw internally has been seen yet.
4597         (connect_traces): When connecting args_size between traces, allow the
4598         incoming values not to match if there is an insn setting it before the
4599         first insn that can throw internally; in that case, force the creation
4600         of a CFI note on this latter insn.
4602 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
4604         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
4605         the base class, because using the injected-class-name was not clearly
4606         specified until DR 176.
4608 2018-10-12  Paul Koning  <ni1d@arrl.net>
4610         * config/pdp11/pdp11.md (doloop_end): New expander.
4611         (doloop_end_insn): renamed from "doloop_end".
4612         (addqi3): New pattern.
4613         (subqi3): New pattern.
4614         * config/pdp11/predicates.md (incdec_operand): New predicate.
4616 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
4618         PR middle-end/81376
4619         * real.c (format_helper::can_represent_integral_type_p): New function
4620         * real.h (format_helper::can_represent_integral_type_p): Ditto.
4621         * match.pd: New pattern.
4623 2018-10-12  Alexandre Oliva <oliva@adacore.com>
4625         * configure.ac: Introduce --enable-large-address-aware
4626         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
4627         * doc/install.texi: Document it.
4628         * configure, config.in: Rebuilt.
4629         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
4630         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
4631         (LINK_SPEC): Insert it.
4632         * config/i386/mingw-w64.h: Likewise.
4634         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
4636 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
4638         PR rtl-optimization/87600
4639         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
4641 2018-10-12  Paul Koning  <ni1d@arrl.net>
4643         * doc/md.texi (doloop_end): Document that the pattern code may
4644         need to check operand mode.
4646 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
4648         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
4649         to zero-extend between int and floating-point registers.
4650         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
4651         ldp into floating-point registers.  Add type and arch attributes.
4652         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
4653         Use f_loads for type attribute.
4655 2018-10-11  Martin Sebor  <msebor@redhat.com>
4657         * doc/extend.texi (attribute packed): Correct typos.
4659 2018-10-11  Martin Sebor  <msebor@redhat.com>
4661         * doc/extend.texi (attribute flatten): Mention interaction with
4662         noinline.
4664 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
4666         PR target/87156
4667         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4668         Set new_decl virtual flag to zero.
4670 2018-10-11  Martin Sebor  <msebor@redhat.com>
4672         PR middle-end/87593
4673         * doc/extend.texi (attribute format_arg): Discuss using multiple
4674         attributes on a single function.
4676 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
4678         PR tree-optimization/86829
4679         * match.pd (sin (atan (x))): New simplification rules.
4680         (cos (atan (x))): Likewise.
4681         * real.c (build_sinatan_real): New function.
4682         * real.h (build_sinatan_real): Prototype.
4684 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
4686         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
4687         function.
4688         (fold_mergeeo_helper): New helper function.
4689         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
4690         intrinsics.  Correct some whitespace indentation issues.
4692 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
4694         PR target/87511
4695         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
4696         Use HOST_WIDE_INT_1U for shift.
4698 2018-10-11  Doug Rupp  <rupp@adacore.com>
4699             Olivier Hainque  <hainque@adacore.com>
4701         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
4702         Pass --relax to the linker for RTPs.
4703         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
4705 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
4706             Jan Hubicka  <jh@suse.cz>
4707             Martin Jambor  <mjambor@suse.cz>
4709         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
4710         the same elements are repeated rather than printing all of them.
4711         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
4712         "repeated" elements.
4713         * read-rtl-function.c (test_loading_repeat): New function.
4714         (read_rtl_function_c_tests): Call test_loading_repeat.
4715         * rtl-tests.c (test_dumping_repeat): New function.
4716         (rtl_tests_c_tests): Call test_dumping_repeat.
4718 2018-10-11  Richard Biener  <rguenther@suse.de>
4720         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
4721         bdver?_cost): Unify to ...
4722         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
4723         * config/i386/i386.c (processor_cost_table): Adjust.
4725 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
4727         PR middle-end/87574
4728         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
4729         the thunk when expanding to GIMPLE.
4731 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4733         * varasm.c (mergeable_string_section): Don't try to move zero-length
4734         strings to the merge section.
4736 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
4738         PR target/87573
4739         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
4741 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
4743         PR target/87550
4744         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
4745         to special_args set.
4747 2018-10-10  Richard Biener  <rguenther@suse.de>
4749         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
4750         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
4751         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
4752         and recursing and pattern terminating the recursion on SSE
4753         vector width using ix86_expand_reduc.
4754         (reduc_sminmax_scal_<mode>): Split into part reducing to half
4755         width and recursing and SSE2 vector variant doing the final
4756         reduction with ix86_expand_reduc.
4757         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
4758         with terminating the recursion at AVX level, splitting that
4759         to SSE there.
4761 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
4763         * genmatch.c (error_cb): Rename to...
4764         (diagnostic_cb): ...this, converting int params to enums.
4765         (fatal_at): Update for renaming.
4766         (warning_at): Likewise.
4767         (main): Likewise.
4768         * input.c (selftest::ebcdic_execution_charset::apply):
4769         Update for renaming of...
4770         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
4771         to...
4772         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
4773         converting level and reason to enums.
4774         (class selftest::lexer_error_sink): Rename to...
4775         (class selftest::lexer_test_options): ...this, renaming field
4776         "m_errors" to "m_diagnostics".
4777         (selftest::lexer_test_options::apply): Update for renaming of...
4778         (selftest::lexer_test_options::on_error): ...this, renaming to...
4779         (selftest::lexer_test_options::on_diagnostic): ...this
4780         converting level and reason to enums.
4781         (selftest::test_lexer_string_locations_raw_string_unterminated):
4782         Update for renamings.
4783         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
4784         "reason".
4786 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
4788         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
4789         * config/rs6000/pmmintrin.h: New file.
4791 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
4793         PR tree-optimization/86659
4794         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
4796 2018-10-09  Richard Biener  <rguenther@suse.de>
4798         PR tree-optimization/63155
4799         * tree-ssa-structalias.c: Include tree-ssa.h.
4800         (get_constraint_for_ssa_var): For undefs return nothing_id.
4801         (find_func_aliases): Cleanup PHI handling.
4803 2018-10-09  Richard Biener  <rguenther@suse.de>
4805         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
4806         replacements.
4808 2018-10-09  Martin Liska  <mliska@suse.cz>
4810         * asan.c (asan_emit_stack_protection): If a stack variable
4811         is located in a same file as current function, then emit
4812         line info into variable definition string.
4814 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4816         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
4817         information.
4819 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4821         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
4822         on the thunk.
4824 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4826         PR tree-optimization/86659
4827         * gimple-match.h (struct gimple_match_op): Add reverse field.
4828         (gimple_match_op::set_op): New overloaded method.
4829         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
4830         the REF_REVERSE_STORAGE_ORDER flag on the value.
4831         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
4832         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
4834 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
4836         PR middle-end/63155
4837         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
4838         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
4840 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
4842         PR target/87517
4843         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
4844         Defined with __builtin_ia32_vfmaddsubpd512_mask.
4846 2018-10-08  Richard Biener  <rguenther@suse.de>
4848         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
4849         cost the same as AVX128 ones.
4851 2018-10-08  Paul Koning  <ni1d@arrl.net>
4853         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
4854         (expand_block_move): New function.
4855         * config/pdp11/pdp11.c (output_block_move): Remove.
4856         (expand_block_move): New function.
4857         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
4858         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
4859         (*movmemhi1): Remove.
4861 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
4863         * config/s390/2827.md: Increase latencies for some FP instructions.
4865 2018-10-08  Richard Biener  <rguenther@suse.de>
4867         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4868         Open a dump scope.
4869         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
4870         * tree-vectorizer.h (dump_stmt_cost): Adjust.
4871         (add_stmt_cost): Dump return value of the hook.
4873 2018-10-08  Richard Biener  <rguenther@suse.de>
4875         PR tree-optimization/63155
4876         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
4877         (ssa_propagation_engine::ssa_propagate): Remove redundant
4878         bitmap bit clearing.
4880 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
4882         PR rtl-optimization/86939
4883         PR rtl-optimization/87479
4884         * ira.h (non_conflicting_reg_copy_p): New prototype.
4885         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
4886         (make_hard_regno_dead): Don't add conflicts for register
4887         ignore_reg_for_conflicts.
4888         (make_object_dead): Likewise.
4889         (non_conflicting_reg_copy_p): New function.
4890         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
4891         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
4892         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
4893         (make_hard_regno_dead): Don't add conflicts for register
4894         ignore_reg_for_conflicts.  Remove special conflict handling of
4895         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
4896         check_pic_pseudo_p and update callers.
4897         (mark_pseudo_dead): Don't add conflicts for register
4898         ignore_reg_for_conflicts.
4899         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
4901 2018-10-05  Andrew Waterman  <andrew@sifive.com>
4902             Jim Wilson  <jimw@sifive.com>
4904         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
4905         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
4906         new pattern using HONOR_SNANS that emits one extra instruction.
4908 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4910         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
4911         patterns): Merge SI and DI patterns to a GPR pattern.
4912         (unnamed define_insn and define_split for record form of that): Merge
4913         to a single define_insn_and_split pattern.
4915 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
4917         PR c++/56856
4918         * input.c (expand_location_to_spelling_point): Add param "aspect"
4919         and use rather than hardcoding LOCATION_ASPECT_CARET.
4920         (get_substring_ranges_for_loc): Handle the case of a single token
4921         within a macro expansion.
4922         * input.h (expand_location_to_spelling_point): Add "aspect" param,
4923         defaulting to LOCATION_ASPECT_CARET.
4925 2018-10-05  Paul Koning  <ni1d@arrl.net>
4927         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
4928         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
4929         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4930         (pdp11_guard_type): New function.
4932 2018-10-05  Paul Koning  <ni1d@arrl.net>
4934         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
4935         * config/pdp11/pdp11.opt (mfloat32): Remove.
4936         (mfloat64): Remove.
4937         * doc/invoke.texi (pdp11 -mfloat32): Remove:
4938         (pdp11 -mfloat64): Remove.
4940 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4942         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
4943         (*cmp<mode>_cc_i387): Ditto.
4944         (*cmpu<mode>_cc_i387): Ditto.
4945         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4946         * config/i386/i386.c (ix86_expand_fp_compare): Remove
4947         "scratch" argument.
4948         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
4949         Emit x86_sahf_1 pattern.
4950         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
4951         (ix86_expand_carry_flag_compare): Ditto.
4953 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4955         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
4956         to reg_or_0_operand.  Add "C" constraint.
4957         (*cmpxf_cc_i387): Ditto.
4958         (*cmp<mode>_i387): Change operand 2 predicate
4959         to nonimm_or_0_operand.  Add "C" constraint.
4960         (*cmp<mode>_cc_i387): Ditto.
4961         (*cmp<mode>_0_i387): Remove insn pattern.
4962         (*cmp<mode>_0_cc_i387): Ditto.
4964 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4966         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
4967         * config/i386/predicates.md (nonimm_or_0_operand): Rename
4968         from vector_move_operand.  Update all uses.
4970 2018-10-05  Martin Sebor  <msebor@redhat.com>
4972         PR tree-optimization/87490
4973         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
4974         consistently.
4976 2018-10-05  Richard Biener  <rguenther@suse.de>
4978         PR tree-optimization/63155
4979         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
4980         vertical space in dumpfiles.
4981         * tree-ssa-propagate.h
4982         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
4983         * tree-ssa-propagate.c (cfg_blocks_back): New global.
4984         (ssa_edge_worklist_back): Likewise.
4985         (curr_order): Likewise.
4986         (cfg_blocks_get): Remove abstraction.
4987         (cfg_blocks_add): Likewise.
4988         (cfg_blocks_empty_p): Likewise.
4989         (add_ssa_edge): Add to current or next worklist based on
4990         RPO index.
4991         (add_control_edge): Likewise.
4992         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
4993         into ...
4994         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
4995         iteration from CFG and SSA edge worklist so we process
4996         everything in RPO order, prioritizing forward progress
4997         over iteration.
4998         (ssa_prop_init): Allocate new worklists, do not dump
4999         immediate uses.
5000         (ssa_prop_fini): Free new worklists.
5002 2018-10-05  Richard Biener  <rguenther@suse.de>
5004         * tree-core.h (tree_block::abstract_flag): Remove.
5005         (tree_block::block_num): Make full 32bits.
5006         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
5007         * tree.h (BLOCK_ABSTRACT): Remove.
5008         * dwarf2out.c (gen_lexical_block_die): Remove dead code
5009         resulting from BLOCK_ABSTRACT being always false.
5010         (gen_inlined_subroutine_die): Likewise.
5011         (gen_block_die): Likewise.
5012         * tree.c (block_ultimate_origin): Likewise.
5013         * tree-pretty-print.c (dump_block_node): Remove code dealing
5014         with BLOCK_ABSTRACT.
5015         * tree-ssa-live.c (dump_scope_block): Likewise.
5016         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
5017         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
5019 2018-10-05   Richard Biener  <rguenther@suse.de>
5021         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
5022         is asked for initialize mode to the component mode of the
5023         vector type.
5025 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
5027         PR target/87522
5028         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
5029         assembler for -mavx.
5030         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
5032 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
5034         PR target/87509
5035         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
5036         RS6000_BTM_DFP.
5037         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
5038         to be DImode.  When using mffscrn, force the operand to a register.
5040 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
5042         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
5043         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
5044         X87MODEF mode iterator.
5045         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
5046         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
5047         X87MODEF mode iterator.
5049 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
5051         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
5052         -Wno-prio-ctor-dtor.
5054 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5056         * Makefile.in (OBJS): Add opt-problem.o.
5057         * dump-context.h: Include "selftest.h.
5058         (selftest::temp_dump_context): New forward decl.
5059         (class dump_context): Make friend of class
5060         selftest::temp_dump_context.
5061         (dump_context::dump_loc_immediate): New decl.
5062         (class dump_pretty_printer): Move here from dumpfile.c.
5063         (class temp_dump_context): Move to namespace selftest.
5064         (temp_dump_context::temp_dump_context): Add param
5065         "forcibly_enable_dumping".
5066         (selftest::verify_dumped_text):
5067         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
5068         (selftest::verify_item):
5069         (ASSERT_IS_TEXT): Move here from dumpfile.c.
5070         (ASSERT_IS_TREE): Likewise.
5071         (ASSERT_IS_GIMPLE): Likewise.
5072         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
5073         to...
5074         (dump_context::dump_loc_immediate): ...this new function.
5075         (class dump_pretty_printer): Move to dump-context.h.
5076         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
5077         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
5078         (temp_dump_context::temp_dump_context): Move to "selftest"
5079         namespace.  Add param "forcibly_enable_dumping", and use it to
5080         conditionalize the use of m_pp;
5081         (selftest::verify_dumped_text): Make non-static.
5082         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
5083         (selftest::verify_item): Make non-static.
5084         (ASSERT_IS_TEXT): Move to dump-context.h.
5085         (ASSERT_IS_TREE): Likewise.
5086         (ASSERT_IS_GIMPLE): Likewise.
5087         (selftest::test_capture_of_dump_calls): Pass "true" for new
5088         param of temp_dump_context.
5089         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
5090         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
5091         TDF_COMPARE_DEBUG.
5092         * opt-problem.cc: New file.
5093         * opt-problem.h: New file.
5094         * optinfo-emit-json.cc
5095         (selftest::test_building_json_from_dump_calls): Pass "true" for
5096         new param of temp_dump_context.
5097         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
5098         (optinfo::emit_for_opt_problem): New function.
5099         (optinfo::emit): Clarity which emit_item is used.
5100         * optinfo.h (optinfo::get_dump_location): New accessor.
5101         (optinfo::emit_for_opt_problem): New decl.
5102         (optinfo::emit): Make const.
5103         * selftest-run-tests.c (selftest::run_tests): Call
5104         selftest::opt_problem_cc_tests.
5105         * selftest.h (selftest::opt_problem_cc_tests): New decl.
5106         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
5107         bool to opt_result, converting fprintf messages to
5108         opt_result::failure_at calls.  Add "stmt" param for use by the
5109         failure_at calls.
5110         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
5111         (runtime_alias_check_p): Convert return type from bool to
5112         opt_result, converting dump_printf calls to
5113         opt_result::failure_at, using the statement DDR_A for their
5114         location.
5115         (find_data_references_in_stmt): Convert return type from bool to
5116         opt_result, converting "return false" to opt_result::failure_at
5117         with a new message.
5118         * tree-data-ref.h: Include "opt-problem.h".
5119         (dr_analyze_innermost): Convert return type from bool to opt_result,
5120         and add a const gimple * param.
5121         (find_data_references_in_stmt): Convert return type from bool to
5122         opt_result.
5123         (runtime_alias_check_p): Likewise.
5124         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
5125         dr_analyze_innermost.
5126         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
5127         Convert return type from bool to opt_result, adding a message for
5128         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
5129         (vect_analyze_data_ref_dependence): Convert return type from bool
5130         to opt_result.  Change sense of return type from "false"
5131         effectively meaning "no problems" to "false" meaning a problem,
5132         so that "return false" becomes "return opt_result::success".
5133         Convert "return true" calls to opt_result::failure_at, using
5134         the location of statement A rather than vect_location.
5135         (vect_analyze_data_ref_dependences): Convert return type from bool
5136         to opt_result.
5137         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
5138         calls to opt_result::failure_at, using the stmt location rather
5139         than vect_location.
5140         (vect_verify_datarefs_alignment): Convert return type from bool
5141         to opt_result.
5142         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
5143         into multiple more-tightly-scoped copies.
5144         (vect_analyze_data_refs_alignment): Convert return type from bool
5145         to opt_result.
5146         (vect_analyze_data_ref_accesses): Likewise, converting a
5147         "return false" to a "return opt_result::failure_at", adding a
5148         new message.
5149         (vect_prune_runtime_alias_test_list): Convert return type from
5150         bool to opt_result, converting dump_printf_loc to
5151         opt_result::failure_at.  Add a %G to show the pertinent statement,
5152         and use the stmt's location rather than vect_location.
5153         (vect_find_stmt_data_reference): Convert return type from
5154         bool to opt_result, converting dump_printf_loc to
5155         opt_result::failure_at, using stmt's location.
5156         (vect_analyze_data_refs):  Convert return type from bool to
5157         opt_result.  Convert "return false" to "return
5158         opt_result::failure_at", adding messages as needed.
5159         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
5160         type from bool to opt_result.
5161         (vect_determine_vf_for_stmt): Likewise.
5162         (vect_determine_vectorization_factor): Likewise, converting
5163         dump_printf_loc to opt_result::failure_at, using location of phi
5164         rather than vect_location.
5165         (vect_analyze_loop_form_1): Convert return type from bool to
5166         opt_result, converting dump_printf_loc calls, retaining the use of
5167         vect_location.
5168         (vect_analyze_loop_form): Convert return type from loop_vec_info
5169         to opt_loop_vec_info.
5170         (vect_analyze_loop_operations): Convert return type from bool to
5171         opt_result, converting dump_printf_loc calls, using the location
5172         of phi/stmt rather than vect_location where available.  Convert
5173         various "return false" to "return opt_result::failure_at" with
5174         "unsupported phi" messages.
5175         (vect_get_datarefs_in_loop): Convert return type from bool to
5176         opt_result.  Add a message for the
5177         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
5178         (vect_analyze_loop_2): Convert return type from bool to
5179         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
5180         each "goto again;", adding new messages where needed.
5181         Add "unsupported grouped {store|load}" messages.
5182         (vect_analyze_loop): Convert return type from loop_vec_info to
5183         opt_loop_vec_info.
5184         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
5185         bool to opt_result.
5186         * tree-vect-stmts.c (process_use): Likewise, converting
5187         dump_printf_loc call and using stmt location, rather than
5188         vect_location.
5189         (vect_mark_stmts_to_be_vectorized): Likeise.
5190         (vect_analyze_stmt): Likewise, adding a %G.
5191         (vect_get_vector_types_for_stmt): Convert return type from bool to
5192         opt_result, converting dump_printf_loc calls and using stmt
5193         location, rather than vect_location.
5194         (vect_get_mask_type_for_stmt): Convert return type from tree to
5195         opt_tree, converting dump_printf_loc calls and using stmt location.
5196         * tree-vectorizer.c: Include "opt-problem.h.
5197         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
5198         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
5199         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
5200         enabled, use it to report at the top level "couldn't vectorize
5201         loop" followed by the problem.
5202         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
5203         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
5204         to opt_result.
5205         (vect_analyze_stmt): Likewise.
5206         (vect_get_vector_types_for_stmt): Likewise.
5207         (tree vect_get_mask_type_for_stmt): Likewise.
5208         (vect_analyze_data_ref_dependences): Likewise.
5209         (vect_enhance_data_refs_alignment): Likewise.
5210         (vect_analyze_data_refs_alignment): Likewise.
5211         (vect_verify_datarefs_alignment): Likewise.
5212         (vect_analyze_data_ref_accesses): Likewise.
5213         (vect_prune_runtime_alias_test_list): Likewise.
5214         (vect_find_stmt_data_reference): Likewise.
5215         (vect_analyze_data_refs): Likewise.
5216         (vect_analyze_loop): Convert return type from loop_vec_info to
5217         opt_loop_vec_info.
5218         (vect_analyze_loop_form): Likewise.
5219         (vect_analyze_slp): Convert return type from bool to opt_result.
5221 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5223         * doc/invoke.texi (-fopt-info): Document new "internals"
5224         sub-option.
5225         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
5226         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
5227         MSG_ALL_KINDS.
5228         (optinfo_verbosity_options): Add "internals".
5229         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
5230         (dump_context::apply_dump_filter_p): New member function.
5231         (dump_context::dump_loc): Use apply_dump_filter_p rather than
5232         explicitly masking the dump_kind.
5233         (dump_context::begin_scope): Increment the scope depth first.  Use
5234         apply_dump_filter_p rather than explicitly masking the dump_kind.
5235         (dump_context::emit_item): Use apply_dump_filter_p rather than
5236         explicitly masking the dump_kind.
5237         (dump_dec): Likewise.
5238         (dump_hex): Likewise.
5239         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
5240         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
5241         (opt_info_switch_p): Update handling of default
5242         MSG_OPTIMIZED_LOCATIONS to cope with default of
5243         MSG_PRIORITY_USER_FACING.
5244         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
5245         masking the dump_kind.
5246         (selftest::test_capture_of_dump_calls): Update test_dump_context
5247         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
5248         than MSG_ALL.  Generalize scope test to be run at all four
5249         combinations of with/without MSG_PRIORITY_USER_FACING and
5250         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
5251         for each of the two values.
5252         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
5253         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
5254         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
5255         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
5256         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
5257         with MSG_PRIORITY_*.
5258         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
5259         dump messages as MSG_PRIORITY_USER_FACING.
5260         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
5261         about the interaction with MSG_PRIORITY_*.
5263 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5265         * varasm.c (output_constant): Add new parameter merge_strings.
5266         Make strings properly zero terminated in merge string sections.
5267         (mergeable_string_section): Don't fail if the last char is non-zero.
5268         (assemble_variable_contents): Handle merge string sections.
5269         (assemble_variable): Likewise.
5270         (assemble_constant_contents): Likewise.
5271         (output_constant_def_contents): Likewise.
5272         (output_constructor_array_range,
5273         output_constructor_regular_field): Adjust call to output_constant.
5274         (output_object_block): Adjust call to assemble_constant_contents
5275         and assemble_variable_contents.
5277 2018-10-04  Martin Liska  <mliska@suse.cz>
5279         PR c/87483
5280         * cgraphunit.c (process_function_and_variable_attributes):
5281         Warn about a function with alias attribute and a body.
5283 2018-10-04  Martin Liska  <mliska@suse.cz>
5285         PR ipa/82625
5286         * multiple_target.c (redirect_to_specific_clone): New function.
5287         (ipa_target_clone): Use it.
5288         * tree-inline.c: Fix comment.
5290 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5292         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
5293         fields.
5294         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
5295         (gcc::dump_manager::register_pass): New member function, adapted
5296         from loop body in gcc::pass_manager::register_pass, adding a
5297         call to update_dfi_for_opt_info.
5298         (gcc::dump_manager::opt_info_enable_passes): Store the
5299         -fopt-info options into the new fields.  Move the loop
5300         bodies into...
5301         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
5302         function.
5303         * dumpfile.h (struct opt_pass): New forward decl.
5304         (gcc::dump_manager::register_pass): New decl.
5305         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
5306         (class gcc::dump_manager): Add fields "m_optgroup_flags",
5307         "m_optinfo_flags", and "m_optinfo_filename".
5308         * passes.c (gcc::pass_manager::register_pass): Move all of the
5309         dump-handling code to gcc::dump_manager::register_pass.
5311 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
5313         PR rtl-optimization/87466
5314         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
5315         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
5316         * doc/tm.texi: Regenerate.
5317         * ira-lives.c (process_bb_node_lives): Use the new target hook.
5318         * lra-lives.c (process_bb_lives): Likewise.
5319         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
5320         Define.
5322 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
5324         * params.c (add_params): Fix initialization.
5326 2018-10-04  Martin Liska  <mliska@suse.cz>
5328         PR gcov-profile/84107
5329         * tree-profile.c (init_ic_make_global_vars):
5330         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
5331         Come up with new ic_tuple* variables.  Emit
5332         __gcov_indirect_call{,_topn} variables.
5333         (gimple_gen_ic_profiler): Access the variable
5334         and emit gimple.
5335         (gimple_gen_ic_func_profiler): Access
5336         __gcov_indirect_call.callee field.
5337         (gimple_init_gcov_profiler): Use ptr_type_node.
5338         * value-prof.c (gimple_ic): Use ptr_type_node.
5340 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5342         PR tree-optimization/85787
5343         * ipa-pure-const.c (malloc_candidate_p_1): Move most of
5344         malloc_candidate_p into this function and add support for
5345         detecting multiple phis.
5346         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
5348 2018-10-04  Martin Liska  <mliska@suse.cz>
5350         PR ipa/87491
5351         * ipa-inline.c (inline_to_all_callers_1):
5352         Call ultimate_alias_target for node being inlined.
5354 2018-10-03  Jeff Law  <law@redhat.com>
5356         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
5357         target's wchar_t.
5358         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
5359         * tree.h (get_typenode_from_name): Prototype.
5361 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
5363         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
5364         Change operand 2 predicate to nonimmediate_operand.
5365         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
5367 2018-10-03  Martin Sebor  <msebor@redhat.com>
5368             Jeff Law  <law@redhat.com>
5370         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
5371         initialize it.
5372         (get_string_length): Detect unterminated arrays.
5373         (format_string): Same.
5374         (format_directive): Warn about unterminated arrays.
5375         (handle_gimple_call): Mark statements with no_warning as needed.
5377 2018-10-03  Jim Wilson  <jimw@sifive.com>
5379         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
5380         also define __riscv_abi_rve.  Delete trailing white space.
5382 2018-10-03  Paul Koning  <ni1d@arrl.net>
5384         Enable LRA register allocator for PDP11.
5385         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
5386         (R): Likewise.
5387         (D): Likewise.
5388         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
5389         * config/pdp11/pdp11.opt (-mlra): New option.
5390         * doc/invoke.texi (PDP-11 Options): Document -mlra.
5392 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
5394         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
5395         (*<absneg:code>extend<mode>xf2): Ditto.
5397 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
5399         PR tree-optimization/87415
5400         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
5401         precision fields.
5403 2018-10-02  Jeff Law  <law@redhat.com>
5405         * gimple-fold.c (get_range_strlen): Only set *nonstr when
5406         an unterminated string is discovered.  Bubble up range
5407         even for unterminated strings.
5408         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
5409         indicates the string was not terminated via NONSTR.
5411 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
5413         * tree-vrp.c (extract_range_from_unary_expr): Special case all
5414         pointer conversions.
5415         Do not do anything special for anti-ranges.
5417 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
5419         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
5420         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
5422 2018-10-03  Martin Liska  <mliska@suse.cz>
5424         PR gcov-profile/86109
5425         * coverage.c (coverage_begin_function): Do not
5426         mark lambdas as artificial.
5427         * tree-core.h (struct GTY): Remove tm_clone_flag
5428         and introduce new lambda_function.
5429         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
5431 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
5433         PR target/87474
5434         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
5435         P8_VECTOR and VSX are enabled.
5437 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
5439         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
5440         0x3907 as CPU model number.
5442 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
5444         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
5445         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
5446         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
5447         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
5448         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
5449         * config/s390/s390.md: Likewise. Rename also the cpu attribute
5450         value from arch12 to z14.
5452 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
5454         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
5455         (isinfxf2): Ditto.
5456         (isinf<mode>2): Ditto.
5458 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
5460         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
5461         before emitting fxam.  Perform calculations in XFmode.
5463 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5465         * match.pd (((X /[ex] A) +- B) * A): New transformation.
5467 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5469         PR middle-end/87319
5470         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
5471         * tree.c (signed_or_unsigned_type_for): Handle complex.
5473 2018-10-02  Jeff Law  <law@redhat.com>
5475         * gimple-fold.c (get_range_strlen): Remove dead code.
5477 2018-10-02  Martin Sebor  <msebor@redhat.com>
5478             Jeff Law  <law@redhat.com>
5480         * builtins.c (unterminated_array): Add new arguments.
5481         If argument is not terminated, bubble up size and exact
5482         state to callers.
5483         (expand_builtin_strnlen): Detect, avoid expanding
5484         and diagnose unterminated arrays.
5485         (c_strlen): Fill in offset of start of unterminated strings.
5486         * builtins.h (unterminated_array): Update prototype.
5488 2018-10-02  Richard Biener  <rguenther@suse.de>
5490         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
5491         of haddv4df, first reduce to SSE width and exploit the fact
5492         that we only need element zero with the reduction result.
5493         (reduc_plus_scal_v2df): Likewise.
5495 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5497         * dojump.h (do_jump): Delete.
5498         (do_jump_1): Likewise.
5499         (split_comparison): Move around.
5500         * dojump.c (do_jump): Make static.
5501         (do_jump_1): Likewise.
5502         (jumpifnot): Move around.
5503         (jumpifnot_1): Likewise.
5504         (jumpif): Likewise.
5505         (jumpif_1): Likewise.
5506         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
5508 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5510         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
5511         insns in the delay slot and add_insn_after for the jump insn.
5513 2018-10-02  Richard Biener  <rguenther@suse.de>
5515         * tree-inline.c (expand_call_inline): Use the location of
5516         the callee declaration for the inline-entry marker.
5517         * final.c (notice_source_line): Remove special-casing of
5518         NOTE_INSN_INLINE_ENTRY.
5520 2018-10-01  Carl Love  <cel@us.ibm.com>
5522         PR 69431
5523         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
5524         (__builtin_mtfsb0): New.
5525         (__builtin_mtfsb1): New.
5526         ( __builtin_set_fpscr_rn): New.
5527         (__builtin_set_fpscr_drn): New.
5528         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
5529         (rs6000_expand_set_fpscr_rn_builtin): Add.
5530         (rs6000_expand_set_fpscr_drn_builtin): Add.
5531         (rs6000_expand_builtin): Add case statement entries for
5532         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
5533         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
5534         RS6000_BUILTIN_MFFSL.
5535         (rs6000_init_builtins): Add ftype initialization and def_builtin
5536         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
5537         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
5538         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
5539         rs6000_mffscdrn): Add define_insn.
5540         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
5541         * doc/extend.texi: Add documentation for the builtins.
5543 2018-10-01  Richard Biener  <rguenther@suse.de>
5545         PR tree-optimization/87465
5546         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
5547         causing branch miscounts.
5549 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5551         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
5552         aarch64_option_default_param):  New.
5553         (params.h): Include.
5554         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
5555         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
5556         stack-clash protection validation code.
5558 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5560         * params.c (validate_param): New.
5561         (add_params): Use it.
5562         (set_param_value): Refactor param validation into validate_param.
5563         (diagnostic.h): Include.
5564         * diagnostic.h (diagnostic_ready_p): New.
5566 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5568         * params.c (set_param_value):
5569         Add index of parameter being validated.
5570         * common/common-target.def (option_validate_param): New.
5571         * common/common-targhooks.h (default_option_validate_param): New.
5572         * common/common-targhooks.c (default_option_validate_param): New.
5573         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
5574         * doc/tm.texi: Regenerate.
5576 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5578         PR target/86486
5579         * config/aarch64/aarch64.c (aarch64_override_options_internal):
5580         Add validation for stack-clash parameters and set defaults.
5582 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5584         PR target/86486
5585         * configure.ac: Add stack-clash-protection-guard-size.
5586         * doc/install.texi: Document it.
5587         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
5588         * params.def: Update comment for guard-size.
5589         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
5590         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
5591         * configure: Regenerate.
5593 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5595         PR target/86486
5596         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
5597         STACK_DYNAMIC_OFFSET): New.
5598         * config/aarch64/aarch64.c (aarch64_layout_frame):
5599         Update outgoing args size.
5600         (aarch64_stack_clash_protection_alloca_probe_range,
5601         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5603 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5605         PR target/86486
5606         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
5607         probe ranges.
5608         * target.def (stack_clash_protection_alloca_probe_range): New.
5609         (stack_clash_protection_final_dynamic_probe): Remove.
5610         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
5611         (default_stack_clash_protection_final_dynamic_probe): Remove.
5612         * targhooks.c: Likewise.
5613         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5614         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
5615         * doc/tm.texi: Regenerate.
5617 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5619         PR target/86486
5620         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
5621         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
5622         aarch64_clamp_to_uimm12_shift): New.
5623         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
5624         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
5626 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5628         PR target/86486
5629         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
5631 2018-10-01  Jeff Law  <law@redhat.com>
5632             Richard Sandiford <richard.sandiford@linaro.org>
5633             Tamar Christina  <tamar.christina@arm.com>
5635         PR target/86486
5636         * config/aarch64/aarch64.md
5637         (probe_stack_range): Add k (SP) constraint.
5638         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
5639         STACK_CLASH_MAX_UNROLL_PAGES): New.
5640         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
5641         stack probes for stack clash.
5642         (aarch64_allocate_and_probe_stack_space): New.
5643         (aarch64_expand_prologue): Use it.
5644         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
5645         (aarch64_sub_sp): Add emit_move_imm optional param.
5647 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
5649         PR tree-optimization/87261
5650         * match.pd: Remove trailing whitespace.
5651         Add (x & y) | ~(x | y) -> ~(x ^ y),
5652         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
5654 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5656         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
5657         constraints.
5658         (ashlsi3_insn): Update instruction constraints.
5659         (ashrsi3_insn): Likewise.
5660         (rotrsi3): Likewise.
5661         (add_shift): Likewise.
5662         * config/arc/constraints.md (Csz): New 32 bit constraint. It
5663         avoids placing in the limm field small constants which, otherwise,
5664         could end into a small instruction.
5666 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5668         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
5669         destination register is not odd-even.
5670         (umaddsidi4_split): Likewise.
5672 2018-10-01  Richard Biener  <rguenther@suse.de>
5674         * tree-inline.c (expand_call_inline): Store origin of fn
5675         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
5676         * tree.c (block_ultimate_origin): Simplify and do some
5677         checking.
5679 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
5681         * config/i386/mmx.md (EMMS): New int iterator.
5682         (emms): New int attribute.
5683         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
5684         EMMS int iterator.  Explicitly declare clobbers.
5685         (mmx_emms): Remove expander.
5686         (mmx_femms): Ditto.
5687         * config/i386/predicates.md (emms_operation): Remove predicate.
5688         (vzeroall_pattern): New predicate.
5689         (vzeroupper_pattern): Rename from vzeroupper_operation.
5690         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
5691         vzeroupper_pattern and vzeroall_pattern predicates.
5693 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
5695         PR rtl-optimization/86939
5696         * ira-lives.c (make_hard_regno_born): Rename from this...
5697         (make_hard_regno_live): ... to this.  Remove update to conflict
5698         information.  Update function comment.
5699         (make_hard_regno_dead): Add conflict information update.  Update
5700         function comment.
5701         (make_object_born): Rename from this...
5702         (make_object_live): ... to this.  Remove update to conflict information.
5703         Update function comment.
5704         (make_object_dead):  Add conflict information update.  Update function
5705         comment.
5706         (mark_pseudo_regno_live): Call make_object_live.
5707         (mark_pseudo_regno_subword_live): Likewise.
5708         (mark_hard_reg_dead): Update function comment.
5709         (mark_hard_reg_live): Call make_hard_regno_live.
5710         (process_bb_node_lives): Likewise.
5711         * lra-lives.c (make_hard_regno_born): Rename from this...
5712         (make_hard_regno_live): ... to this.  Remove update to conflict
5713         information.  Remove now uneeded check_pic_pseudo_p argument.
5714         Update function comment.
5715         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
5716         to conflict information.  Update function comment.
5717         (mark_pseudo_live): Remove update to conflict information.  Update
5718         function comment.
5719         (mark_pseudo_dead): Add conflict information update.
5720         (mark_regno_live): Call make_hard_regno_live.
5721         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
5722         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
5724 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
5726         PR target/87370
5727         * config/i386/i386.c (construct_container): Use TImode for
5728         BLKmode values in 2 integer registers.
5730 2018-09-29  Jeff Law  <law@redhat.com>
5732         * builtins.c (unterminated_array): Pass in c_strlen_data * to
5733         c_strlen rather than just a tree *.
5734         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
5735         Update recursive calls appropriately.  If caller did not provide a
5736         suitable data pointer, create a local one.  When a non-terminated
5737         string is discovered, bubble up information about the string via the
5738         c_strlen_data object.
5739         * builtins.h (c_strlen): Update prototype.
5740         (c_strlen_data): New structure.
5741         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
5742         For a type 2 call, if c_strlen indicates a non-terminated string
5743         use the length of the non-terminated string.
5744         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
5746 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
5748         PR target/87467
5749         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
5750         __m512d type for __A argument rather than __m512.
5752 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
5754         * match.pd (simple_comparison): Don't optimize if either operand is
5755         a function pointer when target needs function pointer canonicalization.
5757 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5759         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
5760         power5 .. power9 to remove indirection.
5761         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
5762         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
5763         ASM_CPU_476_SPEC): Delete.
5764         (ASM_CPU_SPEC): Adjust.
5765         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
5766         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
5768 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5770         * config.in: Delete HAVE_AS_DCI.
5771         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
5772         * config/rs6000/rs6000.h: Ditto.
5773         * configure.ac: Delete HAVE_AS_DCI.
5774         * configure: Regenerate.
5776 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5778         * config.in (HAVE_AS_LWSYNC): Delete.
5779         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5780         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
5781         do it as a .long .
5782         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5783         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
5784         as a .long .
5785         * configure.ac: Delete HAVE_AS_LWSYNC.
5786         * configure: Regenerate.
5788 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5789             Pierre-Marie de Rodat  <derodat@adacore.com>
5791         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
5792         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
5793         (cgraph_node::create_thunk): Add indirect_offset parameter.
5794         (thunk_adjust): Likewise.
5795         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
5796         and initialize the corresponding field with it.
5797         (cgraph_node::dump): Dump indirect_offset field.
5798         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
5799         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
5800         (thunk_adjust): Add indirect_offset parameter and deal with it.
5801         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
5802         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
5803         or if the thunk is external or local.  Fix formatting.  Do not chain
5804         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
5805         if any, in the GIMPLE representation.
5806         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
5807         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
5808         (input_node): Read indirect_offset field.
5809         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
5810         call to thunk_adjust.
5811         * tree-nested.c (struct nesting_info): Add thunk_p field.
5812         (create_nesting_tree): Set it.
5813         (convert_all_function_calls): Copy static chain from targets to thunks.
5814         (finalize_nesting_tree_1): Return early for thunks.
5815         (unnest_nesting_tree_1): Do not finalize thunks.
5816         (gimplify_all_functions): Do not gimplify thunks.
5818 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5820         * opt-suggestions.c (option_proposer::build_option_suggestions):
5821         Release "option_values".
5823 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5825         * coverage.c (get_coverage_counts): Convert problem-reporting dump
5826         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
5827         * dumpfile.c (kind_as_string): New function.
5828         (dump_loc): Rather than a hardcoded prefix of "note: ", use
5829         kind_as_string to vary the prefix based on dump_kind.
5830         (selftest::test_capture_of_dump_calls): Update for above.
5832 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5834         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
5835         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
5837 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5839         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
5840         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
5841         INVALID_REGNUM instead of FPSR_REG.
5842         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
5843         * config/i386/i386.md: Update comment of FP compares.
5844         (fldenv): Do not clobber FPSR_REG.
5846 2018-09-28  Richard Biener  <rguenther@suse.de>
5848         * tree.h (BLOCK_ORIGIN): New.
5849         * omp-expand.c (grid_expand_target_grid_body): Assign
5850         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
5851         * tree-inline.c (remap_block): Likewise.
5852         * auto-profile.c (get_function_decl_from_block): Simplify
5853         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
5854         * langhooks.c (lhd_print_error_function): Likewise.
5855         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
5856         Likewise.
5857         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5858         * tree.c (block_nonartificial_location): Likewise.
5859         (block_ultimate_origin): Likewise.
5860         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
5861         no longer needed LTO case.
5863 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
5864             Jan Hubicka  <jh@suse.cz>
5865             Martin Jambor  <mjambor@suse.cz>
5867         * simplify-rtx.c (simplify_merge_mask): New function.
5868         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
5869         same masks are used in op1 or op2.
5870         (test_vec_merge): New function.
5871         (test_vector_ops): Call test_vec_merge.
5873 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5875         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
5876         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
5877         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
5878         (cypress_costs): Set it.
5879         (supersparc_costs): Likewise.
5880         (hypersparc_costs): Likewise.
5881         (leon_cost): Likewise.
5882         (leon3_costs): Likewise.
5883         (sparclet_costs): Likewise.
5884         (ultrasparc_costs): Likewise.
5885         (ultrasparc_costs): Likewise.
5886         (niagara_costs): Likewise.
5887         (niagara2_costs): Likewise.
5888         (niagara3_costs): Likewise.
5889         (niagara4_costs): Likewise.
5890         (niagara7_costs): Likewise.
5891         (m8_costs): Likewise.
5892         (TARGET_CAN_FOLLOW_JUMP): Define.
5893         (pass_work_around_errata::gate): Minor tweak.
5894         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
5895         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
5896         Automaitcally clear MASK_FSMULD mask for V7 processors.
5897         (sparc_can_follow_jump): New static function.
5898         (output_ubranch): Deal with CROSSING_JUMP_P.
5899         (sparc_use_sched_lookahead): Rewrite using switch statement.
5900         (sparc_issue_rate): Reorder.
5901         (sparc_branch_cost): New function.
5903 2018-09-27  Martin Sebor  <msebor@redhat.com>
5905         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
5906         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
5907         (int_fits_type_p): Same.
5909 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
5911         * config/i386/i386.md (FPCR_REG): Remove.
5912         (UNSPEC_FLDCW): Remove.
5913         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
5914         (x86_fldcw_1): Remove insn pattern.
5915         (fnstenv): Do not clobber FPCR_REG.
5916         (fldenv): Ditto.
5917         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
5918         (CALL_USED_REGISTERS): Ditto.
5919         (REG_ALLOC_ORDER): Ditto.
5920         (REG_CLASS_CONTENTS): Ditto.
5921         (HI_REGISTER_NAMES): Ditto.
5922         (ADDITIONAL_REGISTER_NAMES): Use defines instead
5923         of numerical constants.
5924         * config/i386/i386.c (regclass_map): Remove fpsr register.
5925         (dbx_register_map): Ditto.
5926         (dbx64_register_map): Ditto.
5927         (svr4_dbx_register_map): Ditto.
5928         (print_reg): Do not handle FPCR_REG.
5930 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
5932         PR target/87149
5933         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
5934         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
5935         Delete, always treat as true.
5936         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5937         Ditto.  Simplify remaining code.
5938         * config/powerpcspe/powerpcspe.h: Ditto.
5939         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
5940         Simplify remaining code.
5941         (rs6000_expand_builtin): Ditto.
5942         * config/rs6000/rs6000.h: Ditto.
5943         * configure.ac: Ditto.
5944         * configure: Regenerate.
5946 2018-09-27  Martin Liska  <mliska@suse.cz>
5948         * coverage.c (get_coverage_counts): Revert the formatting
5949         of missing profile opt info.
5951 2018-09-27  Richard Biener  <rguenther@suse.de>
5953         PR debug/37801
5954         PR debug/87440
5955         * dwarf2out.c (set_block_origin_self): Do not mark outermost
5956         block as we do not output that.
5957         (gen_inlined_subroutine_die): Elide the originally outermost
5958         block, matching what we do for concrete instances.
5959         (decls_for_scope): Add parameter specifying whether to recurse
5960         to subblocks.
5962 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
5963             Tom de Vries  <tom@codesourcery.com>
5965         PR 82089
5967         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
5968         STORE_FLAG_VALUE == 1.
5970 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
5972         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
5973         constant definitions.
5974         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
5975         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
5976         ("speculation_barrier"): New expander definition.
5978 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
5980         PR gcov-profile/86957
5981         * common.opt: New warning option -Wmissing-profile.
5982         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
5983         * doc/invoke.texi: Document -Wmissing-profile.
5985 2018-09-26  Jim Wilson  <jimw@sifive.com>
5987         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
5988         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
5990 2018-09-26  Martin Sebor  <msebor@redhat.com>
5992         * tree.c (zerop): Change return type to bool.
5993         (integer_zerop, integer_onep, integer_each_onep): Same.
5994         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5995         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5996         (real_onep, real_minus_onep, chain_index): Same.
5997         (print_type_hash_statistics, type_list_equal): Same.
5998         * tree.h (zerop): Same.
5999         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
6000         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
6001         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
6002         (real_onep, real_minus_onep, chain_index): Same.
6003         (print_type_hash_statistics, type_list_equal): Same.
6005 2018-09-26  Jim Wilson  <jimw@sifive.com>
6007         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
6009 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
6011         PR target/87414
6012         * config/i386/i386.c: Include debug.h and dwarf2out.h.
6013         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
6014         call.
6016 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
6018         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
6020 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
6022         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
6023         and FP_SECOND_SSE_REGS.
6024         (REG_CLASS_NAMES): Ditto.
6025         (REG_CLASS_CONTENTS): Ditto.
6026         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
6027         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
6028         (ix86_preferred_output_reload_class): Ditto.
6029         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
6030         clobber constraint to "=&f".
6031         (fix_truncdi_i387): Ditto.
6032         (lrintxfdi2): Ditto.
6033         (fistdi2_<rounding>): Ditto.
6034         (fpremxf4_i387): Change "=u" constraint to "=f".
6035         (fprem1xf4_i387): Ditto.
6036         (sincosxf3): Ditto.
6037         (fptanxf4_i387): Ditto.
6038         (fxtractxf3_i387): Ditto.
6039         (fscalexf4_i387): Ditto.
6040         (atan2xf3): Change "u" constraint to "f".
6041         (fyl2xxf3_i387): Ditto.
6042         (fyl2xp1xf3_i387): Ditto.
6044 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
6046         PR target/87439
6047         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
6048         for removed I387_MASK_PM entity.
6051 2018-09-26  Jeff Law  <law@redhat.com>
6052         Revert
6053         2018-09-26  Alexey Neyman  <stilor@att.net>
6055         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
6056         headers are no longer pulled in by <isl/val.h>.
6058 2018-09-26  Richard Biener  <rguenther@suse.de>
6060         PR debug/87443
6061         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
6062         or concrete instance DIE to the tree.  Create abstract origin
6063         attributes also for concrete instances.
6065 2018-09-26  Alexey Neyman  <stilor@att.net>
6067         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
6068         headers are no longer pulled in by <isl/val.h>.
6070 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
6072         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
6073         Use new helper functions.
6074         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
6075         Use new helper functions.
6076         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
6077         aarch_mm_needs_release): New declarations.
6078         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
6079         aarch_mm_needs_release): New.
6081 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
6083         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
6084         (arm32_output_mi_thunk): Deal with long calls.
6086 2018-09-26  Richard Biener  <rguenther@suse.de>
6088         PR debug/87428
6089         PR debug/87362
6090         * tree-inline.c (expand_call_inline): When the location
6091         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
6092         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
6093         the inserted BLOCK to make inlined_function_outer_scope_p
6094         recognize it.
6095         * dwarf2out.c (add_call_src_coords_attributes): Do not add
6096         coords for reserved locations.
6098 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6100         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
6101         (*call_indirect_nonlocal_sysv<mode>): Ditto.
6102         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
6103         (*sibcall_nonlocal_sysv<mode>): Ditto.
6104         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
6105         (<bd>_<mode>): Ditto.
6106         (<bd>tf_<mode>): Ditto.
6108 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6110         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
6111         control string as a list of templates instead of as C code.
6112         (*altivec_movti): Ditto.
6113         * config/rs6000/darwin.md (movdf_low_di): Ditto.
6115 2018-09-25  Jim Wilson  <jimw@sifive.com>
6117         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
6118         when target symbol is weak.
6120 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6122         PR c/87387
6123         * builtins.c (unterminated_array): Simplify.
6124         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
6125         where pointer arithmetic is safe.
6127 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6129         PR target/86987
6130         * config/rs6000/altivec.md (altivec_vspltb): Use
6131         const_0_to_15_operand instead of u5bit_cint_operand.
6132         (*altivec_vspltb_internal): Ditto.
6133         (altivec_vspltb_direct): Ditto.
6134         (altivec_vsplth): Use const_0_to_7_operand instead of
6135         u5bit_cint_operand.
6136         (*altivec_vsplth_internal): Ditto.
6137         (altivec_vsplth_direct): Ditto.
6138         (altivec_vspltw): Use const_0_to_3_operand instead of
6139         u5bit_cint_operand.
6140         (*altivec_vspltw_internal): Ditto.
6141         (altivec_vspltw_direct): Ditto.
6142         (altivec_vspltsf): Ditto.
6143         (*altivec_vspltsf_internal): Ditto.
6144         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
6145         various splats with the proper size immediate.  Reorder the various
6146         cases by ascending size of immediate, and put all such together.
6148 2018-09-25  Richard Biener  <rguenther@suse.de>
6150         PR debug/83941
6151         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
6152         GC-ification.
6153         (maybe_create_die_with_external_ref): Do not create
6154         DW_TAG_imported_unit here.
6155         (add_abstract_origin_attribute): Handle external BLOCK refs.
6156         (dwarf2out_abstract_function): Simplify LTO case.
6157         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
6158         rather than using maybe_create_die_with_external_ref.
6160 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
6162         PR target/71278
6163         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
6164         (frndintxf2_mask_pm_i387): Ditto.
6165         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
6166         Enable for !flag_trapping_math.
6167         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
6168         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
6169         Change operand 1 predicate to nonimmediate_operand.
6170         (attr "i387_cw"): Remove mask_pm.
6171         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
6172         (enum ix86_entity): Remove I387_MASK_PM.
6173         * config/i386/i386.c (ix86_i387_mode_needed): Do not
6174         handle I387_MASK_PM.
6175         (ix86_mode_needed): Ditto.
6176         (ix86_mode_after): Ditto.
6177         (ix86_mode_entry): Ditto.
6178         (ix86_mode_exit): Ditto.
6179         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
6181 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
6183         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
6184         to_update_switch_stmts to vNULL instead of calling create on them
6185         immediately.
6187 2018-09-25  Richard Biener  <rguenther@suse.de>
6189         PR tree-optimization/87402
6190         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
6191         (visit_phi): Re-instantiate handling of supposed to be VARYING
6192         but non-VARYING backedge value.
6194 2018-09-25  Richard Biener  <rguenther@suse.de>
6196         PR debug/83941
6197         * dwarf2out.c (struct sym_off_pair): New.
6198         (external_die_map): New global.
6199         (lookup_decl_die): When in LTO create DIEs lazily from the
6200         external_die_map.
6201         (lookup_block_die): New function, create DIEs lazily in LTO.
6202         (equate_block_to_die): New function.
6203         (dwarf2out_die_ref_for_decl): During WPA get the association
6204         from the external DIE map.
6205         (dwarf2out_register_external_die): Record mapping into the
6206         external DIE map.
6207         (maybe_create_die_with_external_ref): New function split out from
6208         DIE generation part of old dwarf2out_register_external_die.
6209         (add_abstract_origin_attribute): Do not return the DIE.  When
6210         in LTO reference externals directly.
6211         (dwarf2out_abstract_function): When in LTO ignore calls for
6212         decls with external DIEs (already present abstract instances).
6213         (gen_call_site_die): Adjust.
6214         (add_high_low_attributes): Likewise.
6215         (gen_lexical_block_die): Likewise.
6216         (gen_inlined_subroutine_die): Likewie.
6217         (gen_block_die): Likewise.
6218         (dwarf2out_inline_entry): Likewise.
6219         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
6220         DIEs.
6222 2018-09-25  Martin Liska  <mliska@suse.cz>
6224         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
6225         integers and not by a float value.
6227 2018-09-25  Martin Liska  <mliska@suse.cz>
6229         PR fortran/87394
6230         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
6231         instead of NULL.
6232         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
6233         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
6234         * godump.c (go_format_type): Remove extra parenthesis.
6236 2018-09-25  Martin Liska  <mliska@suse.cz>
6238         * alias.c (set_dest_equal_p): Remove unused function.
6239         * config/i386/i386.c (def_builtin_pure2): Likewise.
6240         * diagnostic-show-locus.c (class layout): Remove
6241         unused field.
6242         (layout::layout): Likewise here.
6243         * dump-context.h (class temp_dump_context): Likewise.
6244         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
6245         (add_AT_loclistsptr): Likewise.
6246         (add_AT_offset): Likewise.
6247         (get_AT_hi_pc): Likewise.
6248         (is_comdat_die): Likewise.
6249         (type_is_enum): Likewise.
6250         (ceiling): Likewise.
6251         (add_AT_vms_delta): Likewise.
6252         (is_class_die): Likewise.
6253         * edit-context.c (class line_event): Remove unused field.
6254         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
6255         unused function.
6256         * ipa-cp.c (ipa_get_vr_lat): Likewise.
6257         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
6258         (ok_for_base_p_nonstrict): Likewise.
6259         * tree-chrec.c (is_not_constant_evolution): Likewise.
6260         (chrec_fold_poly_cst): Likewise.
6261         * tree-if-conv.c (has_pred_critical_p): Likewise.
6262         * tree-ssa-coalesce.c (print_exprs): Likewise.
6263         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
6264         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
6265         * tree-vrp.c (value_ranges_intersect_p): Likewise.
6266         (value_range_nonnegative_p): Likewise.
6268 2018-09-25  Martin Liska  <mliska@suse.cz>
6270         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
6271         Do not handle "GNU Pascal".
6272         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
6273         Likewise.
6274         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
6275         from documentation. Likewise.
6276         * dbxout.c (dbxout_range_type): Likewise.
6277         * doc/cpp.texi: Likewise.
6278         * doc/extend.texi: Likewise.
6279         * doc/frontends.texi: Likewise.
6280         * doc/invoke.texi: Remove Pascal entry.
6281         * tree.def (CLEANUP_POINT_EXPR): Likewise.
6282         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
6284 2018-09-25  Martin Liska  <mliska@suse.cz>
6286         PR middle-end/86078
6287         * doc/invoke.texi: Document all parameters and remove default
6288         of the parameters.
6290 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
6292         PR bootstrap/87417
6293         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
6294         contains HOST_WIDE_INTs when computing its size.
6296 2018-09-24  Jim Wilson  <jimw@sifive.com>
6298         PR target/87391
6299         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
6300         not TARGET_RVE.
6301         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
6303 2018-09-24  Andrew Pinski  <apinski@marvell.com>
6305         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
6306         access prev before checking it for NULLness in the
6307         AARCH64_FUSE_CMP_BRANCH case.
6309 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
6311         PR target/82699
6312         * config/i386/i386.c (rest_of_insert_endbranch): Set
6313         endbr_queued_at_entrance to true and don't insert ENDBR if
6314         x86_function_profiler will be called.
6315         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
6316         is true.
6317         * config/i386/i386.h (machine_function): Add
6318         endbr_queued_at_entrance.
6320 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6322         * genattrtab.c (mk_attr_alt): Use alternative_mask.
6323         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
6324         types.
6325         (check_attr_test): Use alternative_mask.
6326         (get_attr_value): Likewise.
6327         (compute_alternative_mask): Use alternative_mask and XWINT.
6328         (make_alternative_compare): Use alternative_mask.
6329         (attr_alt_subset_p): Use XWINT.
6330         (attr_alt_subset_of_compl_p): Likewise.
6331         (attr_alt_intersection): Use alternative_mask and XWINT.
6332         (attr_alt_union): Likewise.
6333         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
6334         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
6335         (simplify_test_exp): Use alternative_mask and XWINT.
6336         (write_test_expr): Use alternative_mask and XWINT, adjust bit
6337         number calculation to support 64 bits.  Generate code that
6338         checks 64-bit masks.
6339         (main): Use alternative_mask.
6340         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
6342 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6344         PR target/80080
6345         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
6346         RETURN+USE when returning via %r14.
6348 2018-09-24  Martin Liska  <mliska@suse.cz>
6350         * gcov.c (output_lines): Print colorization legend
6351         for both flag_use_colors and flag_use_hotness_colors.
6352         Reword the help.
6354 2018-09-24  Martin Liska  <mliska@suse.cz>
6356         * coverage.c (get_coverage_counts): Use warning_at
6357         with current_function_decl location. Use %qD in warning
6358         message.
6360 2018-09-24  Martin Liska  <mliska@suse.cz>
6362         * memory-block.h (memory_block_pool::release): Annotate with
6363         valgrind that the memory is not accessible.
6365 2018-09-24  Martin Liska  <mliska@suse.cz>
6367         PR sanitizer/85774
6368         * asan.c: Make asan_handled_variables extern.
6369         * asan.h: Likewise.
6370         * cfgexpand.c (expand_stack_vars): Make sure
6371         a representative is unpoison if another
6372         variable in the partition is handled by
6373         use-after-scope sanitization.
6375 2018-09-24  Richard Biener  <rguenther@suse.de>
6377         PR tree-optimization/63155
6378         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
6379         the worklist when the edge of the respective argument isn't
6380         executable.
6382 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6384         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
6385         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
6386         (MASK_CLASS_P): Update for rename.
6387         (MAYBE_MASK_CLASS_P): Ditto.
6388         (REG_CLASS_NAMES): Update.
6389         (REG_CLASS_CONTENT): Update.
6390         * config/i386/i386.c (regclass_map): Update for MASK_REG
6391         and ALL_MASK_REGS rename.
6392         * config/i386/constraints.md (Yk): Update for rename.
6393         (k): Ditto.
6395 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6397         * config/i386/i386.h (enum reg_class): Remove
6398         EVEX_SSE_REGS and MOD4_SSE_REGS.
6399         (REG_CLASS_NAMES): Update.
6400         (REG_CLASS_CONTENT): Update.
6401         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
6402         registers as ALL_SSE_REGS.
6403         (ix86_additional_allocno_class_p): Remove.
6404         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
6405         (ix86_register_priority): Lower priority of EVEX SSE registers.
6406         Use IN_RANGE macro where appropriate.
6407         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
6408         AVX-5124VNNIW checks.
6409         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
6410         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
6411         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
6412         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
6413         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
6414         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
6415         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
6416         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
6417         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
6418         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
6419         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
6420         * config/i386/constraints.md (Yh): Remove.
6422 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6424         * config/i386/i386.c (regclass_map): Declare integer REX registers
6425         as GENERAL_REGS.
6427 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
6429         * doc/service.texi (Service): Switch the fsf.org link to https.
6431 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
6433         PR target/86798
6434         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6435         Define to speculation_safe_value_not_needed.
6437 2018-09-21  Florian Weimer  <fweimer@redhat.com>
6439         PR middle-end/81035
6440         * doc/extend.texi (Common Function Attributes): Mention that
6441         noreturn suppresses tail call optimization.
6443 2018-09-21  Jeff Law  <law@redhat.com>
6445         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
6446         vr_values::cleanup_edges_and_switches.
6447         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
6448         vr_values class.
6449         (identify_jump_threads): Remove EDGE_IGNORE handling.
6450         (execute_vrp): Move handling of to_remove_edges and
6451         to_update_switch_stmts into vr_values class member functions.
6452         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
6453         (to_update_switch_stmts): Likewise.
6454         * vr-values.c: Include cfghooks.h.
6455         (vr_values::vr_values): Initialize to_remove_edges and
6456         to_update_switch_stmts.
6457         (vr_values::~vr_values): Verify to_remove_edges and
6458         to_update_switch_stmts are empty.
6459         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
6460         (vr_values::cleanup_edges_and_switches): New member function.
6461         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
6462         function.  Add new data members.
6464 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
6466         PR tree-optimization/87309
6467         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
6468         calls with pflags and alt_flags.
6469         (selftest::test_capture_of_dump_calls): Add test of interaction of
6470         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
6472 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6474         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
6476 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6478         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
6480 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6482         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
6483         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
6484         (VXWORKS_LIBS_RTP): Minor reordering.
6486 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6488         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
6489         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
6491 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6493         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
6494         (PTRDIFF_TYPE): Likewise.
6496 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6498         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
6499         triplet, similar to support for VxWorks7.
6500         * config/vxworks-dummy.h: Provide a default definition
6501         of TARGET_VXWORKS64 to 0.
6503 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6505         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
6506         * config/vxworks-dummy.h: here.
6508 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6510         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
6512 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
6513             Bo Zhou  <zbo.zhou@hisilicon.com>
6515         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
6516         * config/aarch64/aarch64-tune.md: Regenerated.
6517         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
6518         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
6519         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
6521 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
6522             Julian Brown  <julian@codesourcery.com>
6524         * builtins.c (get_builtin_sync_mem): Handle address spaces.
6526 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6528         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
6529         if the call takes a static chain.
6531 2018-09-21  Martin Liska  <mliska@suse.cz>
6533         * auto-profile.c (autofdo_source_profile::read): Do not
6534         set sum_all.
6535         (read_profile): Do not add working sets.
6536         (read_autofdo_file): Remove sum_all.
6537         (afdo_callsite_hot_enough_for_early_inline): Remove const
6538         qualifier.
6539         * coverage.c (struct counts_entry): Remove gcov_summary.
6540         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
6541         do not support GCOV_TAG_PROGRAM_SUMMARY.
6542         (get_coverage_counts): Remove summary and expected
6543         arguments.
6544         * coverage.h (get_coverage_counts): Likewise.
6545         * doc/gcov-dump.texi: Remove -w option.
6546         * gcov-dump.c (dump_working_sets): Remove.
6547         (main): Do not support '-w' option.
6548         (print_usage): Likewise.
6549         (tag_summary): Likewise.
6550         * gcov-io.c (gcov_write_summary): Do not dump
6551         histogram.
6552         (gcov_read_summary): Likewise.
6553         (gcov_histo_index): Remove.
6554         (gcov_histogram_merge): Likewise.
6555         (compute_working_sets): Likewise.
6556         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
6557         it not obsolete.
6558         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
6559         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
6560         (GCOV_HISTOGRAM_SIZE): Remove.
6561         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
6562         (struct gcov_summary): Simplify rapidly just
6563         to runs and sum_max fields.
6564         (gcov_histo_index): Remove.
6565         (NUM_GCOV_WORKING_SETS): Likewise.
6566         (compute_working_sets): Likewise.
6567         * gcov-tool.c (print_overlap_usage_message): Remove
6568         trailing empty line.
6569         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
6570         (output_lines): Remove program related line.
6571         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
6572         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
6573         histogram.
6574         (input_profile_summary): Do not read it.
6575         (merge_profile_summaries): And do not merge it.
6576         (input_symtab): Do not call removed function.
6577         * modulo-sched.c (sms_schedule): Do not print sum_max.
6578         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
6579         removed when histogram method was invented.
6580         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
6581         mode.
6582         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
6583         GCOV coding style.
6584         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
6585         and dump selected value.
6586         * profile.c (add_working_set): Remove.
6587         (get_working_sets): Likewise.
6588         (find_working_set): Likewise.
6589         (get_exec_counts): Do not work with working sets.
6590         (read_profile_edge_counts): Do not inform as sum_max is removed.
6591         (compute_branch_probabilities): Likewise.
6592         (compute_value_histograms): Remove argument for call of
6593         get_coverage_counts.
6594         * profile.h: Do not make gcov_summary const.
6596 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
6598         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
6600 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6602         PR tree-optimization/86990
6603         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
6604         Check that the entire merged store group is made of constants only for
6605         overlapping stores.
6607 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
6609         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
6610         (VTABLE_VERIFICATION_SPEC): Likewise.
6611         (SANITIZER_EARLY_SPEC): Likewise.
6612         (SANITIZER_SPEC): Likewise.
6613         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
6614         * doc/invoke.texi (Link Options): Document -r.
6616 2018-09-20  Richard Biener <rguenther@suse.de>
6618         PR middle-end/87054
6619         * gimplify.c (gimplify_expr): Retain alignment of
6620         addressable lvalue in dereference.
6622 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
6624         PR bootstrap/87013
6625         * configure.ac: Check for .loc is_stmt support.
6626         * configure, config.in: Rebuilt.
6627         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
6628         if not supported.
6630 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6632         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
6633         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
6634         -misel=no.
6636 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6638         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6639         VECTOR_OTHER.
6640         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
6641         case VECTOR_OTHER.
6643 2018-09-20  Marek Polacek  <polacek@redhat.com>
6645         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
6647 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6649         PR tree-optimization/87288
6650         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
6651         into account when determining PEELING_FOR_NITERS.
6653 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6655         PR tree-optimization/86877
6656         * tree-vect-loop.c (vect_analyze_loop_2): Call
6657         vect_verify_datarefs_alignment.
6659 2018-09-19  Marek Polacek  <polacek@redhat.com>
6661         * doc/invoke.texi: Document -Wclass-conversion.
6663 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
6665         * config/pa/pa.c (pa_adjust_priority): Delete.
6666         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
6668         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
6669         (atomic_storehi): Likewise.
6670         (atomic_storesi): Likewise.
6671         (atomic_loaddi): Restore compare and swap exchange loop code.
6673 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
6675         PR rtl-optimization/86902
6676         * combine.c (try_combine): When changing the CC mode used, don't change
6677         an unrelated mode in other_insn to that new CC mode.
6679 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
6681         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
6682         with %T in place of calls to dump_generic_expr.
6683         (prune_runtime_alias_test_list): Likewise.
6684         (create_runtime_alias_checks): Likewise.
6685         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
6686         (vect_analyze_data_ref_dependence): Likewise.
6687         (vect_slp_analyze_data_ref_dependence): Likewise.
6688         (vect_record_base_alignment): Likewise.  Use %G in place of call
6689         to dump_gimple_stmt.
6690         (vect_compute_data_ref_alignment): Likewise.
6691         (verify_data_ref_alignment): Likewise.
6692         (vect_find_same_alignment_drs): Likewise.
6693         (vect_analyze_group_access_1): Likewise.
6694         (vect_analyze_data_ref_accesses): Likewise.
6695         (dependence_distance_ge_vf): Likewise.
6696         (dump_lower_bound): Likewise.
6697         (vect_prune_runtime_alias_test_list): Likewise.
6698         (vect_find_stmt_data_reference): Likewise.
6699         (vect_analyze_data_refs): Likewise.
6700         (vect_create_addr_base_for_vector_ref): Likewise.
6701         (vect_create_data_ref_ptr): Likewise.
6702         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
6703         (vect_can_advance_ivs_p): Likewise.
6704         (vect_update_ivs_after_vectorizer): Likewise.
6705         (vect_gen_prolog_loop_niters): Likewise.
6706         (vect_prepare_for_masked_peels): Likewise.
6707         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6708         (vect_determine_vectorization_factor): Likewise.
6709         (vect_is_simple_iv_evolution): Likewise.
6710         (vect_analyze_scalar_cycles_1): Likewise.
6711         (vect_analyze_loop_operations): Likewise.
6712         (report_vect_op): Likewise.
6713         (vect_is_slp_reduction): Likewise.
6714         (check_reduction_path): Likewise.
6715         (vect_is_simple_reduction): Likewise.
6716         (vect_create_epilog_for_reduction): Likewise.
6717         (vect_finalize_reduction:): Likewise.
6718         (vectorizable_induction): Likewise.
6719         (vect_transform_loop_stmt): Likewise.
6720         (vect_transform_loop): Likewise.
6721         (optimize_mask_stores): Likewise.
6722         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
6723         (vect_split_statement): Likewise.
6724         (vect_recog_over_widening_pattern): Likewise.
6725         (vect_recog_average_pattern): Likewise.
6726         (vect_determine_min_output_precision_1): Likewise.
6727         (vect_determine_precisions_from_range): Likewise.
6728         (vect_determine_precisions_from_users): Likewise.
6729         (vect_mark_pattern_stmts): Likewise.
6730         (vect_pattern_recog_1): Likewise.
6731         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6732         (vect_record_max_nunits): Likewise.
6733         (vect_build_slp_tree_1): Likewise.
6734         (vect_build_slp_tree_2): Likewise.
6735         (vect_print_slp_tree): Likewise.
6736         (vect_analyze_slp_instance): Likewise.
6737         (vect_detect_hybrid_slp_stmts): Likewise.
6738         (vect_detect_hybrid_slp_1): Likewise.
6739         (vect_slp_analyze_operations): Likewise.
6740         (vect_slp_analyze_bb_1): Likewise.
6741         (vect_transform_slp_perm_load): Likewise.
6742         (vect_schedule_slp_instance): Likewise.
6743         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6744         (vect_mark_stmts_to_be_vectorized): Likewise.
6745         (vect_init_vector_1): Likewise.
6746         (vect_get_vec_def_for_operand): Likewise.
6747         (vect_finish_stmt_generation_1): Likewise.
6748         (vect_check_load_store_mask): Likewise.
6749         (vectorizable_call): Likewise.
6750         (vectorizable_conversion): Likewise.
6751         (vectorizable_operation): Likewise.
6752         (vectorizable_load): Likewise.
6753         (vect_analyze_stmt): Likewise.
6754         (vect_is_simple_use): Likewise.
6755         (vect_get_vector_types_for_stmt): Likewise.
6756         (vect_get_mask_type_for_stmt): Likewise.
6757         * tree-vectorizer.c (increase_alignment): Likewise.
6759 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
6761         * doc/rtl.texi: Adjust vec_select description.
6762         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
6763         non-constant selectors.
6765 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
6767         * config/aarch64/aarch64-protos.h
6768         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
6769         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
6770         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
6771         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
6772         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
6773         (AARCH64_FL_PROFILE): Move index so flags are ordered.
6774         (AARCH64_ISA_RCPC8_4): New flag.
6775         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
6776         to aarch64_offset_9bit_signed_unscaled_p.
6777         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
6778         and use stlur.
6779         * config/aarch64/constraints.md (Ust): New constraint.
6780         * config/aarch64/predicates.md.
6781         (aarch64_9bit_offset_memory_operand): New predicate.
6782         (aarch64_rcpc_memory_operand): New predicate.
6784 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
6786         PR rtl-optimization/87361
6787         * rtlanal.c (nonzero_bits1): Revert accidental change.
6789 2018-09-19  Richard Biener  <rguenther@suse.de>
6791         PR tree-optimization/87349
6792         PR tree-optimization/87342
6793         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
6795 2018-09-18  Marek Polacek  <polacek@redhat.com>
6797         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
6798         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
6800 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6802         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
6804 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6806         PR rtl-optimization/86882
6807         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
6809 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
6811         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
6812         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
6814 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
6816         PR other/87353
6817         * doc/invoke.texi (Link Options): Fix formatting and grammar.
6819 2018-09-18  Richard Biener  <rguenther@suse.de>
6821         PR middle-end/63155
6822         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
6823         (compute_samebase_partition_bases): Likewise.
6824         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
6825         (gimple_can_coalesce_p): Simplify.
6827 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
6829         Handle a library implementation of ffs calling __builtin_ffs.
6830         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
6831         (mmix_init_libfuncs): New function: make __builtin_ffs expand
6832         to __ffsdi2.
6834 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6836         * diagnostic-show-locus.c (class layout_range): Add field
6837         "m_original_idx".
6838         (layout_range::layout_range): Add "original_idx" param and use it
6839         to initialize new field.
6840         (make_range): Use 0 for original_idx.
6841         (layout::layout): Pass in index to calls to
6842         maybe_add_location_range.
6843         (layout::maybe_add_location_range): Add param "original_idx" and
6844         pass it on to layout_range.
6845         (layout::print_any_labels): Pass on range->m_original_idx to
6846         get_text call.
6847         (gcc_rich_location::add_location_if_nearby): Use 0 for
6848         original_idx.
6849         * gcc-rich-location.h (text_range_label::get_text): Update for new
6850         param.
6851         (range_label_for_type_mismatch::get_text): Likewise.
6853 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6855         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
6857 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6859         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
6860         format_string_diagnostic_t.
6861         (fmtwarn_n): Likewise.
6862         * substring-locations.c
6863         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
6864         (format_warning_n_va): Convert to...
6865         (format_string_diagnostic_t::emit_warning_n_va): ...this.
6866         (format_warning_va): Convert to...
6867         (format_string_diagnostic_t::emit_warning_va): ...this.
6868         (format_warning_at_substring): Convert to...
6869         (format_string_diagnostic_t::emit_warning): ...this.
6870         (format_warning_at_substring_n): Convert to...
6871         (format_string_diagnostic_t::emit_warning_n): ...this.
6872         * substring-locations.h (class format_string_diagnostic_t): New
6873         class.
6874         (format_warning_va): Convert to
6875         format_string_diagnostic_t::emit_warning_va.
6876         (format_warning_n_va): Convert to
6877         format_string_diagnostic_t::emit_warning_n_va.
6878         (format_warning_at_substring): Convert to
6879         format_string_diagnostic_t::emit_warning.
6880         (format_warning_at_substring_n): Convert to
6881         format_string_diagnostic_t::emit_warning_n.
6883 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
6884             Bernd Schmidt <bernds_cb1@t-online.de>
6886         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
6887         SImode args.
6889 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6891         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
6892         operand 0 predicate to nonimmediate operand.
6893         (rint<mode>2_frndint): Remove insn pattern.
6894         (rint<mode>2): Change operand 1 predicate to general_operand.
6895         Extend operand 1 to XFmode and generate rintxf2 insn.
6896         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
6897         Do not use X87MODEF mode macro.
6898         (frndintxf2_<rounding>_i387): Rename from
6899         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
6900         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
6901         to XFmode and generate significandxf3 insn.
6903 2018-09-17  Richard Biener  <rguenther@suse.de>
6905         PR tree-optimization/87328
6906         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
6907         visiting unexecutable backedges when not iterating.
6908         (do_rpo_vn): Mark all edges not executable even when not
6909         iterating.
6911 2018-09-17  Martin Jambor  <mjambor@suse.cz>
6913         PR c/63886
6914         * doc/invoke.texi (Warning Options): Likewise.
6916 2018-09-17  Richard Biener  <rguenther@suse.de>
6918         PR tree-optimization/87301
6919         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
6920         clean EH info from leftover copy assignments.
6922 2018-09-17  Martin Liska  <mliska@suse.cz>
6924         PR gcov-profile/85871
6925         * gcov.c (output_intermediate_file): Fix out of bounds
6926         access.
6928 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
6930         * config/arc/arc.c: Object attributes for core4 not reflected
6931         correctly.
6932         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
6933         core3).
6935 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
6937         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
6939 2018-09-17  Martin Liska  <mliska@suse.cz>
6941         * doc/gcov.texi: Document new option --use-hotness-colors.
6942         * gcov.c (struct source_info): Declare new field.
6943         (source_info::source_info): Set default for maximum_count.
6944         (print_usage): Add new -q option.
6945         (process_args): Process it.
6946         (accumulate_line_info): Save src->maximum_count.
6947         (output_line_beginning): Make color line number if
6948         flag_use_hotness_colors is set.
6949         (output_line_details): Pass default argument value.
6950         (output_lines): Pass src->maximum_count.
6952 2018-09-17  Martin Liska  <mliska@suse.cz>
6954         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
6955         Use processor_names table.
6956         * config/i386/i386.c (ix86_default_align): Use
6957         processor_cost_table for alignment values.
6958         (ix86_option_override_internal): Use processor_names.
6959         (ix86_function_specific_print): Likewise.
6960         * config/i386/i386.h (struct processor_costs):
6961         Add alignment values.
6962         (struct ptt): Remove and replace with const char *.
6963         * config/i386/x86-tune-costs.h (struct processor_costs):
6964         Declare default alignments for all costs.
6966 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
6968         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
6969         symbolics or VR_VARYING ranges for ABS_EXPR.
6970         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
6971         when range will wrap.
6973 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
6975         PR middle-end/86864
6976         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
6977         before and after a JUMP_TABLE_DATA.
6979 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
6981         PR middle-end/87188
6982         * dojump.c (do_compare_and_jump): Canonicalize function pointers
6983         when one operand is a function pointer.  Use POINTER_TYPE_P and
6984         FUNC_OR_METHOD_TYPE_P.
6985         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
6986         * fold-const.c (build_range_check): Likewise.
6987         * match.pd (simple_comparison): Likewise.
6989 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
6991         PR c/82967
6992         * spellcheck.c (get_edit_distance_cutoff): New function.
6993         (selftest::test_edit_distance_unit_test_oneway): Rename to...
6994         (selftest::test_get_edit_distance_one_way): ...this.
6995         (selftest::test_get_edit_distance_unit): Rename to...
6996         (selftest::test_get_edit_distance_both_ways): ...this.
6997         (selftest::test_edit_distances): Move tests to this new function,
6998         and test some more pairs of strings.  Update for above renaming.
6999         (selftest::get_old_cutoff): New function.
7000         (selftest::test_get_edit_distance_cutoff): New function.
7001         (selftest::assert_suggested_for): New function.
7002         (ASSERT_SUGGESTED_FOR): New macro.
7003         (selftest::assert_not_suggested_for): New function.
7004         (ASSERT_NOT_SUGGESTED_FOR): New macro.
7005         (selftest::test_suggestions): New function.
7006         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
7007         tests to selftest::test_edit_distances and call it.  Add calls to
7008         selftest::test_get_edit_distance_cutoff and
7009         selftest::test_suggestions.
7010         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
7011         (best_match::consider): Replace hard-coded cutoff calculation with
7012         a call to...
7013         (best_match::get_cutoff): New declaration.
7014         (best_match::get_best_meaningful_candidate): Likewise.
7016 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7018         * builtins.c (fold_builtin_strlen): Remove TODO comment.
7020 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7022         revert:
7023         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7025         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
7026         terminated string literal.
7028 2018-09-14  Martin Sebor  <msebor@redhat.com>
7030         * builtins.c (unterminated_array): Handle ARRAY_REF.
7031         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
7032         * builtins.h (unterminated_array): Declare extern.
7033         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
7034         arrays.
7035         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
7036         calls.
7038 2018-09-14  Martin Sebor  <msebor@redhat.com>
7039             Jeff Law  <law@redhat.com>
7041         * builtins.c (unterminated_array): New.
7042         (expand_builtin_strcpy): Adjust.
7043         (expand_builtin_strcpy_args): Detect unterminated arrays.
7044         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
7045         unterminated arrays.
7046         * gimple-fold.h (get_maxval_strlen): Add argument.
7047         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
7049         * gimple-fold.c (get_range_strlen): Add argument.
7050         (get_maxval_strlen): Adjust.
7051         * gimple-fold.h (get_range_strlen): Add argument.
7053 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
7055         * config/i386/movdirintrin.h: Fix copyright year.
7057 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
7059         * reg-stack.c: Include regs.h.
7060         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
7061         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
7062         FIRST_STACK_REG, not DFmode.
7063         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
7064         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
7065         (change stack): Default register mode to the reg_raw_mode of
7066         FIRST_STACK_REG, not DFmode.
7067         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
7068         (*swapxf): Rename from swapxf.
7070 2018-09-14  Carl Love  <cel@us.ibm.com>
7072         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
7073         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
7075 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
7077         PR target/87224
7078         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
7079         alternatives.
7081 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
7083         PR target/85628
7084         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
7086 2018-09-14  Jason Merrill  <jason@redhat.com>
7088         Fix --enable-gather-detailed-mem-stats.
7089         * hash-table.c (hash_table_usage): Change from variable to function.
7090         * hash-table.h: Adjust.
7091         * Makefile.in: Add missing dependencies on hash-table.h.
7093 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7095         PR tree-optimization/87259
7096         PR lto/87283
7097         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
7098         execute_cse_reciprocals_1 has tried transforming.
7100 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
7102         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
7103         VR_VARYING for PLUS/MINUS_EXPR.
7105 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
7107         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
7108         formatting.
7110 2018-09-14  Richard Biener  <rguenther@suse.de>
7112         PR middle-end/63155
7113         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
7114         bits for the merged partition.
7116 2018-09-13  Martin Sebor  <msebor@redhat.com>
7117             Bernd Edlinger  <bernd.edlinger@hotmail.de>
7119         * builtins.h (c_srlen): Add argument.
7120         * builtins.c (warn_string_no_nul): New function.
7121         (c_strlen): Add argument and use it.  Update recursive calls.
7122         Pass DECL argument to string_constant to get info on non
7123         terminated strings.  Update *NONSTR as needed.
7124         (fold_builtin_strlen): Add argument to calls to c_strlen.
7125         Warn for unterminated arrays.
7126         (warn_string_no_null): Add prototype.
7127         * expr.c (string_constant): Update arguments.  Update recursive
7128         calls appropriately.  Detect missing NUL terminator and outermost
7129         declaration its missing in.
7130         Improve checks for arrays with nonzero lower bound or elements
7131         that are not a single byte.  Simplify offset computation.
7132         Simplify checks for non-NUL terminated strings.
7133         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
7134         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
7136 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7138         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
7139         correctly.
7140         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
7141         argument.  Fix range checks.
7142         * fold-const.h (c_getstr): Adjust protoype.
7143         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
7144         string is constant but contains no NUL byte.
7146         * expr.c (string_constant): Adjust function comment.
7147         Remove bogus check for zero termination.
7149         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
7151         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
7152         (get_constant_size): Don't make STRING_CSTs larger than they are.
7153         (check_string_literal): New check function for STRING_CSTs.
7154         (output_constant): Use it.
7156 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
7158         PR target/86812
7159         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
7161 2018-09-13  Richard Biener  <rguenther@suse.de>
7163         PR tree-optimization/87263
7164         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
7165         (struct unwind_state): Add max_rpo field.
7166         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
7167         Compute max_rpo, the max RPO number a block can be backwards reached
7168         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
7169         separating it from the iterating mode.
7171 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
7173         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
7174         (rfs_decision): New scheduling decision.
7176 2018-09-13  Richard Biener  <rguenther@suse.de>
7178         PR bootstrap/87134
7179         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
7180         (vn_nary_op_insert_pieces_predicated): Do not write useless
7181         valid_dominated_by_p entry outside of the allocated storage.
7183 2018-09-13  Omar Sandoval  <osandov@osandov.com>
7184             Tom de Vries  <tdevries@suse.de>
7186         PR debug/86985
7187         * dwarf2out.c (is_c): New function.
7188         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
7190 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
7192         PR target/85628
7193         * config/aarch64/aarch64.md (*aarch64_bfxil):
7194         Define.
7195         * config/aarch64/constraints.md (Ulc): Define.
7196         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
7197         Define.
7198         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
7199         New function.
7201 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
7203         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
7204         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
7205         aarch64_layout_frame call.
7206         (aarch64_expand_epilogue): Likewise.
7207         (aarch64_initial_elimination_offset): Likewise.
7208         (aarch64_get_separate_components): Likewise.
7209         (aarch64_use_return_insn_p): Likewise.
7210         (aarch64_layout_frame): Remove unneeded check.
7212 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7214         * configure.ac: Only append
7215         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
7216         gcc_config_arguments if it was never reconfigured or last reconfigure
7217         was with different arguments.
7218         * configure: Regenerated.
7220 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7221             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7223         PR middle-end/87290
7224         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
7225         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
7227 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7229         PR tree-optimization/87287
7230         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
7231         X % C == 0 to X % (unsigned) C == 0 optimization to ...
7232         * match.pd (X % C == 0): ... here.  New optimization.
7234 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7236         PR middle-end/82853
7237         * expr.h (maybe_optimize_mod_cmp): Declare.
7238         * expr.c (mod_inv): New function.
7239         (maybe_optimize_mod_cmp): New function.
7240         (do_store_flag): Use it.
7241         * cfgexpand.c (expand_gimple_cond): Likewise.
7243 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
7244             Julian Brown  <julian@codesourcery.com>
7246         PR middle-end/86336
7247         * gimplify.c (gimplify_scan_omp_clauses): Set
7248         target_firstprivatize_array_bases in OpenACC parallel and kernels
7249         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
7250         OpenACC data regions.
7252 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
7254         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
7255         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
7257 2018-09-12  Richard Biener  <rguenther@suse.de>
7259         PR tree-optimization/87280
7260         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
7261         edge but unreachable target.
7262         (do_rpo_vn): For conservatively handling a PHI only mark
7263         the backedge executable but not the block reachable.
7265 2018-09-12  Richard Biener  <rguenther@suse.de>
7267         PR tree-optimization/87266
7268         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
7269         visited blocks.
7271 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
7273         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
7274         constants.
7275         ("trunc<BFP:mode><DFP_ALL:mode>2")
7276         ("trunc<DFP_ALL:mode><BFP:mode>2")
7277         ("extend<BFP:mode><DFP_ALL:mode>2")
7278         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
7279         according to the target operand type.
7281 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7282             Andreas Krebbel  <krebbel@linux.ibm.com>
7284         PR tree-optimization/86844
7285         * gimple-ssa-store-merging.c
7286         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
7287         there are any overlapping stores in between them, make sure they are
7288         also coalesced or we give up completely.
7290 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7292         PR middle-end/87248
7293         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
7294         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
7296 2018-09-12  Tom de Vries  <tdevries@suse.de>
7298         * common.opt (gdescribe-dies): Add option.
7299         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
7300         attribute for artifical and nameless decls.
7301         (dwarf2out_register_external_die): Add description attribute to
7302         external reference die.
7303         (add_desc_attribute): New functions.
7304         (gen_subprogram_die): Add description attribute to
7305         DW_TAG_call_site_parameter.
7306         * tree-pretty-print.c (print_generic_expr_to_str): New function.
7307         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
7308         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
7309         -gno-describe-dies.
7310         (@item -gdescribe-dies): Add.
7312 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
7314         * tree-vrp.c (vrp_shift_undefined_p): Remove.
7315         (extract_range_from_binary_expr_1: Call
7316         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
7317         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
7318         depend on sign.
7320 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
7322         * gimple-ssa-warn-alloca.c
7323         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
7324         field for ALLOCA_BOUND_*_LARGE.
7326 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7328         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
7330 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7332         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
7333         for clobbers.  Remove obsolete comment.
7335 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7337         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
7338         mpxchk, mpxld and mpxst types.
7339         (define_attr length_immediate): Remove all processing of mpx types.
7340         (define_attr prefix_0f): Ditto.
7341         (define_attr memory): Ditto.
7343 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7345         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
7346         (log<mode>2): Change operand 1 predicate to general_operand.
7347         Extend operand 1 to XFmode and generate logxf3 insn.
7348         (log10<mode>2): Change operand 1 predicate to general_operand.
7349         Extend operand 1 to XFmode and generate log10xf3 insn.
7350         (log2<mode>2): Change operand 1 predicate to general_operand.
7351         Extend operand 1 to XFmode and generate log2xf3 insn.
7352         (fyl2xp1_extend<mode>xf3_i387): Remove.
7353         (log1p<mode>2): Change operand 1 predicate to general_operand.
7354         Extend operand 1 to XFmode and generate log1pxf3 insn.
7355         (fxtract_extend<mode>xf3_i387): Remove.
7356         (logb<mode>2): Change operand 1 predicate to general_operand.
7357         Extend operand 1 to XFmode and generate logbxf3 insn.
7358         (ilogb<mode>2): Change operand 1 predicate to general_operand.
7359         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
7360         (significand<mode>2): Change operand 1 predicate to general_operand.
7361         Extend operand 1 to XFmode and generate significandxf3 insn.
7363 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7365         * gcc.c (perror_with_name, pfatal_with_name): Delete.
7366         (load_specs): Use fatal_error.
7367         (DELETE_IF_ORDINARY, process_command): Use error.
7368         (execute, run_attempt): Use fatal_error.
7370 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
7372         * diagnostic-core.h (sorry_at): New prototype.
7373         * diagnostic.c (sorry_at): New function.
7375 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
7377         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
7378         by zero as VR_UNDEFINED.
7380 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
7382         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
7383         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
7384         (<sincos>mode2): New expander.
7385         (sincos_extend<mode>xf3_i387): Remove insn pattern.
7386         (sincos -> sin, cos splitters): Remove splitter patterns.
7387         (sincos<mode>3): Change operand 2 predicate to general_operand.
7388         Extend operand 2 to XFmode and generate sincosxf3 insn.
7389         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
7390         Change operand 3 predicate to const1_operand.
7391         (fptan_extend<mode>xf4_i387): Remove insn pattern.
7392         (tanxf2): Update operands in the call to fptanxf4_i387.
7393         (tan<mode>2): Change operand 1 predicate to general_operand.
7394         Extend operand 1 to XFmode and generate tanxf3 insn.
7395         (atan2xf3): Rename from *fpatanxf3_i387.
7396         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
7397         (atan2xf3): Remove expander.
7398         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
7399         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
7400         (atan<mode>2): Change operand 1 predicate to general_operand.
7401         Extend operand 1 to XFmode and generate atanxf3 insn.
7403 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
7405         * config/i386/i386.md (x87/SSE constant load splitter): Use
7406         memory_operand instead of nonimmediate_operand for input operand
7407         predicate.
7409 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
7411         * config/i386/i386.md (float partial SSE register stall splitter): Move
7412         splitter near its instruction pattern.
7413         (float_extend partial SSE register stall splitter): Ditto.
7414         (float_truncate partial SSE register stall splitter): Ditto.
7416 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
7418         PR target/86794
7419         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7420         to speculation_safe_value_not_needed.
7422         PR target/85666
7423         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
7424         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
7425         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
7426         leaf_function_p, instead use has_hard_reg_initial_val.
7428 2018-09-09  Nathan Sidwell  <nathan@acm.org>
7430         * gcc.h (pfatal_with_name): Don't declare here.
7431         * gcc.c (pfatal_with_name): Make static.
7433 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
7435         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
7436         earlyclobber.
7438 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
7440         PR rtl-optimization/85458
7441         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
7442         priority hook to reduce the priority of EXPR.
7444 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
7446         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
7447         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
7448         with X87_ENABLE_FLOAT.  Remove preparation code.
7449         (*float<SWI48:mode><MODEF:mode>2): Rename from
7450         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
7451         math using "enabled" attribute.
7452         (*floatdi<MODEF:mode>2_i387): Rename from
7453         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
7454         enable for 32bit targets only.
7455         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
7456         splitter.
7457         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
7458         as operand 1 predicate.  Rewrite as define_insn_and_split.
7459         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
7461 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7463         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
7464         to fallthru to FLOAT case.
7466 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7468         PR target/86731
7469         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
7470         around folding of vec_sl to handle out of range shift values.
7472 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7474         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
7475         Update callers to gen_fix_trunc<mode>_i387_fisttp
7476         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
7477         nonimmediate_operand.
7478         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
7479         and corresponding splitters.
7480         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
7481         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
7482         (fix_truncdi_i387_with_temp): Remove insn pattern
7483         and corresponding splitters.
7484         (fix_trunc<mode>_i387): Change operand 0 predicate to
7485         nonimmediate_operand.
7486         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
7487         and corresponding splitters.
7488         (*fistdi2_1): Remove.
7489         (fistdi2): Ditto.
7490         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
7491         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
7492         (*fist<mode>2_1): Remove.
7493         (fist<mode>2): Ditto.
7494         (fist<mode>2_with_temp): Remove insn pattern and corresponding
7495         splitters.
7496         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
7497         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
7498         (fistdi2_<rounding>): Change operand 0 predicate to
7499         nonimmediate_operand.
7500         (fistdi2_<rounding>_with_temp): Remove insn pattern
7501         and corresponding splitters.
7502         (fist<mode>2_<rounding>): Change operand 0 predicate to
7503         nonimmediate_operand.
7504         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
7505         and corresponding splitters.
7507         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
7509 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7511         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
7512         the init value.
7514 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7516         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7517         early gimple folding of vec_splat().
7518         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
7519         * gimple-fold.h: Add an extern define for tree_vec_extract().
7521 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7523         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
7524         wrappers around TREE_TYPE comparisons.
7526 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7528         PR target/80080
7529         * config/s390/predicates.md: Add nonsym_memory_operand.
7530         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
7531         contains a SYMBOL_REF, load it into an intermediate pseudo.
7532         (s390_emit_compare_and_swap): Legitimize operand.
7533         * config/s390/s390.md: Use the new nonsym_memory_operand
7534         with UNSPECV_CAS patterns.
7536 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7538         PR target/80080
7539         * config/s390/s390-passes.def: New file.
7540         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
7541         declaration.
7542         (make_pass_s390_early_mach): Add declaration.
7543         * config/s390/s390.c (make_pass_s390_early_mach):
7544         (s390_option_override): Remove dynamic registration.
7545         * config/s390/t-s390: Add s390-passes.def.
7547 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7549         * config/s390/s390.c (s390_decompose_constant_pool_ref):
7550         Remove UNSPEC_LTREL_BASE check.
7551         (annotate_constant_pool_refs): Likewise.
7552         (find_constant_pool_ref): Likewise.
7553         (find_ltrel_base): Removed.
7554         (replace_ltrel_base): Removed.
7555         (s390_mainpool_finish): Remove replace_ltrel_base call.
7556         (s390_chunkify_start): Remove pending LTREL_BASE logic.
7557         (s390_chunkify_finish): Remove replace_ltrel_base call.
7558         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
7560 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
7562         PR target/86779
7563         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7564         to speculation_safe_value_not_needed.
7566 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
7567             Bernd Schmidt  <bernds_cb1@t-online.de>
7569         * config/nvptx/nvptx-opts.h: New file.
7570         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
7571         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
7572         (ASM_SPEC): Define.
7573         (TARGET_SM35): New macro.
7574         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
7575         correct predicate.
7576         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
7577         values.
7578         (misa=): New option.
7579         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
7581 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
7583         * config/i386/i386.md (truncdfsf2): Remove expander.
7584         (truncdfsf2_with_temp): Ditto.
7585         (truncxf<mode>2): Ditto.
7586         (*truncdfsf_fast_mixed): Remove insn pattern.
7587         (*truncdfsf_fast_i387): Ditto.
7588         (*truncdfsf_mixed): Ditto.
7589         (*truncdfsf_i387): Ditto.
7590         (*truncdfsf2_i387_1): Ditto.
7591         (*truncxfsf2_mixed): Ditto.
7592         (*truncxfdf2_mixed): Ditto.
7593         (*truncxf<mode>2_i387_noop): Ditto. Update callers
7594         to call gen_truncxf<mode>2 instead.
7595         (*truncxf<mode>2_i387): Remove.
7596         (reg->reg splitters): Remove splitter pattern.
7597         (reg->mem splitters): Ditto.
7599         (truncdfsf2): New insn pattern.
7600         (truncxf<mode>2): Ditto.
7602 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7604         * tree-ssa-math-opts.c (is_mult_by): New function.
7605         (is_square_of): Use the above.
7606         (optimize_recip_sqrt): New function.
7607         (pass_cse_reciprocals::execute): Use the above.
7609 2018-09-05  Richard Biener  <rguenther@suse.de>
7611         PR bootstrap/87134
7612         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
7613         to zero-init the emplaced vec.
7615 2018-09-05  Martin Liska  <mliska@suse.cz>
7617         PR tree-optimization/87205
7618         * tree-switch-conversion.c (pass_lower_switch::execute):
7619         Group cases for switch statements.
7621 2018-09-05  Richard Biener  <rguenther@suse.de>
7623         PR tree-optimization/87217
7624         * tree-ssa-sccvn.c (vuse_valueize): New.
7625         (vn_reference_lookup_pieces): Use it.
7626         (vn_reference_lookup): Likewise.
7628 2018-09-05  Nathan Sidwell  <nathan@acm.org>
7630         PR c++/87137
7631         * stor-layout.c (place_field): Scan forwards to check last
7632         bitfield when ms_bitfield_placement is in effect.
7634 2018-09-05  Richard Biener  <rguenther@suse.de>
7636         PR bootstrap/87225
7637         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
7638         return.
7640 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7641             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
7643         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
7644         * config.gcc (extra_objs): Build it.
7645         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
7646         Likewise.
7647         * config/aarch64/aarch64-passes.def
7648         (pass_tag_collision_avoidance): New pass.
7649         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
7650         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
7651         (aarch64_classify_address): Remove static qualifier.
7652         (aarch64_address_info, aarch64_address_type): Move to...
7653         * config/aarch64/aarch64-protos.h: ... here.
7654         (make_pass_tag_collision_avoidance): New function.
7655         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
7656         New tuning flag.
7658 2018-09-05  Martin Liska  <mliska@suse.cz>
7660         * doc/gcov.texi: Update documentation of humar
7661         readable mode.
7662         * gcov.c (format_count): Print one decimal place, it provides
7663         more fine number of situations like '1G' vs. '1.4G'.
7665 2018-09-05  Martin Liska  <mliska@suse.cz>
7667         PR target/87164
7668         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
7669         * optc-gen.awk: Allow 'Var' for Deprecated options in order
7670         to generate a MASK value.
7672 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
7674         PR debug/86593
7675         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
7676         if frame pointer isn't used.
7677         (compute_frame_pointer_to_fb_displacement): Likewise.
7679 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
7681         PR target/87198
7682         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
7683         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
7684         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
7685         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
7686         and OPTION_MASK_ISA_XSAVEC_UNSET.
7688 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
7690         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
7691         XOR operations in NAND case.
7693 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7695         * wide-int-range.cc (wide_int_range_convert): New.
7696         * wide-int-range.h (wide_int_range_convert): New.
7697         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
7698         code into wide_int_range_convert.
7699         (extract_range_into_wide_ints): Do not munge anti range constants
7700         into the entire domain.  Just return the range back.
7702 2018-09-04  Martin Liska  <mliska@suse.cz>
7704         * genmatch.c (output_line_directive): Add new argument
7705         fnargs.
7706         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
7708 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7710         * doc/invoke.texi (Option Summary): Add whitespace.
7712         * doc/invoke.texi (Option Summary): Add -Waligned-new.
7714 2018-09-04  Richard Biener  <rguenther@suse.de>
7716         PR tree-optimization/87211
7717         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
7718         backedge value we're supposed to treat as VARYING also number
7719         the PHI to VARYING in case it got a different value-number already.
7721 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7723         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
7724         (extract_range_from_binary_expr_1): Do not call
7725         vrp_can_optimize_bit_op.
7726         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
7727         static.
7728         (wide_int_range_get_mask_and_bounds): New.
7729         (wide_int_range_optimize_bit_op): New.
7730         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
7731         (wide_int_range_bit_and): Same.
7732         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
7733         (wide_int_range_optimize_bit_op): New.
7734         (wide_int_range_get_mask_and_bounds): New.
7736 2018-09-04  Richard Biener  <rguenther@suse.de>
7738         PR tree-optimization/87176
7739         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
7740         variable.  When value-numbering a virtual PHI node make sure
7741         to not value-number to the backedge value.
7743 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7745         * doc/extend.texi (Long Long, Hex Floats): Document support for
7746         long long and hex floats in more recent versions of ISO C++.
7748 2018-09-03  Richard Biener  <rguenther@suse.de>
7750         PR tree-optimization/87177
7751         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
7752         cleanup.
7754 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7756         * bb-reorder.c (edge_order): Convert to C-qsort-style
7757         tri-state comparator.
7758         (reorder_basic_blocks_simple): Change std::stable_sort to
7759         gcc_stablesort.
7761 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7763         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
7764         tri-state comparator.
7765         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
7767 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7769         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
7770         (mergesort): ... here as maximum count for using netsort.
7771         (gcc_qsort): Set nlim to 3 if stable sort is requested.
7772         (gcc_stablesort): New.
7773         * system.h (gcc_stablesort): Declare.
7775 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7777         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
7778         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
7779         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
7781 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
7783         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
7784         lxsdx and stxsdx alternatives.
7785         (*mov<mode>_hardfloat64): Ditto.
7786         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
7788 2018-09-03  Richard Biener  <rguenther@suse.de>
7790         PR tree-optimization/87200
7791         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
7792         simplify result.
7794 2018-09-03  Martin Liska  <mliska@suse.cz>
7796         PR tree-optimization/87201
7797         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7798         Fix parenthesis in an expression.
7800 2018-09-03  Richard Biener  <rguenther@suse.de>
7802         PR tree-optimization/87197
7803         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
7804         visited.  CSE the VN_INFO hashtable lookup.
7806         PR tree-optimization/87169
7807         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
7808         iterating make sure there's no extra backedges from irreducible
7809         regions feeding the header.  Mark the destination block
7810         executable.
7812 2018-09-03  Martin Liska  <mliska@suse.cz>
7814         PR driver/83193
7815         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
7816         * common/common-targhooks.c (default_get_valid_option_values):
7817         New function.
7818         * common/common-targhooks.h (default_get_valid_option_values):
7819         Likewise.
7820         * common/config/i386/i386-common.c: Move processor_target_table
7821         from i386.c.
7822         (ix86_get_valid_option_values): New function.
7823         (TARGET_GET_VALID_OPTION_VALUES): New macro.
7824         * config/i386/i386.c (struct ptt): Move to i386-common.c.
7825         (PTA_*): Move all defined masks into i386-common.c.
7826         (ix86_function_specific_restore): Use new processor_cost_table.
7827         * config/i386/i386.h (struct ptt): Moved from i386.c.
7828         (struct pta): Likewise.
7829         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
7830         * doc/tm.texi.in: Likewise.
7831         * opt-suggestions.c (option_proposer::suggest_option):
7832         Pass prefix to build_option_suggestions.
7833         (option_proposer::get_completions): Likewise.
7834         (option_proposer::build_option_suggestions): Use the new target
7835         hook.
7836         * opts.c (struct option_help_tuple): New struct.
7837         (print_filtered_help): Use the new target hook.
7839 2018-09-03  Martin Liska  <mliska@suse.cz>
7841         PR middle-end/59521
7842         * predict.c (set_even_probabilities): Add likely_edges
7843         argument and handle cases where we have precisely one
7844         likely edge.
7845         (combine_predictions_for_bb): Catch also likely_edges.
7846         (tree_predict_by_opcode): Handle gswitch statements.
7847         * tree-cfg.h (find_case_label_for_value): New declaration.
7848         (find_taken_edge_switch_expr): Likewise.
7849         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7850         Find pivot in decision tree based on probabily, not by number of
7851         nodes.
7853 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
7855         * doc/standards.texi (Standards): Update Objective-C reference.
7857 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7859         * doc/install.texi (Prerequisites): Update link for MPC.
7861 2018-09-01  Michael Matz  <matz@suse.de>
7863         PR tree-optimization/87074
7864         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
7865         PHIs for outer-loop uses.
7867 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7869         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
7870         * doc/invoke.texi (C Dialect Options): Ditto.
7872 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7874         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
7876 2018-08-31  Richard Biener  <rguenther@suse.de>
7878         PR tree-optimization/87168
7879         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
7880         (rpo_elim::eliminate_avail): When OP was not visited it must
7881         be available.
7883 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
7885         * tree-vrp.c (copy_value_range): Convert param "from" from
7886         "value_range *" to "const value_range *".
7887         (range_is_null): Likewise for param "vr".
7888         (range_int_cst_p): Likewise.
7889         (range_int_cst_singleton_p): Likewise.
7890         (symbolic_range_p): Likewise.
7891         (value_ranges_intersect_p): Likewise for both params.
7892         (value_range_nonnegative_p): Likewise for param "vr".
7893         (value_range_constant_singleton): Likewise.
7894         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
7895         (extract_range_into_wide_ints): Likewise for param "vr".
7896         (extract_range_from_multiplicative_op): Likewise for params "vr0"
7897         and "vr1".
7898         (vrp_can_optimize_bit_op): Likewise.
7899         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
7900         "vr1_".
7901         (extract_range_from_unary_expr): Likewise.
7902         (debug_value_range): Likewise for param "vr".
7903         (value_range::dump): Add "const" qualifier.
7904         (vrp_prop::check_array_ref): Convert local "vr" from
7905         "value_range *" to "const value_range *".
7906         (vrp_prop::check_mem_ref): Likewise.
7907         (vrp_prop::visit_stmt): Likewise for local "old_vr".
7908         (vrp_intersect_ranges_1): Likewise for param "vr_1".
7909         (vrp_intersect_ranges): Likewise.
7910         (simplify_stmt_for_jump_threading): Likewise for local "vr".
7911         (vrp_prop::vrp_finalize): Likewise.
7912         * tree-vrp.h (value_range::dump): Add "const" qualifier.
7913         (vrp_intersect_ranges): Add "const" qualifier to params as above.
7914         (extract_range_from_unary_expr): Likewise.
7915         (value_range_constant_singleton): Likewise.
7916         (symbolic_range_p): Likewise.
7917         (copy_value_range): Likewise.
7918         (extract_range_from_binary_expr_1): Likewise.
7919         (range_int_cst_p): Likewise.
7920         (vrp_set_zero_nonzero_bits): Likewise.
7921         (range_int_cst_singleton_p): Likewise.
7923 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
7925         * config/aarch64/arm_neon.h (vabsd_s64): New.
7926         (vnegd_s64): Likewise.
7928 2018-08-31  Martin Jambor  <mjambor@suse.cz>
7930         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
7932 2018-08-31  Martin Liska  <mliska@suse.cz>
7934         * ipa-icf.c (sem_item::add_type): Use
7935         sem_item::m_type_hash_cache.
7936         * ipa-icf.h: Move the cache from sem_item_optimizer
7937         to sem_item.
7939 2018-08-31  Nathan Sidwell  <nathan@acm.org>
7941         * doc/extend.texi (Backwards Compatibility): Remove implicit
7942         extern C leeway of () being (...).
7944 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7946         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
7948 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
7950         PR target/86684
7951         PR target/87149
7952         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
7954 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
7956         PR middle-end/87138
7957         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
7958         gen_int_mode.  Formatting fixes.
7960 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
7962         * target.def (custom_function_descriptors): Improve documentation.
7963         * doc/tm.texi.in (Trampolines): Expand discussion of function
7964         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
7965         beginning of the section.
7966         * doc/tm.texi: Regenerated.
7968 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
7970         * cfg.h (class auto_edge_flag): Spell out the template-id of the
7971         base class in the initializer list.  This is a workaround for
7972         building with older GCC.
7973         (class auto_bb_flag): Likewise.
7975 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
7977         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
7978         (altivec_vcmpequ<VI_char>_p): Remove star.
7979         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
7980         vector load modes.
7981         (expand_strncmp_vec_sequence): New function.
7982         (emit_final_str_compare_vec): New function.
7983         (expand_strn_compare): Add support for vector strncmp.
7984         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
7985         length specification to bytes.
7986         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
7987         (vcmpnezb_p): New pattern.
7988         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
7989         for option -mstring-compare-inline-limit.
7991 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
7993         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
7994         (PTA_SKYLAKE): Add PTA_AES.
7995         (PTA_GOLDMONT): Likewise.
7997 2018-08-29  Jan Hubicka  <jh@suse.cz>
7999         PR lto/86517
8000         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
8001         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
8003 2018-08-29  Jan Hubicka  <jh@suse.cz>
8005         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
8006         TYPE_STUB_DECL.
8007         (hash_tree): Do not visit TYPE_STUB_DECL.
8008         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
8009         stream TYPE_STUB_DECL.
8010         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
8011         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
8012         after free_lang_data.
8013         (type_in_anonymous_namespace_p): Likewise.
8015 2018-08-29  Jan Hubicka  <jh@suse.cz>
8017         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
8018         comment that it has to be even number.
8019         (class sreal): Change m_sig type to int32_t.
8020         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
8021         int64_t for temporary calculations.
8022         (sreal_verify_basics): Drop one bit from minimum and maximum.
8024 2018-08-30  Richard Biener  <rguenther@suse.de>
8026         PR tree-optimization/87147
8027         * tree-ssa-sccvn.c (SSA_VISITED): New function.
8028         (visit_phi): When the degenerate result is from the backedge and
8029         we didn't visit its definition yet drop to VARYING.
8030         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
8032 2018-08-29  Jan Hubicka  <jh@suse.cz>
8034         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
8035         DECL_VINDEX.
8036         (hash_tree): Likewise.
8038 2018-08-29  Jan Hubicka  <jh@suse.cz>
8040         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
8041         and TYPE_NEXT_REF_TO.
8043 2018-08-29  Jan Hubicka  <jh@suse.cz>
8045         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
8046         comment that it has to be even number.
8047         (class sreal): Change m_sig type to int32_t.
8048         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
8049         int64_t for temporary calculations.
8050         (sreal_verify_basics): Drop one bit from minimum and maximum.
8052 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
8054         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
8056 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
8058         PR middle-end/86995
8059         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
8060         if to_add is negative.
8062 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8064         PR middle-end/87053
8065         * builtins.c (c_strlen): Improve range checks.
8067 2018-08-29  Martin Sebor  <msebor@redhat.com>
8068             Jeff Law  <law@redhat.com>
8070         PR tree-optimization/86714
8071         PR tree-optimization/86711
8072         * builtins.c (c_strlen): Add arguments to call to string_constant.
8073         * expr.c (string_constant): Add argument.  Detect missing nul
8074         terminator and outermost declaration it's missing in.
8075         * expr.h (string_constant): Add argument.
8076         * fold-const.c (read_from_constant_string): Add arguments to call to
8077         string_constant.
8078         (c_getstr): Likewise.
8079         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
8080         to string_constant.
8081         * tree-ssa-strlen.c (get_stridx): Likewise.
8083 2018-08-29  Jan Hubicka  <jh@suse.cz>
8085         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
8086         Do not stream DECL_VINDEX.
8087         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
8088         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
8089         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
8091 2018-08-29  Richard Biener  <rguenther@suse.de>
8093         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
8094         virtual operands that are not default defs to honor region
8095         boundaries.
8096         (rpo_vn_valueize): Remove ineffective code here.
8098 2018-08-29  Richard Biener  <rguenther@suse.de>
8100         PR tree-optimization/87132
8101         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
8102         when skipping defs reachable over backedges.
8104 2018-08-29  Richard Biener  <rguenther@suse.de>
8106         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
8107         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
8108         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
8109         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
8110         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
8111         (vn_reference_lookup_3): Remove use of const_parms.
8112         (free_rpo_vn): Do not free const_parms.
8113         (do_rpo_vn): Do not call init_const_parms.
8114         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
8115         SSA_NAME_POINTS_TO_READONLY_MEMORY.
8116         (call_may_clobber_ref_p_1): Likewise.
8118 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
8120         PR other/86726
8121         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
8122         (-O): Ditto.
8123         (-ftree-scev-cprop): Document.
8125 2018-08-29  Jan Hubicka  <jh@suse.cz>
8127         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
8128         parameters.
8129         (sreal constructor): Update.
8130         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
8131         sreal:operator/): Update.
8133 2018-08-29  Martin Liska  <mliska@suse.cz>
8135         * tree-switch-conversion.c (switch_conversion::expand):
8136         Strenghten assumption about gswitch statements.
8138 2018-08-29  Richard Biener  <rguenther@suse.de>
8140         PR tree-optimization/87117
8141         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
8142         re-value-number released SSA VDEFs.
8144 2018-08-29  Richard Biener  <rguenther@suse.de>
8146         PR tree-optimization/87126
8147         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
8149 2018-08-28  Jim Wilson  <jimw@sifive.com>
8151         * config/riscv/pic.md: Rewrite.
8152         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
8153         invalid address.
8154         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
8155         (SOFTF, default_load, softload, softstore): New.
8157 2018-08-28  Jeff Law  <law@redhat.com>
8159         * fold-const.c (fold_binary_loc): Remove recently added assert.
8161 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
8163         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
8164         to OP parmeter of generated function.
8166 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
8168         PR tree-optimization/87009
8169         * match.pd: Add boolean optimizations.
8171 2018-08-28  Martin Sebor  <msebor@redhat.com>
8173         PR middle-end/86631
8174         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
8175         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
8176         (pass_walloca::gate): Use it.
8177         (alloca_call_type): Same.
8178         (pass_walloca::execute): Same.
8179         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
8181 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
8183         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
8184         GCC_VERSION for usage of "__gcc_dump_printf__" format from
8185         >= 3005 to >= 9000.
8187 2018-08-28  Richard Biener  <rguenther@suse.de>
8189         PR tree-optimization/87124
8190         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
8191         constants before looking up avail.
8193 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
8195         PR middle-end/87099
8196         * calls.c (maybe_warn_nonstring_arg): Punt early if
8197         warn_stringop_overflow is zero.  Don't call get_range_strlen
8198         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
8199         Swap comparison operands to have constants on rhs.  Only use
8200         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
8201         increment lenrng[0].
8203 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
8205         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
8206         use of tree_to_shwi.  Remove duplicated test for the size being
8207         a whole number of bytes.
8209 2018-08-28  Richard Biener  <rguenther@suse.de>
8211         PR tree-optimization/87117
8212         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
8213         Handle removed stmt without LHS (GIMPLE_NOP).
8215 2018-08-28  Richard Biener  <rguenther@suse.de>
8217         PR tree-optimization/87117
8218         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
8219         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
8221 2018-08-28  Richard Biener  <rguenther@suse.de>
8223         PR tree-optimization/87117
8224         * tree-ssa-pre.c (compute_avail): Do not make expressions
8225         with predicated values available.
8226         (get_expr_value_id): Assert we do not run into predicated value
8227         expressions.
8229 2018-08-28  Richard Biener  <rguenther@suse.de>
8231         PR tree-optimization/87117
8232         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
8233         get virtual operands.
8234         (get_expr_operands): Handle STRING_CST like other decls.
8236 2018-08-28  Martin Liska  <mliska@suse.cz>
8238         * tree.h: Update documentation of fndecl_built_in_p
8239         functions.
8242 2018-08-27  Jeff Law  <law@redhat.com>
8243         PR tree-optimization/87110
8244         * tree-ssa-dse.c (compute_trims): Handle non-constant
8245         TYPE_SIZE_UNIT.
8247 2018-08-27  Martin Sebor  <msebor@redhat.com>
8249         PR tree-optimization/86914
8250         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
8252 2018-08-27  Martin Sebor  <msebor@redhat.com>
8254         PR tree-optimization/87112
8255         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
8256         the type of the bound argument.
8258 2018-08-27  Jeff Law  <law@redhat.com>
8260         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
8261         type does not have a TYPE_SIZE_UNIT.
8263 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
8265         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
8266         with include of backend.h.
8268 2018-08-27  Richard Biener  <rguenther@suse.de>
8270         PR tree-optimization/86927
8271         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
8272         use const cond reduction code.
8274 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
8276         PR tree-optimization/85758
8277         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
8279 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
8281         PR c++/87091
8282         * diagnostic-show-locus.c (class layout_range): Update for
8283         conversion of show_caret_p to a tri-state.
8284         (layout_range::layout_range): Likewise.
8285         (make_range): Likewise.
8286         (layout::maybe_add_location_range): Likewise.
8287         (layout::should_print_annotation_line_p): Don't show annotation
8288         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
8289         (layout::get_state_at_point): Update for conversion of
8290         show_caret_p to a tri-state.  Bail out early for
8291         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
8292         underlining or source colorization.
8293         (gcc_rich_location::add_location_if_nearby): Update for conversion
8294         of show_caret_p to a tri-state.
8295         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
8296         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
8297         Likewise.
8298         (selftest::test_one_liner_labels): Likewise.
8299         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
8300         conversion of show_caret_p to a tri-state.
8301         * pretty-print.c (text_info::set_location): Likewise.
8302         * pretty-print.h (text_info::set_location): Likewise.
8303         * substring-locations.c (format_warning_n_va): Likewise.
8304         * tree-diagnostic.c (default_tree_printer): Likewise.
8305         * tree-pretty-print.c (newline_and_indent): Likewise.
8307 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
8309         PR c++/87091
8310         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
8311         line above for line-insertion fix-it hints.
8312         (selftest::test_fixit_insert_containing_newline): Update the
8313         expected results, and add a test with line-numbering enabled.
8315 2018-08-27  Martin Liska  <mliska@suse.cz>
8317         PR sanitizer/86962
8318         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
8319         params with DECL_HAS_VALUE_EXPR_P.
8321 2018-08-27  Martin Liska  <mliska@suse.cz>
8323         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
8324         selected expansion strategy.
8326 2018-08-27  Martin Liska  <mliska@suse.cz>
8328         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
8329         * builtins.c (is_builtin_fn): Likewise.
8330         * attribs.c (diag_attr_exclusions): Use new function
8331         fndecl_built_in_p and remove check for FUNCTION_DECL if
8332         possible.
8333         (builtin_mathfn_code): Likewise.
8334         (fold_builtin_expect): Likewise.
8335         (fold_call_expr): Likewise.
8336         (fold_builtin_call_array): Likewise.
8337         (fold_call_stmt): Likewise.
8338         (set_builtin_user_assembler_name): Likewise.
8339         (is_simple_builtin): Likewise.
8340         * calls.c (gimple_alloca_call_p): Likewise.
8341         (maybe_warn_nonstring_arg): Likewise.
8342         * cfgexpand.c (expand_call_stmt): Likewise.
8343         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
8344         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
8345         (cgraph_node::verify_node): Likewise.
8346         * cgraphclones.c (build_function_decl_skip_args): Likewise.
8347         (cgraph_node::create_clone): Likewise.
8348         * config/arm/arm.c (arm_insert_attributes): Likewise.
8349         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
8350         * dse.c (scan_insn): Likewise.
8351         * expr.c (expand_expr_real_1): Likewise.
8352         * fold-const.c (operand_equal_p): Likewise.
8353         (fold_binary_loc): Likewise.
8354         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
8355         * gimple-low.c (lower_stmt): Likewise.
8356         * gimple-pretty-print.c (dump_gimple_call): Likewise.
8357         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8358         Likewise.
8359         * gimple.c (gimple_build_call_from_tree): Likewise.
8360         (gimple_call_builtin_p): Likewise.
8361         (gimple_call_combined_fn): Likewise.
8362         * gimplify.c (gimplify_call_expr): Likewise.
8363         (gimple_boolify): Likewise.
8364         (gimplify_modify_expr): Likewise.
8365         (gimplify_addr_expr): Likewise.
8366         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
8367         * ipa-cp.c (determine_versionability): Likewise.
8368         * ipa-fnsummary.c (compute_fn_summary): Likewise.
8369         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
8370         * ipa-split.c (visit_bb): Likewise.
8371         (split_function): Likewise.
8372         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
8373         * lto-cgraph.c (input_node): Likewise.
8374         * lto-streamer-out.c (write_symbol): Likewise.
8375         * omp-low.c (setjmp_or_longjmp_p): Likewise.
8376         (lower_omp_1): Likewise.
8377         * predict.c (strip_predict_hints): Likewise.
8378         * print-tree.c (print_node): Likewise.
8379         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
8380         * trans-mem.c (is_tm_irrevocable): Likewise.
8381         (is_tm_load): Likewise.
8382         (is_tm_simple_load): Likewise.
8383         (is_tm_store): Likewise.
8384         (is_tm_simple_store): Likewise.
8385         (is_tm_abort): Likewise.
8386         (tm_region_init_1): Likewise.
8387         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
8388         * tree-cfg.c (verify_gimple_call): Likewise.
8389         (move_stmt_r): Likewise.
8390         (stmt_can_terminate_bb_p): Likewise.
8391         * tree-eh.c (lower_eh_constructs_2): Likewise.
8392         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
8393         * tree-inline.c (remap_gimple_stmt): Likewise.
8394         (copy_bb): Likewise.
8395         (estimate_num_insns): Likewise.
8396         (fold_marked_statements): Likewise.
8397         * tree-sra.c (scan_function): Likewise.
8398         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
8399         (optimize_stack_restore): Likewise.
8400         (pass_fold_builtins::execute): Likewise.
8401         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
8402         (mark_all_reaching_defs_necessary_1): Likewise.
8403         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
8404         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
8405         (pass_forwprop::execute): Likewise.
8406         * tree-ssa-loop-im.c (stmt_cost): Likewise.
8407         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
8408         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
8409         * tree-ssa-strlen.c (get_string_length): Likewise.
8410         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
8411         (find_func_aliases_for_call): Likewise.
8412         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
8413         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
8414         * tree-tailcall.c (find_tail_calls): Likewise.
8415         * tree.c (need_assembler_name_p): Likewise.
8416         (free_lang_data_in_decl): Likewise.
8417         (get_call_combined_fn): Likewise.
8418         * ubsan.c (is_ubsan_builtin_p): Likewise.
8419         * varasm.c (incorporeal_function_p): Likewise.
8420         * tree.h (DECL_BUILT_IN): Remove and replace with
8421         fndecl_built_in_p.
8422         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
8423         (fndecl_built_in_p): New.
8425 2018-08-27  Martin Liska  <mliska@suse.cz>
8427         PR tree-optimization/86847
8428         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
8429         Dump also subtree probability.
8430         (switch_decision_tree::do_jump_if_equal): New function.
8431         (switch_decision_tree::emit_case_nodes): Handle special
8432         situations in balanced tree that can be emitted much simpler.
8433         Fix calculation of probabilities that happen in tree expansion.
8434         * tree-switch-conversion.h (struct cluster): Add
8435         is_single_value_p.
8436         (struct simple_cluster): Likewise.
8437         (struct case_tree_node): Add new function has_child.
8438         (do_jump_if_equal): New.
8440 2018-08-27  Martin Liska  <mliska@suse.cz>
8442         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
8443         Add new argument to bit_test_cluster constructor.
8444         (bit_test_cluster::emit): Set bits really number of values
8445         handlel by a test.
8446         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
8447         probability argument.
8448         * tree-switch-conversion.h (struct bit_test_cluster):
8449         Add m_handles_entire_switch.
8451 2018-08-27  Martin Liska  <mliska@suse.cz>
8453         PR tree-optimization/86702
8454         * tree-switch-conversion.c (jump_table_cluster::emit):
8455         Make probabilities even for values in jump table
8456         according to number of cases handled.
8457         (switch_decision_tree::compute_cases_per_edge): Pass
8458         argument to reset_out_edges_aux function.
8459         (switch_decision_tree::analyze_switch_statement): Likewise.
8460         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
8461         Make it static.
8463 2018-08-27  Martin Liska  <mliska@suse.cz>
8465         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
8466         cfun argument explicitly.
8467         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
8468         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
8469         function gimple_switch_default_bb.
8470         (convert_switch_statements):
8471         (expand_builtins):
8472         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
8473         * stmt.c (label_to_block_fn): Use label_to_block and pass
8474         cfun argument explicitly and use gimple_switch_label_bb.
8475         (expand_case): Likewise.
8476         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
8477         cfun argument explicitly. Likewise.
8478         (make_edges_bb): Likewise.
8479         (make_cond_expr_edges): Likewise.
8480         (get_cases_for_edge): Likewise.
8481         (make_gimple_switch_edges): Likewise.
8482         (label_to_block_fn): Likewise.
8483         (label_to_block): Likewise.
8484         (make_goto_expr_edges): Likewise.
8485         (make_gimple_asm_edges): Likewise.
8486         (main_block_label): Likewise.
8487         (group_case_labels_stmt): Likewise.
8488         (find_taken_edge_computed_goto): Likewise.
8489         (find_taken_edge_switch_expr): Likewise.
8490         (gimple_verify_flow_info): Likewise.
8491         (gimple_redirect_edge_and_branch): Likewise.
8492         (gimple_switch_label_bb): New function.
8493         (gimple_switch_default_bb): Likewise.
8494         (gimple_switch_edge): Likewise.
8495         (gimple_switch_default_edge): Likewise.
8496         * tree-cfg.h (label_to_block_fn): Remove and replace ...
8497         (label_to_block): ... with this.
8498         (gimple_switch_label_bb): New.
8499         (gimple_switch_default_bb): Likewise.
8500         (gimple_switch_edge): Likewise.
8501         (gimple_switch_default_edge): Likewise.
8502         * tree-cfgcleanup.c (convert_single_case_switch): Use
8503         new gimple functions and pass new argument to label_to_block.
8504         (cleanup_control_flow_bb):
8505         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
8506         cfun argument explicitly.
8507         (make_eh_edges): Likewise.
8508         (redirect_eh_dispatch_edge): Likewise.
8509         (lower_resx): Likewise.
8510         (lower_eh_dispatch): Likewise.
8511         (maybe_remove_unreachable_handlers): Likewise.
8512         (unsplit_eh): Likewise.
8513         (cleanup_empty_eh): Likewise.
8514         (verify_eh_edges): Likewise.
8515         (verify_eh_dispatch_edge): Likewise.
8516         * tree-ssa-dom.c (record_edge_info): Likewise.
8517         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
8518         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
8519         (thread_through_normal_block): Likewise.
8520         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
8521         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
8522         * tree-switch-conversion.c (switch_conversion::collect): Use new
8523         gimple functions.
8524         (switch_conversion::check_final_bb): Likewise.
8525         (switch_conversion::gather_default_values): Pass new argument
8526         to label_to_block.
8527         (switch_conversion::build_constructors): Likewise.
8528         (switch_decision_tree::compute_cases_per_edge): Use new
8529         gimple_switch_edge function.
8530         (switch_decision_tree::analyze_switch_statement): Pass new argument
8531         to label_to_block.
8532         (switch_decision_tree::try_switch_expansion): Use
8533         gimple_switch_default_edge.
8534         * tree-vrp.c (find_switch_asserts): Pass new argument
8535         to label_to_block.
8536         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
8537         (vr_values::simplify_switch_using_ranges): Likewise.
8539 2018-08-27  Richard Biener  <rguenther@suse.de>
8541         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
8542         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
8544         * tree-ssa-sccvn.h (struct vn_pval): New structure.
8545         (struct vn_nary_op_s): Add unwind_to member.  Add
8546         predicated_values flag and put result into a union together
8547         with a linked list of vn_pval.
8548         (struct vn_ssa_aux): Add name member to make maintaining
8549         a map of SSA name to vn_ssa_aux possible.  Remove no longer
8550         needed info, dfsnum, low, visited, on_sccstack, use_processed
8551         and range_info_anti_range_p members.
8552         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
8553         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
8554         New functions.
8555         (vn_valueize): New global.
8556         (vn_context_bb): Likewise.
8557         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
8558         VN_INFO_PTR_INFO): Remove.
8559         * tree-ssa-sccvn.c: ... (rewrite)
8560         (pass_fre::execute): For -O2+ initialize loops and run
8561         RPO VN in optimistic mode (iterating).  For -O1 and -Og
8562         run RPO VN in non-optimistic mode.
8563         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
8564         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
8565         * doc/invoke.texi (sccvn-max-scc-size): Remove.
8566         (rpo-vn-max-loop-depth): Document.
8567         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
8568         when valuezing the VUSE signals we walked out of the region.
8569         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
8570         (phi_translate): Set VN context block to use for availability
8571         lookup.
8572         (compute_avail): Likewise.
8573         (pre_valueize): New function.
8574         (pass_pre::execute): Adjust to the RPO VN API.
8576         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
8577         (propagate_constants_for_unrolling): Remove.
8578         (tree_unroll_loops_completely): Perform value-numbering
8579         on the unrolled bodies loop parent.
8581 2018-08-27  Richard Biener  <rguenther@suse.de>
8583         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
8584         for partial antic compute.
8586 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
8588         PR rtl-optimization/87065
8589         * combine.c (simplify_if_then_else): Formatting fix.
8590         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
8591         check.
8592         (known_cond): Don't return const_true_rtx for vector modes.  Use
8593         CONST0_RTX instead of const0_rtx.  Formatting fixes.
8595 2018-08-27  Martin Liska  <mliska@suse.cz>
8597         PR gcov-profile/87069
8598         * gcov.c (process_file): Record files already processed
8599         and warn about a file being processed multiple times.
8601 2018-08-27  Martin Liska  <mliska@suse.cz>
8603         PR driver/83193
8604         * config/aarch64/aarch64.c (aarch64_override_options_internal):
8605         Set default values for x_aarch64_*_string strings.
8606         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
8607         prefix.  For -mabi do not print '=ABI' in help and use
8608         <option_value> format for -msve-vector-bits and -moverride
8609         options.
8611 2018-08-26  Jeff Law  <law@redhat.com>
8613         * config/mips/frame-header-opt.c: Include "backend.h" rather than
8614         "cfg.h"
8616 2018-08-26  Marek Polacek  <polacek@redhat.com>
8618         PR c++/87029, Implement -Wredundant-move.
8619         * doc/invoke.texi: Document -Wredundant-move.
8621 2018-08-25  Martin Sebor  <msebor@redhat.com>
8623         PR tree-optimization/87059
8624         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
8625         to the same type as the other.
8626         * fold-const.c (fold_binary_loc): Assert expectation.
8628 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
8630         * config/darwin.c (machopic_legitimize_pic_address): Clean up
8631         extraneous parentheses, dead code section and formatting.
8633 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
8635         PR c++/87091
8636         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
8637         wide enough for jumps in the line-numbering to be visible.
8638         (layout::print_gap_in_line_numbering): New member function.
8639         (layout::calculate_line_spans): When using line numbering, merge
8640         line spans that are only 1 line apart.
8641         (diagnostic_show_locus): When printing line numbers, show gaps in
8642         line numbering directly, rather than printing headers.
8643         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
8644         line-numbering with multiple line spans.
8645         (selftest::test_fixit_insert_containing_newline_2): Add test of
8646         line-numbering, in which the spans are close enough to be merged.
8648 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8650         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
8651         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
8652         * tree-vrp.c (range_is_nonnull): Remove.
8653         (range_includes_zero_p): Accept value_range instead of min/max.
8654         (extract_range_from_binary_expr_1): Do not early bail on
8655         POINTER_PLUS_EXPR.
8656         Use range_includes_zero_p instead of range_is_nonnull.
8657         (extract_range_from_unary_expr): Use range_includes_zero_p instead
8658         of range_is_nonnull.
8659         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
8660         special case VR_ANTI_RANGE.
8661         (vrp_finalize): Same.
8662         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
8663         instead of min/max.
8664         (range_is_nonnull): Remove.
8665         * vr-values.c (vrp_stmt_computes_nonzero): Use
8666         range_includes_zero_p instead of range_is_nonnull.
8667         (extract_range_basic): Pass value_range to range_includes_zero_p
8668         instead of range_is_nonnull.
8670 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
8672         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
8673         * emit-rtl.h (rtl_data): Remove return_bnd.
8674         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
8675         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
8676         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
8677         (POINTER_BOUNDS_MODE): Remove definition.
8678         (make_pointer_bounds_mode): Remove.
8679         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
8680         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
8681         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
8682         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
8683         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
8684         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
8685         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
8687         * config/i386/i386-modes.def (BND32, BND64): Remove.
8688         * config/i386/i386.c (dbx_register_map): Remove bound registers.
8689         (dbx64_register_map): Ditto.
8690         (svr4_dbx_register_map): Ditto.
8691         (indirect_thunk_bnd_needed): Remove.
8692         (indirect_thunks_bnd_used): Ditto.
8693         (indirect_return_bnd_needed): Ditto.
8694         (indirect_return_via_cx_bnd): Ditto.
8695         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
8696         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
8697         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
8698         (output_indirect_thunk_function): Remove handling of
8699         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
8700         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
8701         (ix86_save_reg): Remove handling of crtl->return_bnd.
8702         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
8703         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
8704         and UNSPEC_BNDLX_ADDR.
8705         (ix86_output_indirect_branch_via_reg): Remove handling of
8706         indirect_thunk_prefix_bnd.
8707         (ix86_output_indirect_branch_via_push): Ditto.
8708         (ix86_output_function_return): Ditto.
8709         (ix86_output_indirect_function_return): Ditto.
8710         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
8711         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
8712         (CALL_USED_REGISTERS): Ditto.
8713         (REG_ALLOC_ORDER): Update for removal of bound registers.
8714         (HI_REGISTER_NAMES): Ditto.
8715         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
8716         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
8717         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
8718         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
8719         (FIRST_PSEUDO_REG): Update.
8720         (BND): Remove mode iterator.
8721         * config/i386/predicates.md (bnd_mem_operator): Remove.
8723 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8725         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
8726         vectors.
8728 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8730         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
8731         the case in which the permute needs only a single element and
8732         repeats for every vector of the result.  Extend that case to
8733         handle variable-length vectors.
8734         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8736 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
8738         PR debug/79342
8739         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
8740         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
8742 2018-08-24  Richard Biener  <rguenther@suse.de>
8744         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
8745         bb_flags_allocated members.
8746         (auto_flag): New RAII class for allocating flags.
8747         (auto_edge_flag): New RAII class for allocating edge flags.
8748         (auto_bb_flag): New RAII class for allocating bb flags.
8749         * cfgloop.c (verify_loop_structure): Allocate temporary edge
8750         flag dynamically.
8751         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
8752         in favor of temporarily allocated BB flag.
8753         * hsa-brig.c: Re-order includes.
8754         * hsa-dump.c: Likewise.
8755         * hsa-regalloc.c: Likewise.
8756         * print-rtl.c: Likewise.
8757         * profile-count.c: Likewise.
8759 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
8761         PR target/86989
8762         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
8763         the TOC register.
8765 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8767         PR 87073/bootstrap
8768         * wide-int-range.cc (wide_int_range_div): Do not ignore result
8769         from wide_int_range_multiplicative_op.
8771 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
8773         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
8774         "permutaion".
8776 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
8778         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
8779         to 'expanded'.
8781 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
8783         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
8784         full GENERIC expression used for replacement.
8786 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
8788         * tree-vrp.c (abs_extent_range): Remove.
8789         (extract_range_into_wide_ints): Pass wide ints by reference.
8790         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
8791         Pass wide ints by reference in all calls to
8792         extract_range_into_wide_ints.
8793         * wide-int-range.cc (wide_int_range_div): New.
8794         * wide-int-range.h (wide_int_range_div): New.
8795         (wide_int_range_includes_zero_p): New.
8796         (wide_int_range_zero_p): New.
8798 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
8800         * config/aarch64/aarch64.md (arches): New enum.
8801         (arch): New enum attr.
8802         (arch_enabled): New attr.
8803         (enabled): Now uses arch_enabled only.
8804         (simd, sve, fp16): Removed attribute.
8805         (fp): Attr now defined in terms of 'arch'.
8806         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
8807         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
8808         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
8809         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
8810         attributes into 'arch'.
8811         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
8812         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
8813         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
8814         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
8815         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
8816         'simd' attribute into 'arch'.
8817         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
8818         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
8819         Convert use of 'fp' attribute to 'arch'.
8820         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
8821         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
8822         into 'arch'.
8823         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
8824         (different modes) Merge 'fp' and 'simd' into 'arch'.
8825         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
8826         'simd' into 'arch'.
8828 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
8830         PR rtl-optimization/87026
8831         * expmed.c (canonicalize_comparison): If we can no longer create
8832         pseudoregisters, don't.
8834 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
8836         PR target/86951
8837         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
8838         prototype.
8839         * config/arm/arm.c (speculation_barrier_libfunc): New static
8840         variable.
8841         (arm_init_libfuncs): Initialize it.
8842         (arm_emit_speculation_barrier): New function.
8843         * config/arm/arm.md (speculation_barrier): Call
8844         arm_emit_speculation_barrier for architectures that do not have
8845         DSB or ISB.
8846         (speculation_barrier_insn): Only match on Armv7 or later.
8848 2018-08-23  Richard Biener  <rguenther@suse.de>
8850         PR middle-end/87024
8851         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
8852         calls.
8854 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8856         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
8857         of single-vector TBLs.
8858         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
8859         one input is given.
8861 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8863         PR target/85910
8864         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
8865         aarch64_evpc_tbl guard.
8867 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8869         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
8870         behaviour.
8872 2018-08-22  Martin Sebor  <msebor@redhat.com>
8874         PR middle-end/87052
8875         * tree-pretty-print.c (pretty_print_string): Add argument.
8876         (dump_generic_node): Call to pretty_print_string with string size.
8878 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
8880         PR rtl-optimization/86771
8881         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
8882         of two SETs into those two SETs, one to be placed at i2, if that SETs
8883         destination is modified between i2 and i3.
8885 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8887         PR tree-optimization/86725
8888         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
8889         function.
8890         (vect_analyze_scalar_cycles_1): Check it.
8892 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8894         PR tree-optimization/86725
8895         * tree-vect-loop.c (vect_is_simple_reduction): When treating
8896         an outer loop phi as a double reduction, make sure that the
8897         single user of the phi result is an inner loop phi.
8899 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8901         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
8902         grouped stores with gaps to a strided group.
8904 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8906         * tree-vect-stmts.c (get_group_load_store_type)
8907         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
8908         first statement in a group.
8910 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8912         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
8913         the sequence used in gcc/gcc.c.
8915 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8917         PR other/704
8918         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
8919         building it.
8921 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8923         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
8924         Darwin10-specific unwinder-shim.
8925         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
8926         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
8927         New to cater for Darwin10 Rosetta.
8929 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8931         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
8932         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
8934 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8936         PR bootstrap/81033
8937         PR target/81733
8938         PR target/52795
8939         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
8940         (dwarf2out_switch_text_section): Generate a local label for the
8941         second function sub-section and apply it as the second FDE start
8942         label.
8943         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
8944         second sub-section start.
8946 2018-08-22  Richard Biener  <rguenther@suse.de>
8948         PR tree-optimization/86988
8949         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
8951 2018-08-22  Richard Biener  <rguenther@suse.de>
8953         PR tree-optimization/86945
8954         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
8956 2018-08-22  Alexandre Oliva <oliva@adacore.com>
8958         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
8959         a comment about how uses of r2 for .sdata2 come about.
8961 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
8963         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
8965 2018-08-21  Marek Polacek  <polacek@redhat.com>
8967         PR c++/86981, Implement -Wpessimizing-move.
8968         * doc/invoke.texi: Document -Wpessimizing-move.
8970 2018-08-21  Jan Hubicka  <jh@suse.cz>
8972         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
8973         * tree.h (is_redundant_typedef): Remove.
8974         * dwarf2out.c (is_redundant_typedef): Turn into static function.
8976 2018-08-21  Jan Hubicka  <jh@suse.cz>
8978         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
8979         when possible.
8981 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
8983         * expmed.c (extract_low_bits): Reject invalid subregs early.
8985 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8987         PR middle-end/86121
8988         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
8989         behaviour.
8991 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8993         * config/vxworks.h: Guard vxworks_asm_out_constructor and
8994         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
8995         * config/vxworks.c: Likewise.
8997 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8999         * config/vxworks.c: Set targetm.have_ctors_dtors
9000         if HAVE_INITFINI_ARRAY_SUPPORT.
9001         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
9002         if HAVE_INITFINI_ARRAY_SUPPORT.
9004 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
9006         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
9007         default search path for VxWorks < 7.
9009 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9011         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
9012         (get_format_string): Refer to c_getstr.
9014 2018-08-21  Tom de Vries  <tdevries@suse.de>
9016         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
9017         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
9018         (debuginfo_early_stop): Declare.
9019         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
9020         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
9021         (debuginfo_early_stop): New function.
9022         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
9023         and debuginfo_early_stop.
9024         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
9025         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
9026         (general_init): Call debuginfo_early_init.
9027         (finalize): Call debuginfo_fini.
9028         (do_compile): Call debuginfo_init.
9029         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
9030         -fdump-early-debug.
9031         (@item -fdump-debug, @item -fdump-earlydebug): Add.
9033 2018-08-21  Tom de Vries  <tdevries@suse.de>
9035         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
9036         flag_dump_noaddr and flag_dump_unnumbered.
9038 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
9040         * wide-int-range.cc (wide_int_range_abs): New.
9041         (wide_int_range_order_set): Rename from wide_int_range_min_max.
9042         * wide-int-range.h (wide_int_range_abs): New.
9043         (wide_int_range_min_max): New.
9044         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
9045         case to call wide_int_range_abs.
9046         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
9047         (extract_range_from_abs_expr): Delete.
9049 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
9051         PR target/87033
9052         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
9053         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
9054         for indexed loads.
9056 2018-08-20  Nathan Sidwell  <nathan@acm.org>
9057             Jeff Law <law@redhat.com>
9059         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
9060         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
9062 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
9064         PR other/84889
9065         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
9066         (decl_attributes): Likewise.
9067         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
9068         instance.
9069         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9070         * diagnostic-core.h (class auto_diagnostic_group): New class.
9071         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
9072         (diagnostic_report_diagnostic): Handle the first diagnostics within
9073         a group.
9074         (emit_diagnostic): Add auto_diagnostic_group instance.
9075         (inform): Likewise.
9076         (inform_n): Likewise.
9077         (warning): Likewise.
9078         (warning_at): Likewise.
9079         (warning_n): Likewise.
9080         (pedwarn): Likewise.
9081         (permerror): Likewise.
9082         (error): Likewise.
9083         (error_n): Likewise.
9084         (error_at): Likewise.
9085         (sorry): Likewise.
9086         (fatal_error): Likewise.
9087         (internal_error): Likewise.
9088         (internal_error_no_backtrace): Likewise.
9089         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
9090         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
9091         * diagnostic.h (struct diagnostic_context): Add fields
9092         "diagnostic_group_nesting_depth",
9093         "diagnostic_group_emission_count", "begin_group_cb",
9094         "end_group_cb".
9095         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
9096         Add auto_diagnostic_group instance(s).
9097         (find_explicit_erroneous_behavior): Likewise.
9098         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
9099         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
9100         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
9101         (gimplify_va_arg_expr): Likewise.
9102         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
9103         (HSA_SORRY_AT): Likewise.
9104         * ipa-devirt.c (compare_virtual_tables): Likewise.
9105         (warn_odr): Likewise.
9106         * multiple_target.c (expand_target_clones): Likewise.
9107         * opts-common.c (cmdline_handle_error): Likewise.
9108         * reginfo.c (globalize_reg): Likewise.
9109         * substring-locations.c (format_warning_n_va): Likewise.
9110         * tree-inline.c (expand_call_inline): Likewise.
9111         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
9112         * tree-ssa-loop-niter.c
9113         (do_warn_aggressive_loop_optimizations): Likewise.
9114         * tree-ssa-uninit.c (warn_uninit): Likewise.
9115         * tree.c (warn_deprecated_use): Likewise.
9117 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
9119         PR target/87014
9120         * config/i386/i386.md (eh_return): Always update EH return
9121         address in word_mode.
9123 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
9125         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
9126         TARGET_SPLIT_COMPLEX_ARG is defined.
9128 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9130         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
9132 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9134         PR target/86984
9135         * expr.c (expand_assignment): Assert that bitpos is positive.
9136         (store_field): Likewise
9137         (expand_expr_real_1): Make sure that bitpos is positive.
9138         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
9139         integer overflow.
9141 2018-08-20  Nathan Sidwell  <nathan@acm.org>
9143         * Makefile.in (CPP_ID_DATA_H): Delete.
9144         (CPP_INTERNAL_H): Don't add it.
9145         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
9146         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
9148 2018-08-20  Richard Biener  <rguenther@suse.de>
9150         PR tree-optimization/78655
9151         * tree-vrp.c (extract_range_from_binary_expr_1): Make
9152         pointer + offset nonnull if either operand is nonnull work.
9154 2018-08-20  Tom de Vries  <tdevries@suse.de>
9156         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
9157         unless the referenced die describes the added property using
9158         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
9159         Otherwise, add a DW_AT_location to the referenced die.
9161 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
9163         PR target/86994
9164         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
9165         register_operand when calling ix86_set_reg_reg_cost.
9166         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
9167         Set *total to 0 for operands that satisfy x86_64_immediate_operand
9168         predicate and to 1 otherwise.
9170 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
9172         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
9173         emit a diagnostic that it is not supported and reset the option.
9174         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
9175         supported and consume the option.  (ASM_FINAL_SPEC): New.
9177 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
9179         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
9180         a sentence.
9182 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
9184         C-SKY port: Documentation
9186         * doc/extend.texi (C-SKY Function Attributes): New section.
9187         * doc/invoke.texi (Option Summary): Add C-SKY options.
9188         (C-SKY Options): New section.
9189         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
9191 2018-08-17  Jojo  <jijie_rong@c-sky.com>
9192             Huibin Wang  <huibin_wang@c-sky.com>
9193             Sandra Loosemore  <sandra@codesourcery.com>
9194             Chung-Lin Tang  <cltang@codesourcery.com>
9196         C-SKY port: Backend implementation
9198         * config/csky/*: New.
9199         * common/config/csky/*: New.
9201 2018-08-17  Jojo  <jijie_rong@c-sky.com>
9202             Huibin Wang  <huibin_wang@c-sky.com>
9203             Sandra Loosemore  <sandra@codesourcery.com>
9204             Chung-Lin Tang  <cltang@codesourcery.com>
9205             Andrew Jenner  <andrew@codesourcery.com>
9207         C-SKY port: Configury
9209         * config.gcc (csky-*-*): New.
9210         * configure.ac: Add csky to targets for dwarf2 debug_line support.
9211         * configure: Regenerated.
9213 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
9215         * dump-context.h: Include "dumpfile.h".
9216         (dump_context::dump_printf_va): Convert final param from va_list
9217         to va_list *.  Convert from ATTRIBUTE_PRINTF to
9218         ATTRIBUTE_GCC_DUMP_PRINTF.
9219         (dump_context::dump_printf_loc_va): Likewise.
9220         * dumpfile.c: Include "stringpool.h".
9221         (make_item_for_dump_printf_va): Delete.
9222         (make_item_for_dump_printf): Delete.
9223         (class dump_pretty_printer): New class.
9224         (dump_pretty_printer::dump_pretty_printer): New ctor.
9225         (dump_pretty_printer::emit_items): New member function.
9226         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
9227         function.
9228         (dump_pretty_printer::emit_item): New member function.
9229         (dump_pretty_printer::stash_item): New member function.
9230         (dump_pretty_printer::format_decoder_cb): New member function.
9231         (dump_pretty_printer::decode_format): New member function.
9232         (dump_context::dump_printf_va): Reimplement in terms of
9233         dump_pretty_printer.
9234         (dump_context::dump_printf_loc_va): Convert final param from va_list
9235         to va_list *.
9236         (dump_context::begin_scope): Reimplement call to
9237         make_item_for_dump_printf.
9238         (dump_printf): Update for change to dump_printf_va.
9239         (dump_printf_loc): Likewise.
9240         (selftest::test_capture_of_dump_calls): Convert "stmt" from
9241         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
9242         with %T, %E, and %G.
9243         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
9244         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
9245         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
9246         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
9247         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
9248         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
9249         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
9250         within a dump_printf_loc call to "%wu".
9251         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
9252         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
9253         missing space after "=".
9254         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
9255         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
9256         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
9257         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
9258         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
9259         duplicate "vectorized" from message.
9261 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
9263         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
9264         polyNxK_t element's TYPE_STRING_FLAG.
9266 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
9268         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
9269         (they were unnamed before).  Fix comments.
9271 2018-08-17  Nathan Sidwell  <nathan@acm.org>
9273         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
9275 2018-08-17  Richard Biener  <rguenther@suse.de>
9277         PR tree-optimization/86841
9278         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
9280 2018-08-17  Martin Liska  <mliska@suse.cz>
9282         * common.opt: Remove Warn, Init and Report for options with
9283         Ignore/Deprecated flag. Warning is done automatically for
9284         Deprecated flags.
9285         * config/i386/i386.opt: Likewise.
9286         * config/ia64/ia64.opt: Likewise.
9287         * config/rs6000/rs6000.opt: Likewise.
9288         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
9289         Remove usage of flag_check_pointer_bounds.
9290         * lto-wrapper.c (merge_and_complain): Do not handle
9291         OPT_fcheck_pointer_bounds.
9292         (append_compiler_options): Likewise.
9293         * opt-functions.awk: Do not handle Deprecated.
9294         * optc-gen.awk: Check that Var, Report and Init are not
9295         used for an option with Ignore/Deprecated flag.
9296         * opts-common.c (decode_cmdline_option): Do not report
9297         CL_ERR_DEPRECATED.
9298         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
9299         options.
9300         * opts.h (struct cl_option): Remove cl_deprecated flag.
9301         (CL_ERR_DEPRECATED): Remove error enum value.
9303 2018-08-17  Richard Biener  <rguenther@suse.de>
9305         PR middle-end/86505
9306         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
9307         across a va-arg-pack using call adjust its return value accordingly.
9309 2018-08-16  Martin Sebor  <msebor@redhat.com>
9311         PR tree-optimization/86853
9312         * gimple-ssa-sprintf.c (struct format_result): Rename member.
9313         (struct fmtresult): Add member and initialize it in ctors.
9314         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
9315         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
9316         (format_directive): Set POSUNDER4K when MAYFAIL is set.
9317         (parse_directive): Handle %C same as %c.
9318         (sprintf_dom_walker::compute_format_length): Adjust.
9319         (is_call_safe): Adjust.
9321 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9323         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
9324         for determining how to count the elements.
9325         * builtins.h (c_strlen): Adjust prototype.
9326         * expr.c (string_constant): Add new parameter mem_size.
9327         Set *mem_size appropriately.
9328         * expr.h (string_constant): Adjust protoype.
9329         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
9330         * gimple-fold.h (get_range_strlen): Adjust prototype.
9331         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
9332         (format_string): Call get_string_length with eltsize.
9334 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9336         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
9337         to emit the span, rather than setting it as the prefix.
9339 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9341         * diagnostic-show-locus.c (layout::start_annotation_line): Add
9342         "margin_char" parameter, defaulting to space.  Use it in place
9343         of pp_space for the initial part of the margin.
9344         (layout::print_leading_fixits): Use '+' when filling the margin
9345         of line-insertion fix-it hints.
9347 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9349         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
9350         Delete.
9352 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9354         * config/rs6000/altivec.md: Don't set length attribute to the default
9355         value.
9356         * config/rs6000/darwin.md: Ditto.
9357         * config/rs6000/dfp.md: Ditto.
9358         * config/rs6000/htm.md: Ditto.
9359         * config/rs6000/rs6000.md: Ditto.
9360         * config/rs6000/sync.md: Ditto.
9361         * config/rs6000/vsx.md: Ditto.
9363 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9365         * config/rs6000/altivec.md: Don't set length attribute to the default
9366         value, for branch instructions.
9367         * config/rs6000/darwin.md: Ditto.
9368         * config/rs6000/rs6000.md: Ditto.
9370 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9372         * config/rs6000/rs6000.md (length): Always define as const_int 4.
9373         (unnamed conditional branch define_insn): Set length to 4 or 8
9374         depending on offset.
9375         (<bd>_<mode>): Similar, for alternative 0.
9376         (<bd>tf_<mode>): Ditto.
9378 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9380         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
9382 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
9384         * doc/rtl.texi: Replace old RTX class names with new names.
9387 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
9389         * expmed.h (canonicalize_comparison): New declaration.
9390         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
9391         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
9392         * optabs.c (prepare_cmp_insn): Likewise.
9393         * rtl.h (unsigned_condition_p): New function which checks if a
9394         comparison operator is unsigned.
9396 2018-08-16  Nathan Sidwell  <nathan@acm.org>
9398         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
9399         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
9401 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9403         PR target/84711
9404         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
9405         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
9406         (mov<mov>): ..this and enable unconditionally.
9408 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9410         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
9412 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
9414         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
9415         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
9416         "Common" with "Target".
9418 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
9420         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
9421         * doc/invoke.texi (mmitigate-rop): Remove.
9422         * config/i386/i386.c: Do not include "regrename.h".
9423         (ix86_rop_should_change_byte_p, reg_encoded_number)
9424         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
9425         Remove.
9426         (ix86_reorg): Remove call to ix86_mitigate_rop.
9427         * config/i386/i386.md (attr "modrm_class"): Remove.
9428         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
9429         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
9430         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
9432 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
9434         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
9435         allow folding of mergeh() and mergel() for the float and double types.
9436         (fold_mergehl_helper): Rework to handle building a permute tree
9437         for float vectors.
9439 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
9441         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
9442         for TARGET_SSE.
9444 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
9446         * common.opt (fdiagnostics-show-labels): New option.
9447         * diagnostic-show-locus.c (class layout_range): Add field
9448         "m_label".
9449         (class layout): Add field "m_show_labels_p".
9450         (layout_range::layout_range): Add param "label" and use it to
9451         initialize m_label.
9452         (make_range): Pass in NULL for new "label" param of layout_range's
9453         ctor.
9454         (layout::layout): Initialize m_show_labels_p.
9455         (layout::maybe_add_location_range): Pass in loc_range->m_label
9456         when constructing layout_range instances.
9457         (struct line_label): New struct.
9458         (layout::print_any_labels): New member function.
9459         (layout::print_line): Call it if label-printing is enabled.
9460         (selftest::test_one_liner_labels): New test.
9461         (selftest::test_diagnostic_show_locus_one_liner): Call it.
9462         * diagnostic.c (diagnostic_initialize): Initialize
9463         context->show_labels_p.
9464         * diagnostic.h (struct diagnostic_context): Add field
9465         "show_labels_p".
9466         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9467         -fno-diagnostics-show-labels.
9468         * dwarf2out.c (gen_producer_string): Add
9469         OPT_fdiagnostics_show_labels to the ignored options.
9470         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
9471         param.
9472         (gcc_rich_location::maybe_add_expr): Likewise.
9473         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
9474         label" param, defaulting to NULL.
9475         (gcc_rich_location::add_expr): Add "label" param.
9476         (gcc_rich_location::maybe_add_expr): Likewise.
9477         (class text_range_label): New class.
9478         (class range_label_for_type_mismatch): New class.
9479         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
9480         of format_warning_va.
9481         (fmtwarn_n): Likewise for new params of format_warning_n_va.
9482         * lto-wrapper.c (merge_and_complain): Add
9483         OPT_fdiagnostics_show_labels to the "pick one setting" options.
9484         (append_compiler_options): Likewise to the dropped options.
9485         (append_diag_options): Likewise to the passed-on options.
9486         * opts.c (common_handle_option): Handle the new option.
9487         * selftest-diagnostic.c
9488         (test_diagnostic_context::test_diagnostic_context): Enable
9489         show_labels_p.
9490         * substring-locations.c: Include "gcc-rich-location.h".
9491         (format_warning_n_va): Add "fmt_label" and "param_label" params
9492         and use them as appropriate.
9493         (format_warning_va): Add "fmt_label" and "param_label" params,
9494         passing them on to format_warning_n_va.
9495         (format_warning_at_substring): Likewise.
9496         (format_warning_at_substring_n): Likewise.
9497         * substring-locations.h (format_warning_va): Add "fmt_label" and
9498         "param_label" params.
9499         (format_warning_n_va): Likewise.
9500         (format_warning_at_substring): Likewise.
9501         (format_warning_at_substring_n): Likewise.
9502         * toplev.c (general_init): Initialize global_dc->show_labels_p.
9504 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
9506         PR testsuite/86519
9507         * builtins.c (expand_builtin_memcmp): Do not expand the call
9508         when overflow is detected.
9510 2018-08-15  Martin Sebor  <msebor@redhat.com>
9512         PR tree-optimization/71625
9513         * config/aarch64/aarch64-builtins.c
9514         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
9516 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
9518         * config/s390/s390.c (s390_reorg): Remove loop.
9520 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9522         * config/darwin.c
9523          (darwin_function_switched_text_sections): Delete.
9524         * gcc/config/darwin.h
9525          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
9527 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9529         PR target/81685
9530         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
9531         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
9532         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
9534 2018-08-15  Martin Liska  <mliska@suse.cz>
9536         PR tree-optimization/86925
9537         * predict.c (expr_expected_value_1): When taking
9538         later predictor, assign also probability.
9539         Use fold_build2_initializer_loc in order to fold
9540         the expression in -frounding-math.
9542 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
9544         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
9545         patterns.
9546         (expand_vec_perm_1): Try the new method.
9548 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
9550         PR target/86547
9551         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
9552         Check whether lra_live_max_point is 0 before dividing.
9554 2018-08-14  Martin Sebor  <msebor@redhat.com>
9556         PR tree-optimization/86650
9557         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
9558         (vrp_prop::check_mem_ref): Same.
9560 2018-08-13  Liu Hao <lh_mouse@126.com>
9562         * pretty-print.c (eat_esc_sequence): Swap the foreground and
9563         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
9564         and clear it thereafter, as it only works for DBCS.
9566 2018-08-13  Liu Hao <lh_mouse@126.com>
9568         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
9569         handle returned by _get_osf_handle().
9571 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9573         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
9574         for folding vec_perm.
9576 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9578         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
9579         Add support for gimple-folding of vec_pack() and vec_unpack()
9580         intrinsics.
9582 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
9584         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
9585         vec_xst variants to the list.
9586         (rs6000_gimple_fold_builtin): Add support for folding unaligned
9587         vector loads and stores.
9589 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
9591         * config.gcc (rs6000-ibm-aix4.x): Delete.
9592         (rs6000-ibm-aix5.1): Delete.
9593         (rs6000-ibm-aix5.2): Delete.
9594         (rs6000-ibm-aix5.3): Delete.
9595         * config/rs6000/aix43.h: Delete.
9596         * config/rs6000/aix51.h: Delete.
9597         * config/rs6000/aix52.h: Delete.
9598         * config/rs6000/t-aix43: Delete.
9600 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
9602         * config/s390/s390.c (s390_decompose_constant_pool_ref):
9603         New function.
9604         (s390_decompose_address): Factor out constant pool ref
9605         decomposition.
9607 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9609         * config/nds32/nds32-predicates.c
9610         (nds32_can_use_bclr_p): Change return type as bool.
9611         (nds32_can_use_bset_p): Ditto.
9612         (nds32_can_use_btgl_p): Ditto.
9613         (nds32_can_use_bitci_p): Ditto.
9614         * config/nds32/nds32-protos.h
9615         (nds32_can_use_bclr_p): Change declaration.
9616         (nds32_can_use_bset_p): Ditto.
9617         (nds32_can_use_btgl_p): Ditto.
9618         (nds32_can_use_bitci_p): Ditto.
9620 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9622         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
9623         Support -msched-prolog-epilog option.
9624         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
9626 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9628         * common/config/nds32/nds32-common.c
9629         (nds32_option_optimization_table): Enalbe -malways-align.
9631 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9633         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
9634         extra_headers.
9635         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
9636         OPT_misr_secure_ case.
9637         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
9638         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
9639         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
9640         secure attribute.
9641         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
9642         (nds32_isr_info): New field security_level.
9643         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
9644         * config/nds32/nds32.md (return_internal): Consider critical attribute.
9645         * config/nds32/nds32.opt (misr-secure): New option.
9646         * config/nds32/nds32_init.inc: New file.
9647         * config/nds32/nds32_isr.h: New file.
9649 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
9651         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
9652         Update comment for atomic instructions.
9653         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
9654         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
9655         Remove.
9656         (atomic_loaddi): Revise fence expansion to only emit fence prior to
9657         load for __ATOMIC_SEQ_CST model.
9658         (atomic_loaddi_1): Remove float register target.
9659         (atomic_storedi): Handle CONST_INT values.
9660         (atomic_storedi_1): Remove float register source.  Add special case
9661         for zero value.
9662         (memory_barrier): New expander and insn.
9664 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
9666         PR tree-optimization/86835
9667         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
9668         new_stmt after def_gsi, make sure to insert new_square_stmt after
9669         that stmt, not 2 stmts before it.
9671 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
9673         PR target/82418
9674         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
9675         instead of SWI48.
9677 2018-08-10  Martin Liska  <mliska@suse.cz>
9679         PR target/83610
9680         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
9681         function type.
9682         * builtins.c (expand_builtin_expect_with_probability):
9683         New function.
9684         (expand_builtin_expect_with_probability): New function.
9685         (build_builtin_expect_predicate): Add new argumnet probability
9686         for BUILT_IN_EXPECT_WITH_PROBABILITY.
9687         (fold_builtin_expect):
9688         (fold_builtin_2):
9689         (fold_builtin_3):
9690         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
9691         * builtins.h (fold_builtin_expect): Set new argument.
9692         * doc/extend.texi: Document __builtin_expect_with_probability.
9693         * doc/invoke.texi: Likewise.
9694         * gimple-fold.c (gimple_fold_call): Pass new argument.
9695         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
9696         also BUILT_IN_EXPECT_WITH_PROBABILITY.
9697         * predict.c (get_predictor_value): New function.
9698         (expr_expected_value): Add new argument probability. Assume
9699         that predictor and probability are always non-null.
9700         (expr_expected_value_1): Likewise.  For __builtin_expect and
9701         __builtin_expect_with_probability set probability.  Handle
9702         combination in binary expressions.
9703         (tree_predict_by_opcode): Simplify code by simply calling
9704         get_predictor_value.
9705         (pass_strip_predict_hints::execute): Add handling of
9706         BUILT_IN_EXPECT_WITH_PROBABILITY.
9707         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
9708         new predictor.
9709         * tree.h (DECL_BUILT_IN_P): New function.
9711 2018-08-10  Martin Liska  <mliska@suse.cz>
9713         PR tree-optimization/85799
9714         * passes.def: Add argument for pass_strip_predict_hints.
9715         * predict.c (class pass_strip_predict_hints): Add new argument
9716         early_p.
9717         (strip_predictor_early): New function.
9718         (pass_strip_predict_hints::execute): Call the function to
9719         strip predictors.
9720         (strip_predict_hints): New function.
9721         * predict.def: Fix comment.
9723 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
9725         * Makefile.in: Clarify which tm.texi to copy over to assert the
9726         right to grant a GFDL license for all.
9728 2018-08-09  Jeff Law  <law@redhat.com>
9730         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
9731         unused variable.
9733 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
9735         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
9736         prototype.
9738 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9740         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
9741         reductions for variable-length vectors.
9743 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
9745         PR other/84889
9746         * common.opt (fdiagnostics-show-line-numbers): New option.
9747         * diagnostic-show-locus.c (class layout): Add fields
9748         "m_show_line_numbers_p" and "m_linenum_width";
9749         (num_digits): New function.
9750         (test_num_digits): New function.
9751         (layout::layout): Initialize new fields.  Update m_x_offset
9752         logic to handle any left margin.
9753         (layout::print_source_line): Print line number when requested.
9754         (layout::start_annotation_line): New member function.
9755         (layout::print_annotation_line): Call it.
9756         (layout::print_leading_fixits): Likewise.
9757         (layout::print_trailing_fixits): Likewise.  Update calls to
9758         move_to_column for new parameter.
9759         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
9760         it to potentially call start_annotation_line.
9761         (layout::show_ruler): Call start_annotation_line.
9762         (selftest::test_line_numbers_multiline_range): New selftest.
9763         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
9764         and selftest::test_line_numbers_multiline_range.
9765         * diagnostic.c (diagnostic_initialize): Initialize
9766         show_line_numbers_p.
9767         * diagnostic.h (struct diagnostic_context): Add field
9768         "show_line_numbers_p".
9769         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9770         -fno-diagnostics-show-line-numbers.
9771         * dwarf2out.c (gen_producer_string): Add
9772         OPT_fdiagnostics_show_line_numbers to the ignored options.
9773         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
9774         one setting" options.
9775         (append_compiler_options): Likewise to the dropped options.
9776         (append_diag_options): Likewise to the passed-on options.
9777         * opts.c (common_handle_option): Handle the new option.
9778         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
9780 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9782         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
9783         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
9784         third argument of type "const signed char" to descriptions of
9785         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
9786         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
9787         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
9788         __builtin_bcdsub_ov functions.
9790 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9792         PR tree-optimization/86858
9793         * tree-vect-loop.c (vect_is_simple_reduction): Restore
9794         flow_bb_inside_loop_p calls.
9796 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9798         PR tree-optimization/86871
9799         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
9800         instead of gimple_assign_lhs.
9802 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
9804         PR target/86887
9805         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
9806         register constraint to operand 0.
9807         (add<mode>3_carryinC): Likewise.
9808         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
9810 2018-08-09  Martin Liska  <mliska@suse.cz>
9812         PR c/86895
9813         * common.opt: Remove extra line.
9815 2018-08-09  Martin Liska  <mliska@suse.cz>
9817         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
9818         at the end of a line, make first letter capital and end up
9819         a sentence with a dot.
9820         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
9821         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
9822         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
9823         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
9824         (PARAM_MAX_ISL_OPERATIONS): Likewise.
9825         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
9826         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
9827         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
9828         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
9829         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
9830         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
9831         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
9832         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
9833         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
9834         (PARAM_TREE_REASSOC_WIDTH): Likewise.
9835         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
9836         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
9837         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
9839 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
9841         PR target/84332
9842         * config/s390/s390.c (s390_option_override_internal): Reduce the
9843         stack-clash-protection-probe-interval param if it would be too big
9844         for z900.
9846 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
9848         PR target/46179
9849         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
9850         * config/m68k/m68k.c (handle_move_double): Don't call
9851         m68k_final_prescan_insn.
9852         (m68k_adjust_decorated_operand): Renamed from
9853         m68k_final_prescan_insn, remove first and third operand and
9854         simplify.
9855         (print_operand): Call it.
9856         (print_operand_address): Call it.
9858 2018-08-08  Nathan Sidwell  <nathan@acm.org>
9860         * diagnostic.c (diagnostic_report_current_module): Use
9861         linemap_included_from & linemap_included_from_linemap.
9863 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
9865         * config/aarch64/aarch64-cores.def: Add phecda core.
9866         * config/aarch64/aarch64-tune.md: Regenerate.
9867         * doc/invoke.texi: Add phecda core.
9869 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9871         PR target/85295
9872         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
9873         definitions.
9874         * config/s390/s390.md ("movti"): Add more alternatives for
9875         constant to GPR copies.
9877 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9879         * config/s390/s390.c: Fix whitespace damage throughout the file.
9880         * config/s390/s390.h: Likewise.
9881         * config/s390/tpf.h: Likewise.
9883 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9885         * config/s390/s390.c (s390_loadrelative_operand_p):
9886         Remove TARGET_CPU_ZARCH usages.
9887         (s390_rtx_costs): Likewise.
9888         (s390_legitimate_constant_p): Likewise.
9889         (s390_cannot_force_const_mem): Likewise.
9890         (legitimate_reload_constant_p): Likewise.
9891         (s390_preferred_reload_class): Likewise.
9892         (legitimize_pic_address): Likewise.
9893         (legitimize_tls_address): Likewise.
9894         (s390_split_branches): Removed.
9895         (s390_add_execute): Removed.
9896         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
9897         (s390_mainpool_start): Likewise.
9898         (s390_mainpool_finish): Likewise.
9899         (s390_mainpool_cancel): Removed.
9900         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
9901         (s390_chunkify_cancel): Likewise.
9902         (s390_return_addr_rtx): Likewise.
9903         (s390_register_info): Remove split_branches_pending_p uages.
9904         (s390_optimize_register_info): Likewise.
9905         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
9906         split_branches_pending_p usages.
9907         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
9908         (s390_load_got): Likewise.
9909         (s390_expand_split_stack_prologue): Likewise.
9910         (output_asm_nops): Likewise.
9911         (s390_function_profiler): Likewise.
9912         (s390_emit_call): Likewise.
9913         (s390_conditional_register_usage): Likewise.
9914         (s390_optimize_prologue): Likewise.
9915         (s390_reorg): Remove TARGET_CPU_ZARCH and
9916         split_branches_pending_p usages.
9917         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
9918         usages.
9919         (s390_output_indirect_thunk_function): Likewise.
9920         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
9921         (TARGET_CPU_ZARCH_P): Removed.
9922         (struct machine_function): Remove split_branches_pending_p.
9923         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
9925 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9927         * common/config/s390/s390-common.c (processor_flags_table):
9928         Remove flags.
9929         * config.gcc: Remove with_arch/with_tune support.
9930         * config/s390/2064.md: Remove cpu attribute comparisons.
9931         * config/s390/driver-native.c (s390_host_detect_local_cpu):
9932         Remove MTN.
9933         * config/s390/linux.h (ASM_SPEC):
9934         Remove -march support.
9935         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
9936         Use a table to get an arch level.
9937         * config/s390/s390-opts.h (enum processor_type):
9938         Remove enum values.
9939         * config/s390/s390.c
9940         (processor_table): Remove entries, add arch_level values.
9941         (s390_issue_rate): Remove cases.
9942         (s390_option_override): Adjust
9943         s390_option_override_internal() call.
9944         (s390_option_override_internal): Remove deprecation warning.
9945         (s390_valid_target_attribute_tree): Adjust
9946         s390_option_override_internal() call.
9947         * config/s390/s390.h (struct s390_processor):
9948         Share with s390-c.c, add arch_level field.
9949         * config/s390/s390.md:
9950         Remove occurrences in cpu attribute.
9951         * config/s390/s390.opt: Remove -march/-mtune support.
9952         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
9953         * doc/invoke.texi: Remove deprecation warning.
9955 2018-08-08  Luis Machado  <luis.machado@linaro.org>
9957         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
9958         global.
9959         (qdf24xx_tunings): Set vector cost structure to
9960         qdf24xx_vector_cost.
9962         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
9963         <register_sextend>: Set to 3.
9965 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
9967         PR target/86838
9968         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
9969         * config/aarch64/aarch64-simd.md
9970         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
9971         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
9972         (aarch64_frecpx<mode>): ...this new pattern.
9973         * config/aarch64/aarch64-simd-builtins.def: Remove comment
9974         about aarch64_frecp<FRECP:frecp_suffix><mode>.
9976 2018-08-07  Martin Liska  <mliska@suse.cz>
9978         PR middle-end/83023
9979         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
9980         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
9981         * predict.def (PRED_MALLOC_NONNULL): New predictor.
9982         * doc/extend.texi: Document that malloc attribute adds
9983         hit to compiler.
9985 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
9987         PR target/86785
9988         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9989         Define to speculation_safe_value_not_needed.
9991 2018-08-06  Jeff Law  <law@redhat.com>
9993         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
9994         the vr_values instance to cprop_into_stmt.
9995         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
9996         (cprop_operand): Also query EVRP to determine if OP is a constant.
9998 2018-08-06  Nathan Sidwell  <nathan@acm.org>
10000         * diagnostic.c (diagnostic_report_current_module): Reroll
10001         included-at loop.  Translate text.
10003 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
10005         * function-tests.c (selftest::test_expansion_to_rtl): Call
10006         free_after_compilation.
10008 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10010         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
10012 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
10014         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
10015         loops with memory block operations from getting unrolled.
10017 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
10019         PR target/86807
10020         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10021         Define to speculation_safe_value_not_needed.
10023 2018-08-06  Jeff Law  <law@redhat.com>
10025         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
10026         assert.
10028 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10030         PR target/86662
10031         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
10032         with all enabled __intN types.
10034         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
10036 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10038         * alias.c (record_set): Check for clobber high.
10039         * cfgexpand.c (expand_gimple_stmt): Likewise.
10040         * combine-stack-adj.c (single_set_for_csa): Likewise.
10041         * combine.c (find_single_use_1): Likewise.
10042         (set_nonzero_bits_and_sign_copies): Likewise.
10043         (get_combine_src_dest): Likewise.
10044         (is_parallel_of_n_reg_sets): Likewise.
10045         (try_combine): Likewise.
10046         (record_dead_and_set_regs_1): Likewise.
10047         (reg_dead_at_p_1): Likewise.
10048         (reg_dead_at_p): Likewise.
10049         * dce.c (deletable_insn_p): Likewise.
10050         (mark_nonreg_stores_1): Likewise.
10051         (mark_nonreg_stores_2): Likewise.
10052         * df-scan.c (df_find_hard_reg_defs): Likewise.
10053         (df_uses_record): Likewise.
10054         (df_get_call_refs): Likewise.
10055         * dwarf2out.c (mem_loc_descriptor): Likewise.
10056         * haifa-sched.c (haifa_classify_rtx): Likewise.
10057         * ira-build.c (create_insn_allocnos): Likewise.
10058         * ira-costs.c (scan_one_insn): Likewise.
10059         * ira.c (equiv_init_movable_p): Likewise.
10060         (rtx_moveable_p): Likewise.
10061         (interesting_dest_for_shprep): Likewise.
10062         * jump.c (mark_jump_label_1): Likewise.
10063         * postreload-gcse.c (record_opr_changes): Likewise.
10064         * postreload.c (reload_cse_simplify): Likewise.
10065         (struct reg_use): Add source expr.
10066         (reload_combine): Check for clobber high.
10067         (reload_combine_note_use): Likewise.
10068         (reload_cse_move2add): Likewise.
10069         (move2add_note_store): Likewise.
10070         * print-rtl.c (print_pattern): Likewise.
10071         * recog.c (decode_asm_operands): Likewise.
10072         (store_data_bypass_p): Likewise.
10073         (if_test_bypass_p): Likewise.
10074         * regcprop.c (kill_clobbered_value): Likewise.
10075         (kill_set_value): Likewise.
10076         * reginfo.c (reg_scan_mark_refs): Likewise.
10077         * reload1.c (maybe_fix_stack_asms): Likewise.
10078         (eliminate_regs_1): Likewise.
10079         (elimination_effects): Likewise.
10080         (mark_not_eliminable): Likewise.
10081         (scan_paradoxical_subregs): Likewise.
10082         (forget_old_reloads_1): Likewise.
10083         * reorg.c (find_end_label): Likewise.
10084         (try_merge_delay_insns): Likewise.
10085         (redundant_insn): Likewise.
10086         (own_thread_p): Likewise.
10087         (fill_simple_delay_slots): Likewise.
10088         (fill_slots_from_thread): Likewise.
10089         (dbr_schedule): Likewise.
10090         * resource.c (update_live_status): Likewise.
10091         (mark_referenced_resources): Likewise.
10092         (mark_set_resources): Likewise.
10093         * rtl.c (copy_rtx): Likewise.
10094         * rtlanal.c (reg_referenced_p): Likewise.
10095         (single_set_2): Likewise.
10096         (noop_move_p): Likewise.
10097         (note_stores): Likewise.
10098         * sched-deps.c (sched_analyze_reg): Likewise.
10099         (sched_analyze_insn): Likewise.
10101 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10103         * cse.c (invalidate_reg): New function extracted from...
10104         (invalidate): ...here.
10105         (canonicalize_insn): Check for clobber high.
10106         (invalidate_from_clobbers): invalidate clobber highs.
10107         (invalidate_from_sets_and_clobbers): Likewise.
10108         (count_reg_usage): Check for clobber high.
10109         (insn_live_p): Likewise.
10110         * cselib.c (cselib_expand_value_rtx_1):Likewise.
10111         (cselib_invalidate_regno): Check for clobber in setter.
10112         (cselib_invalidate_rtx): Pass through setter.
10113         (cselib_invalidate_rtx_note_stores):
10114         (cselib_process_insn): Check for clobber high.
10115         * cselib.h (cselib_invalidate_rtx): Add operand.
10117 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10119         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
10120         (mark_not_eliminable): Likewise.
10121         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
10122         * lra-lives.c (process_bb_lives): Check for clobber high.
10123         * lra.c (new_insn_reg): Remember clobber highs.
10124         (collect_non_operand_hard_regs): Check for clobber high.
10125         (lra_set_insn_recog_data): Likewise.
10126         (add_regs_to_insn_regno_info): Likewise.
10127         (lra_update_insn_regno_info): Likewise.
10129 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10131         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
10132         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
10134 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10136         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
10137         (copy_insn_1): Likewise.
10138         (gen_hard_reg_clobber_high): New gen function.
10139         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
10140         * genemit.c (gen_exp): Likewise.
10141         (gen_emit_seq): Pass through info.
10142         (gen_insn): Check for CLOBBER_HIGH.
10143         (gen_expand): Pass through info.
10144         (gen_split): Likewise.
10145         (output_add_clobbers): Likewise.
10146         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
10147         (remove_clobbers): Likewise.
10148         * rtl.h (gen_hard_reg_clobber_high): New declaration.
10150 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10152         * doc/rtl.texi (clobber_high): Add.
10153         (parallel): Add in clobber high
10154         * rtl.c (rtl_check_failed_code3): Add function.
10155         * rtl.def (CLOBBER_HIGH): Add expression.
10156         * rtl.h (RTL_CHECKC3): Add macro.
10157         (rtl_check_failed_code3): Add declaration.
10158         (XC3EXP): Add macro.
10160 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
10162         PR target/86386
10163         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
10164         cfun->machine->max_used_stack_alignment if needed.
10166 2018-08-04  Martin Sebor  <msebor@redhat.com>
10168         PR tree-optimization/86571
10169         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
10170         NaN output to 4.
10172 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
10174         PR target/86799
10175         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10176         Define.
10178 2018-08-03  Jeff Law  <law@redhat.com>
10180         PR target/86795
10181         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10182         Define to speculation_safe_value_not_needed.
10184 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
10186         * doc/gcov.texi (-x): Remove duplicate "to".
10187         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
10188         (-Wif-not-aligned): Remove duplicate "is".
10189         (-flto): Remove duplicate "the".
10190         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
10191         duplicate "v5.00.b".
10192         (MSP430 Options): Remove duplicate "and" from the description
10193         of "-mgprel-sec=regexp".
10194         (x86 Options): Remove duplicate copies of "vmldLog102" and
10195         vmlsLog104 from description of "-mveclibabi=type".
10197 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
10199         * internal-fn.h (first_commutative_argument): Declare.
10200         * internal-fn.c (first_commutative_argument): New function.
10201         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
10202         restrictions for pattern statements.  Use first_commutative_argument
10203         to look for commutative operands in calls to internal functions.
10205 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
10207         * Makefile.in (wide-int-range.o): New.
10208         * tree-vrp.c: Move all the wide_int_* functions to...
10209         * wide-int-range.cc: ...here.
10210         * tree-vrp.h: Move all the wide_int_* prototypes to...
10211         * wide-int-range.h: ...here.
10213 2018-08-03  Tom de Vries  <tdevries@suse.de>
10215         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
10216         UI_NONE.
10217         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
10218         * except.c (output_function_exception_table): Do early exit if
10219         targetm_common.except_unwind_info (&global_options) == UI_NONE.
10221 2018-08-03  Martin Liska  <mliska@suse.cz>
10223         * predict.c (dump_prediction): Change to 2 digits
10224         in fraction part.
10226 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
10228         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
10229         neon_dup_q to...
10230         (falkor_am_1_gtov_gtov): ... a new insn reservation.
10232 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
10234         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
10235         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
10236         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
10237         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
10238         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
10240 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
10242         * diagnostic-show-locus.c (diagnostic_show_locus): Use
10243         pp_take_prefix when saving the existing prefix.
10244         * diagnostic.c (diagnostic_append_note): Likewise.
10245         * langhooks.c (lhd_print_error_function): Likewise.
10246         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
10247         param's type.  Free the existing prefix.
10248         (pp_take_prefix): New function.
10249         (pretty_printer::pretty_printer): Drop the prefix parameter.
10250         Rename the length parameter to match the comment.
10251         (pretty_printer::~pretty_printer): Free the prefix.
10252         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
10253         parameter.
10254         (struct pretty_printer): Drop the "const" from "prefix" field's
10255         type and clarify memory management.
10256         (pp_set_prefix): Drop the "const" from the 2nd param.
10257         (pp_take_prefix): New decl.
10259 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
10261         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
10262         for word_mode_ok here instead of passing as argument.
10263         (expand_block_compare): Change select_block_compare_mode() call.
10264         (expand_strncmp_gpr_sequence): New function.
10265         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
10267 2018-08-02  Jeff Law  <law@redhat.com>
10269         PR target/86790
10270         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10271         Define to speculation_safe_value_not_needed.
10273         PR target/86784
10274         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10275         Define to speculation_safe_value_not_needed.
10277 2018-08-02  Tom de Vries  <tdevries@suse.de>
10279         PR target/86660
10280         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
10281         function.  Return UI_TARGET unconditionally.
10282         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
10283         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
10285 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10287         * genemit.c (print_overload_test): Fix typo.
10289 2018-08-02  Richard Biener  <rguenther@suse.de>
10291         PR tree-optimization/86816
10292         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
10293         which checks for value availability before querying it.
10294         (gvn_uses_equal): Use it.
10295         (same_succ_hash): Likewise.
10296         (gimple_equal_p): Likewise.
10298 2018-08-02  Nick Clifton  <nickc@redhat.com>
10300         PR target/86813
10301         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10302         Define to speculation_safe_value_not_needed.
10304         PR target/86810
10305         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10306         Define to speculation_safe_value_not_needed.
10308         PR target/86810
10309         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10310         Define to speculation_safe_value_not_needed.
10312         PR target/86803
10313         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10314         Define to speculation_safe_value_not_needed.
10316         PR target/86797
10317         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10318         Define to speculation_safe_value_not_needed.
10320         PR target/86791
10321         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10322         Define to speculation_safe_value_not_needed.
10324         PR target/86789
10325         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10326         Define to speculation_safe_value_not_needed.
10328         PR target/86787
10329         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10330         Define to speculation_safe_value_not_needed.
10332         PR target/86782
10333         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10334         speculation_safe_value_not_needed.
10336         PR target/86781
10337         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
10338         to speculation_safe_value_not_needed.
10340 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10342         * doc/md.texi: Expand the documentation of instruction names
10343         to mention port-local uses.  Document '@' in pattern names.
10344         * read-md.h (overloaded_instance, overloaded_name): New structs.
10345         (mapping): Declare.
10346         (md_reader::handle_overloaded_name): New member function.
10347         (md_reader::get_overloads): Likewise.
10348         (md_reader::m_first_overload): New member variable.
10349         (md_reader::m_next_overload_ptr): Likewise.
10350         (md_reader::m_overloads_htab): Likewise.
10351         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
10352         m_next_overload_ptr and m_overloads_htab.
10353         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
10354         (get_mode_token, get_code_token, get_int_token): New functions.
10355         (map_attr_string): Add an optional argument that passes back
10356         the associated iterator.
10357         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
10358         (md_reader::handle_overloaded_name, add_overload_instance): New
10359         functions.
10360         (apply_iterators): Handle '@' names.  Report an error if '@'
10361         is used without iterators.
10362         (initialize_iterators): Initialize the new iterator_group fields.
10363         * genopinit.c (handle_overloaded_code_for)
10364         (handle_overloaded_gen): New functions.
10365         (main): Use them to print declarations of maybe_code_for_* and
10366         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
10367         * genemit.c (print_overload_arguments, print_overload_test)
10368         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
10369         (main): Use it to print definitions of maybe_code_for_* and
10370         maybe_gen_* functions.
10371         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
10372         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
10373         instead of explicit mode checks.
10374         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
10375         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
10376         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
10377         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
10378         (aarch64_expand_compare_and_swap): Likewise
10379         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
10380         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
10381         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
10382         (aarch64_constant_pool_reload_icode): Delete.
10383         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
10384         instead of aarch64_constant_pool_reload_icode.  Use
10385         code_for_aarch64_reload_mov instead of explicit mode checks.
10386         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
10387         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
10388         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
10389         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
10390         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
10391         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
10392         (aarch64_atomic_load_op_code): Delete.
10393         (aarch64_emit_atomic_load_op): Likewise.
10394         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
10395         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
10396         instead of aarch64_emit_atomic_load_op.
10397         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
10398         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
10399         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
10400         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
10401         character before the pattern name.
10402         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
10403         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
10404         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
10405         (aarch64_frecps<mode>): Likewise.
10406         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
10407         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
10408         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
10409         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
10410         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
10412 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10414         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
10415         Allow HFmode constants if TARGET_FP_F16INST.
10417 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
10419         PR target/86014
10420         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
10421         No longer check last store for clobber of address register.
10423 2018-08-02  Martin Liska  <mliska@suse.cz>
10425         PR gcov-profile/86817
10426         * gcov.c (process_all_functions): New function.
10427         (main): Call it.
10428         (process_file): Move functions processing to
10429         process_all_functions.
10431 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
10433         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
10434         "const" to the "gimple *" and "rtx_insn *" parameters.
10435         * dumpfile.h (dump_user_location_t::dump_user_location_t):
10436         Likewise.
10437         (dump_location_t::dump_location_t): Likewise.
10439 2018-08-01  Martin Sebor  <msebor@redhat.com>
10441         PR tree-optimization/86650
10442         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
10443         rather than a "gcall *".  Directly pass the data of interest
10444         to percent_K_format, rather than building a temporary CALL_EXPR
10445         to hold it.
10446         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
10447         (gimple_fold_builtin_strncat): Adjust.
10448         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
10449         gcall* argument with gimple*.
10450         * gimple-ssa-warn-restrict.c (check_call): Same.
10451         (wrestrict_dom_walker::before_dom_children): Same.
10452         (builtin_access::builtin_access): Same.
10453         (check_bounds_or_overlap): Same
10454         (maybe_diag_overlap): Same.
10455         (maybe_diag_offset_bounds): Same.
10456         * tree-diagnostic.c (default_tree_printer): Move usage of
10457         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
10458         to this callsite.
10459         * tree-pretty-print.c (percent_K_format): Add argument.
10460         * tree-pretty-print.h: Add argument.
10461         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
10462         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
10463         (maybe_diag_stxncpy_trunc): Same.
10464         (handle_builtin_stxncpy): Same.
10465         (handle_builtin_strcat): Same.
10467 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10469         * match.pd: Optimise pointer range checks.
10471 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10473         PR tree-optimization/86758
10474         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
10475         to remove pattern statements.
10477 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10479         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
10480         result of dfs_enumerate_from when constructing stmt_vec_infos,
10481         instead of additionally calling get_loop_body.
10483 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10485         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
10486         parameter.
10487         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
10488         When creating an iv, assert that the step is not known to be zero.
10489         (vect_setup_realignment): Update call accordingly.
10490         * tree-vect-stmts.c (vectorizable_store): Likewise.
10491         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
10493 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10495         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
10496         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
10497         (vectorizable_reduction): Likewise.
10498         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10499         (vect_detect_hybrid_slp_stmts): Likewise.
10500         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10502 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
10504         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
10505         (wide_int_set_zero_nonzero_bits): ...this.
10506         (zero_nonzero_bits_from_vr): Rename to...
10507         (vrp_set_zero_nonzero_bits): ...this.
10508         (extract_range_from_multiplicative_op_1): Abstract wide int
10509         code...
10510         (wide_int_range_multiplicative_op): ...here.
10511         (extract_range_from_binary_expr_1): Extract wide int binary
10512         operations into their own functions.
10513         (wide_int_range_lshift): New.
10514         (wide_int_range_can_optimize_bit_op): New.
10515         (wide_int_range_shift_undefined_p): New.
10516         (wide_int_range_bit_xor): New.
10517         (wide_int_range_bit_ior): New.
10518         (wide_int_range_bit_and): New.
10519         (wide_int_range_trunc_mod): New.
10520         (extract_range_into_wide_ints): New.
10521         (vrp_shift_undefined_p): New.
10522         (extract_range_from_multiplicative_op): New.
10523         (vrp_can_optimize_bit_op): New.
10524         * tree-vrp.h (value_range::dump): New.
10525         (wide_int_range_multiplicative_op): New.
10526         (wide_int_range_lshift):New.
10527         (wide_int_range_shift_undefined_p): New.
10528         (wide_int_range_bit_xor): New.
10529         (wide_int_range_bit_ior): New.
10530         (wide_int_range_bit_and): New.
10531         (wide_int_range_trunc_mod): New.
10532         (zero_nonzero_bits_from_bounds): Rename to...
10533         (wide_int_set_zero_nonzero_bits): ...this.
10534         (zero_nonzero_bits_from_vr): Rename to...
10535         (vrp_set_zero_nonzero_bits): ...this.
10536         (range_easy_mask_min_max): Rename to...
10537         (wide_int_range_can_optimize_bit_op): this.
10538         * vr-values.c (simplify_bit_ops_using_ranges): Rename
10539         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
10541 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10543         * tree-vectorizer.h (vect_orig_stmt): New function.
10544         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
10545         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
10546         (vect_create_epilog_for_reduction): Likewise.
10547         (vectorizable_live_operation): Likewise.
10548         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10549         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
10550         * tree-vect-stmts.c (vectorizable_call): Likewise.
10551         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
10553 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10555         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
10556         argument.
10557         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
10558         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
10559         (vect_transform_loop): Likewise.
10560         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10562 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10564         * tree-vectorizer.h (vect_schedule_slp): Return void.
10565         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10566         (vect_schedule_slp): Likewise.
10568 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10570         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
10571         argument.
10572         (vect_transform_loop): Update calls accordingly.  Schedule SLP
10573         instances before the main loop, if any exist.
10575 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10577         PR tree-optimization/86749
10578         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
10579         If the lhs is used in a COND_EXPR, check that it is being used
10580         as the "then" or "else" value.
10582 2018-08-01  Tom de Vries  <tdevries@suse.de>
10584         PR target/86800
10585         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10586         speculation_safe_value_not_needed.
10588 2018-08-01  Richard Biener  <rguenther@suse.de>
10590         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
10591         as base and offset.
10593 2018-08-01  Martin Liska  <mliska@suse.cz>
10595         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
10596         format how successful transformation is dumped.
10597         (gimple_mod_pow2_value_transform): Likewise.
10598         (gimple_mod_subtract_transform): Likewise.
10599         (gimple_stringops_transform): Likewise.
10601 2018-08-01  Martin Liska  <mliska@suse.cz>
10603         PR value-prof/35543
10604         * value-prof.c (interesting_stringop_to_profile_p):
10605         Simplify the code and add BUILT_IN_MEMMOVE.
10606         (gimple_stringops_transform): Likewise.
10608 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
10610         * config/aarch64/aarch64-simd.md
10611         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
10612         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
10613         use GPI iterator instead of SI mode.
10615 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10617         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
10618         rs6000_speculation_barrier.
10619         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
10620         new barrier pattern name.
10622 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10624         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
10625         (speculation_barrier): New insn.
10627 2018-08-01  Richard Biener  <rguenther@suse.de>
10629         PR bootstrap/86724
10630         * graphite.h: Include isl/id.h and isl/space.h to allow build
10631         with ISL 0.20.
10633 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
10635         PR target/86651
10636         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10637         mode for COFF targets.
10638         * defaults.h (TARGET_COFF): Define.
10639         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
10640         TARGET_COFF): Define.
10641         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
10642         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
10643         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
10645 2018-07-31  Alexandre Oliva <oliva@adacore.com>
10647         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
10648         * gimple-streamer-out.c (output_bb): Save it.
10649         * lto-streamer-in.c (input_struct_function_base): Restore
10650         instance discriminator if available.  Create map on demand.
10651         * lto-streamer-out.c (output_struct_function_base): Save it if
10652         available.
10653         * final.c (decl_to_instance_map): Document LTO strategy.
10655 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
10656             Olivier Hainque  <hainque@adacore.com>
10658         * debug.h (decl_to_instance_map_t): New type.
10659         (decl_to_instance_map): Declare.
10660         (maybe_create_decl_to_instance_map): New inline function.
10661         * final.c (bb_discriminator, last_bb_discriminator): New statics,
10662         to track basic block discriminators.
10663         (final_start_function_1): Initialize them.
10664         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
10665         bb_discriminator.
10666         (decl_to_instance_map): New variable.
10667         (map_decl_to_instance, maybe_set_discriminator): New functions.
10668         (notice_source_line): Set discriminator.
10670 2018-07-31  Ian Lance Taylor  <iant@golang.org>
10672         * targhooks.c (default_have_speculation_safe_value): Add
10673         ATTRIBUTE_UNUSED.
10675 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10677         * dump-context.h: Include "pretty-print.h".
10678         (dump_context::refresh_dumps_are_enabled): New decl.
10679         (dump_context::emit_item): New decl.
10680         (class dump_context): Add fields "m_test_pp" and
10681         "m_test_pp_flags".
10682         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10683         (temp_dump_context::get_dumped_text): New decl.
10684         (class temp_dump_context): Add field "m_pp".
10685         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
10686         (dump_context::refresh_dumps_are_enabled): ...and add a test for
10687         m_test_pp.
10688         (set_dump_file): Update for above change.
10689         (set_alt_dump_file): Likewise.
10690         (dump_loc): New overload, taking a pretty_printer *.
10691         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
10692         to any test pretty-printer.
10693         (make_item_for_dump_gimple_stmt): New function, adapted from
10694         optinfo::add_gimple_stmt.
10695         (dump_context::dump_gimple_stmt): Call it, and use the result,
10696         eliminating the direct usage of dump_file and alt_dump_file in
10697         favor of indirectly using them via emit_item.
10698         (make_item_for_dump_gimple_expr): New function, adapted from
10699         optinfo::add_gimple_expr.
10700         (dump_context::dump_gimple_expr): Call it, and use the result,
10701         eliminating the direct usage of dump_file and alt_dump_file in
10702         favor of indirectly using them via emit_item.
10703         (make_item_for_dump_generic_expr): New function, adapted from
10704         optinfo::add_tree.
10705         (dump_context::dump_generic_expr): Call it, and use the result,
10706         eliminating the direct usage of dump_file and alt_dump_file in
10707         favor of indirectly using them via emit_item.
10708         (make_item_for_dump_printf_va): New function, adapted from
10709         optinfo::add_printf_va.
10710         (make_item_for_dump_printf): New function.
10711         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
10712         and use the result, eliminating the direct usage of dump_file and
10713         alt_dump_file in favor of indirectly using them via emit_item.
10714         (make_item_for_dump_dec): New function.
10715         (dump_context::dump_dec): Call it, and use the result,
10716         eliminating the direct usage of dump_file and alt_dump_file in
10717         favor of indirectly using them via emit_item.
10718         (make_item_for_dump_symtab_node): New function, adapted from
10719         optinfo::add_symtab_node.
10720         (dump_context::dump_symtab_node): Call it, and use the result,
10721         eliminating the direct usage of dump_file and alt_dump_file in
10722         favor of indirectly using them via emit_item.
10723         (dump_context::begin_scope): Reimplement, avoiding direct usage
10724         of dump_file and alt_dump_file in favor of indirectly using them
10725         via emit_item.
10726         (dump_context::emit_item): New member function.
10727         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10728         Set up test pretty-printer on the underlying context.  Call
10729         refresh_dumps_are_enabled.
10730         (temp_dump_context::~temp_dump_context): Call
10731         refresh_dumps_are_enabled.
10732         (temp_dump_context::get_dumped_text): New member function.
10733         (selftest::verify_dumped_text): New function.
10734         (ASSERT_DUMPED_TEXT_EQ): New macro.
10735         (selftest::test_capture_of_dump_calls): Run all tests twice, with
10736         and then without optinfo enabled.  Add uses of
10737         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
10738         * dumpfile.h: Update comment for the dump_* API.
10739         * optinfo-emit-json.cc
10740         (selftest::test_building_json_from_dump_calls): Update for new
10741         param for temp_dump_context ctor.
10742         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
10743         and "m_owned" field.
10744         (optinfo_item::~optinfo_item): Likewise.
10745         (optinfo::add_item): New member function.
10746         (optinfo::emit): Update comment.
10747         (optinfo::add_string): Delete.
10748         (optinfo::add_printf): Delete.
10749         (optinfo::add_printf_va): Delete.
10750         (optinfo::add_gimple_stmt): Delete.
10751         (optinfo::add_gimple_expr): Delete.
10752         (optinfo::add_tree): Delete.
10753         (optinfo::add_symtab_node): Delete.
10754         (optinfo::add_dec): Delete.
10755         * optinfo.h (class dump_context): New forward decl.
10756         (optinfo::add_item): New decl.
10757         (optinfo::add_string): Delete.
10758         (optinfo::add_printf): Delete.
10759         (optinfo::add_printf_va): Delete.
10760         (optinfo::add_gimple_stmt): Delete.
10761         (optinfo::add_gimple_expr): Delete.
10762         (optinfo::add_tree): Delete.
10763         (optinfo::add_symtab_node): Delete.
10764         (optinfo::add_dec): Delete.
10765         (optinfo::add_poly_int): Delete.
10766         (optinfo_item::optinfo_item): Remove "owned" param.
10767         (class optinfo_item): Remove field "m_owned".
10769 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10771         PR middle-end/86705
10772         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
10773         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
10774         requested variable alignment.
10775         (expand_one_ssa_partition): Likewise.
10776         (expand_one_var): Likewise.
10778 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10780         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10781         to speculation_safe_value_not_needed.
10783 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10785         * targhooks.h (speculation_safe_value_not_needed): New prototype.
10786         * targhooks.c (speculation_safe_value_not_needed): New function.
10787         * target.def (have_speculation_safe_value): Update documentation.
10788         * doc/tm.texi: Regenerated.
10790 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10792         * config/aarch64/iterators.md (ALLI_TI): New iterator.
10793         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
10794         expand.
10795         (despeculate_copy<ALLI:mode>_insn): New insn.
10796         (despeculate_copyti_insn): New insn.
10797         (despeculate_simple<ALLI:mode>): New insn
10798         (despeculate_simpleti): New insn.
10799         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
10800         function.
10801         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
10802         aarch64_speculation_safe_value.
10803         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
10805 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10807         * config/aarch64/aarch64-speculation.cc: New file.
10808         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
10809         before pass_reorder_blocks.
10810         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
10811         prototype.
10812         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
10813         X14 and X15 when tracking speculation.
10814         * config/aarch64/aarch64.md (register name constants): Add
10815         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
10816         (unspec): Add UNSPEC_SPECULATION_TRACKER.
10817         (speculation_barrier): New insn attribute.
10818         (cmp<mode>): Allow SP in comparisons.
10819         (speculation_tracker): New insn.
10820         (speculation_barrier): Add speculation_barrier attribute.
10821         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
10822         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
10823         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
10825 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10827         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
10828         aarch64_track_speculation is true.
10829         (tb<optab><mode>1): Likewise.
10830         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
10831         generate CB[N]Z when tracking speculation.
10832         (aarch64_split_compare_and_swap): Likewise.
10833         (aarch64_split_atomic_op): Likewise.
10835 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10837         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
10839 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10841         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
10842         (speculation_barrier): New insn.
10844 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10846         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
10847         * config/arm/arm.md (speculation_barrier): New expand.
10848         (speculation_barrier_insn): New pattern.
10850 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10852         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
10853         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
10854         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
10855         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
10856         list.
10857         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
10858         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
10859         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
10860         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
10861         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
10862         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
10863         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
10864         * builtins.c (expand_speculation_safe_value): New function.
10865         (expand_builtin): Call it.
10866         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
10867         * doc/extend.texi: Document __builtin_speculation_safe_value.
10868         * doc/md.texi: Document "speculation_barrier" pattern.
10869         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
10870         TARGET_HAVE_SPECULATION_SAFE_VALUE.
10871         * doc/tm.texi: Regenerated.
10872         * target.def (have_speculation_safe_value, speculation_safe_value): New
10873         hooks.
10874         * targhooks.c (default_have_speculation_safe_value): New function.
10875         (default_speculation_safe_value): New function.
10876         * targhooks.h (default_have_speculation_safe_value): Add prototype.
10877         (default_speculation_safe_value): Add prototype.
10879 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10881         * dump-context.h (dump_context::dump_loc): New decl.
10882         * dumpfile.c (dump_context::dump_loc): New member function.
10883         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
10884         and dump_gimple_stmt.
10885         (dump_context::dump_gimple_expr_loc): Likewise, using
10886         dump_gimple_expr.
10887         (dump_context::dump_generic_expr_loc): Likewise, using
10888         dump_generic_expr.
10889         (dump_context::dump_printf_loc_va): Likewise, using
10890         dump_printf_va.
10891         (dump_context::begin_scope): Explicitly using the global function
10892         "dump_loc", rather than the member function.
10894 2018-07-31  Martin Sebor  <msebor@redhat.com>
10896         PR tree-optimization/86741
10897         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
10899 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
10901         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
10902         depend on whether prefetch instructions will be emitted or not.
10903         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
10904         will be emitted or not.
10905         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
10906         (TARGET_SETMEM_PFD): New macros.
10908 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10910         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
10911         (NULL_STMT_VEC_INFO): Delete.
10912         (stmt_vec_info::operator*): Likewise.
10913         (stmt_vec_info::operator gimple *): Likewise.
10914         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
10915         of NULL_STMT_VEC_INFO.
10916         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10917         (vect_reassociating_reduction_p): Likewise.
10918         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10919         (vectorizable_store): Likewise.
10920         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
10921         (vec_info::free_stmt_vec_infos): Likewise.
10923 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10925         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
10926         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
10928 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10930         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
10931         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10932         (vec_info::free_stmt_vec_info): New private member functions.
10933         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
10934         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
10935         * tree-parloops.c (gather_scalar_reductions): Remove calls to
10936         set_stmt_vec_info_vec and free_stmt_vec_infos.
10937         * tree-vect-loop.c (_loop_vec_info): Remove call to
10938         set_stmt_vec_info_vec.
10939         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
10940         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
10941         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
10942         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10943         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
10944         assignments in {vec_info::,}new_stmt_vec_info that are redundant
10945         with the clearing in the xcalloc.
10946         (stmt_vec_info_vec): Delete.
10947         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
10948         (vectorize_loops): Likewise.
10949         (vec_info::~vec_info): Remove argument from call to
10950         free_stmt_vec_infos.
10951         (vec_info::add_stmt): Remove vinfo argument from call to
10952         new_stmt_vec_info.
10954 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10956         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
10957         rather than a gimple stmt.
10958         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
10959         information for pattern statements when passed the original
10960         statement; instead wait to be passed the pattern statement itself.
10961         Don't call set_vinfo_for_stmt here.
10962         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
10963         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
10964         stmt_vec_infos here.
10965         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
10966         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
10967         stmt_vec_infos entry.
10969 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10971         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
10972         * tree-vectorizer.c (vec_info::replace_stmt): New function.
10973         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
10974         * tree-vect-stmts.c (vectorizable_call): Likewise.
10975         (vectorizable_simd_clone_call): Likewise.
10977 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10979         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
10980         * tree-vectorizer.c (vec_info::remove_stmt): New function.
10981         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
10982         * tree-vect-loop.c (vect_transform_loop): Likewise.
10983         * tree-vect-slp.c (vect_schedule_slp): Likewise.
10984         * tree-vect-stmts.c (vect_remove_stores): Likewise.
10986 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10988         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
10989         (vect_dr_stmt): Delete.
10990         * tree-vectorizer.c (vec_info::lookup_dr): New function.
10991         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
10992         of DR_VECT_AUX.
10993         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10994         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
10995         (vect_verify_datarefs_alignment, vect_peeling_supportable)
10996         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10997         (vect_analyze_data_refs): Likewise.
10998         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
10999         argument.
11000         (vect_find_same_alignment_drs): Likewise.
11001         (vect_slp_analyze_node_dependences): Update calls accordingly.
11002         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
11003         instead of DR_VECT_AUX.
11004         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
11005         of a vector data references.  Use vec_info::lookup_dr instead of
11006         DR_VECT_AUX.
11007         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
11008         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
11009         instead of DR_VECT_AUX.
11011 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11013         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
11014         dr_vec_info.
11015         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
11016         accordingly.
11017         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11018         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11019         (vect_gen_prolog_loop_niters): Likewise.
11021 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11023         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
11024         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
11025         (vect_known_alignment_in_bytes, vect_dr_behavior)
11026         (vect_get_scalar_dr_size): Take references as dr_vec_infos
11027         instead of data_references.  Update calls to other routines for
11028         which the same change has been made.
11029         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
11030         dr_vec_infos instead of stmt_vec_infos.
11031         (vect_analyze_data_ref_dependence): Update call accordingly.
11032         (vect_slp_analyze_data_ref_dependence)
11033         (vect_record_base_alignments): Use DR_VECT_AUX.
11034         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
11035         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
11036         (vector_alignment_reachable_p, vect_get_data_access_cost)
11037         (vect_peeling_supportable, vect_analyze_group_access_1)
11038         (vect_analyze_group_access, vect_analyze_data_ref_access)
11039         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
11040         (vect_compile_time_alias, vect_small_gap_p)
11041         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
11042         (vect_supportable_dr_alignment): Take references as dr_vec_infos
11043         instead of data_references.  Update calls to other routines for
11044         which the same change has been made.
11045         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
11046         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
11047         (vect_slp_analyze_and_verify_node_alignment)
11048         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
11049         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
11050         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
11051         above changes.
11052         (_vect_peel_info::dr): Replace with...
11053         (_vect_peel_info::dr_info): ...this new field.
11054         (vect_peeling_hash_get_most_frequent)
11055         (vect_peeling_hash_choose_best_peeling): Update accordingly.
11056         (vect_peeling_hash_get_lowest_cost):
11057         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
11058         routines for which the same change has been made.
11059         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
11060         data_reference.
11061         * tree-vect-loop-manip.c (get_misalign_in_elems)
11062         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
11063         above changes.
11064         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11065         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
11066         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
11067         (get_group_load_store_type, get_negative_load_store_type)
11068         (vect_get_data_ptr_increment, vectorizable_store)
11069         (vectorizable_load): Likewise.
11070         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
11071         Update calls to other routines for which the same change has been made.
11073 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11075         * tree-vectorizer.h (vec_info::move_dr): New member function.
11076         (dataref_aux): Rename to...
11077         (dr_vec_info): ...this and add "dr" and "stmt" fields.
11078         (_stmt_vec_info::dr_aux): Update accordingly.
11079         (_stmt_vec_info::data_ref_info): Delete.
11080         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
11081         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
11082         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
11083         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
11084         of data_ref.
11085         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
11086         (STMT_VINFO_DR_INFO): New macro.
11087         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
11088         (set_dr_misalignment): Update after rename of dataref_aux.
11089         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
11090         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
11091         initialization of STMT_VINFO_DATA_REF.
11092         * tree-vectorizer.c (vec_info::move_dr): New function.
11093         * tree-vect-patterns.c (vect_recog_bool_pattern)
11094         (vect_recog_mask_conversion_pattern)
11095         (vect_recog_gather_scatter_pattern): Use it.
11096         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
11097         the "dr" and "stmt" fields of dr_vec_info instead of
11098         STMT_VINFO_DATA_REF.
11100 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11102         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
11103         (is_pattern_stmt_p): Use it.
11104         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
11105         on pattern statements.
11107 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11109         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
11110         original stmt as a stmt_vec_info rather than a gimple stmt.
11111         (vect_pattern_recog_1): Take the statement directly as a
11112         stmt_vec_info, rather than via a gimple_stmt_iterator.
11113         Update call to vect_mark_pattern_stmts.
11114         (vect_pattern_recog): Update calls accordingly.
11116 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11118         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
11119         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
11120         a vect_def_type for the first argument.
11121         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
11122         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
11123         operand if it isn't defined by a vectorized statement.
11124         (vect_build_gather_load_calls): Remove the mask_dt argument and
11125         update calls to vect_get_vec_def_for_stmt_copy.
11126         (vectorizable_bswap): Likewise the dt argument.
11127         (vectorizable_call): Update calls to vectorizable_bswap and
11128         vect_get_vec_def_for_stmt_copy.
11129         (vectorizable_simd_clone_call, vectorizable_assignment)
11130         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
11131         (vectorizable_comparison): Update calls to
11132         vect_get_vec_def_for_stmt_copy.
11133         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
11134         vect_is_simple_use.
11135         (vect_get_loop_based_defs): Remove dt argument and update call
11136         to vect_get_vec_def_for_stmt_copy.
11137         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
11138         and vect_get_vec_def_for_stmt_copy.
11139         (vectorizable_load): Update calls to vect_build_gather_load_calls
11140         and vect_get_vec_def_for_stmt_copy.
11141         * tree-vect-loop.c (vect_create_epilog_for_reduction)
11142         (vectorizable_reduction, vectorizable_live_operation): Update calls
11143         to vect_get_vec_def_for_stmt_copy.
11145 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11147         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
11148         and gimple stmt arguments with a stmt_vec_info.
11149         (vect_record_base_alignments): Update calls accordingly.
11150         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
11151         and gimple stmt arguments with a stmt_vec_info.
11152         (vect_build_slp_tree_1): Remove vinfo argument and update call
11153         to vect_record_max_nunits.
11154         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
11155         and vect_record_max_nunits.
11157 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11159         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
11160         file and take a stmt_vec_info instead of a gimple stmt.
11161         (supportable_widening_operation, vect_finish_replace_stmt)
11162         (vect_finish_stmt_generation, vect_get_store_rhs)
11163         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
11164         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
11165         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
11166         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
11167         (vect_create_data_ref_ptr, bump_vector_ptr)
11168         (vect_permute_store_chain, vect_setup_realignment)
11169         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
11170         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
11171         (vectorizable_reduction, vectorizable_induction)
11172         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
11173         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
11174         than gimple stmts as arguments.
11175         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
11176         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
11177         (can_group_stmts_p, vect_check_gather_scatter)
11178         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
11179         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
11180         (vect_permute_load_chain, vect_shift_permute_load_chain)
11181         (vect_transform_grouped_load)
11182         (vect_record_grouped_load_vectors): Likewise.
11183         * tree-vect-loop.c (vect_fixup_reduc_chain)
11184         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
11185         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
11186         (vectorizable_reduction, vectorizable_induction)
11187         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
11188         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
11189         (vect_get_load_store_mask): Likewise.
11190         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
11191         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
11192         * tree-vect-stmts.c (vect_mark_relevant)
11193         (is_simple_and_all_uses_invariant)
11194         (exist_non_indexing_operands_for_use_p, process_use)
11195         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
11196         (vect_get_vec_def_for_operand, vect_get_vec_defs)
11197         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
11198         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
11199         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
11200         (get_negative_load_store_type, get_load_store_type)
11201         (vect_check_load_store_mask, vect_check_store_rhs)
11202         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
11203         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11204         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11205         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11206         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
11207         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
11208         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
11209         (supportable_widening_operation): Likewise.
11211 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11213         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
11214         a stmt_vec_info instead of a gcall.
11215         (vect_check_gather_scatter): Update call accordingly.
11216         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
11217         of a gphi.
11218         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
11219         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
11220         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
11221         instead of a gimple stmt.
11222         (vect_transform_loop): Update calls accordingly.
11223         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
11224         stmt_vec_infos instead of gimple stmts.
11225         (vect_analyze_slp_instance): Update use accordingly.
11226         * tree-vect-stmts.c (read_vector_array, write_vector_array)
11227         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
11228         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
11229         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
11230         (vect_gen_widened_results_half, vect_get_loop_based_defs)
11231         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
11232         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
11233         down to subroutines.
11235 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11237         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
11238         of the worklist from a vector of gimple stmts to a vector of
11239         stmt_vec_infos.
11240         * tree-vect-stmts.c (vect_mark_relevant, process_use)
11241         (vect_mark_stmts_to_be_vectorized): Likewise
11243 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11245         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
11246         statement before passing it to vect_analyze_stmt.
11247         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
11248         the chain of phi vector definitions.  Track the exit phi via its
11249         stmt_vec_info.
11250         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
11251         STMT_VINFO_REDUC_DEF.
11252         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
11253         stmt_vec_infos to handle the statement chains.
11254         (vect_get_slp_defs): Record the first statement in the node
11255         using a stmt_vec_info.
11256         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
11257         statements here and pass their stmt_vec_info down to subroutines.
11258         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
11259         down to vect_finish_stmt_generation.
11260         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
11261         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
11262         stmt_vec_infos to subroutines.
11263         (vect_remove_stores): Use stmt_vec_infos to handle the statement
11264         chains.
11266 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11268         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
11269         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
11270         (vect_permute_store_chain, vect_setup_realignment)
11271         (vect_permute_load_chain, vect_shift_permute_load_chain)
11272         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
11273         stmts internally, and when passing values to other vectorizer routines.
11274         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11275         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
11276         (vect_analyze_loop_operations, get_initial_def_for_reduction)
11277         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
11278         (vectorizable_reduction, vectorizable_induction)
11279         (vectorizable_live_operation, vect_transform_loop_stmt)
11280         (vect_transform_loop): Likewise.
11281         * tree-vect-patterns.c (vect_reassociating_reduction_p)
11282         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
11283         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
11284         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11285         (vect_slp_analyze_node_operations_1): Likewise.
11286         * tree-vect-stmts.c (vect_mark_relevant, process_use)
11287         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
11288         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
11289         (vect_finish_stmt_generation_1, get_group_load_store_type)
11290         (get_load_store_type, vect_build_gather_load_calls)
11291         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11292         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11293         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11294         (vectorizable_store, vectorizable_load, vectorizable_condition)
11295         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
11296         (supportable_widening_operation): Likewise.
11297         (vect_get_vector_types_for_stmt): Likewise.
11298         * tree-vectorizer.h (vect_dr_behavior): Likewise.
11300 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11302         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11303         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
11304         (vect_permute_store_chain, vect_permute_load_chain)
11305         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
11306         repeated stmt_vec_info lookups.
11307         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11308         (vect_update_ivs_after_vectorizer): Likewise.
11309         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
11310         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11311         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
11312         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11313         (vect_bb_slp_scalar_cost): Likewise.
11314         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
11316 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11318         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
11319         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
11320         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
11321         to get gassigns and gcalls, rather than operating on generc gimple
11322         stmts.
11323         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
11324         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
11325         (vectorizable_load, vect_analyze_stmt): Likewise.
11326         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
11328 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11330         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
11331         return stmt_vec_infos rather than gimple stmts.  Do not accept
11332         null arguments.
11333         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
11334         of a gimple stmt.
11335         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
11336         Update use of get_later_stmt.
11337         (vect_get_constant_vectors): Update call accordingly.
11338         (vect_schedule_slp_instance): Likewise
11339         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
11340         (vect_slp_analyze_instance_dependence): Likewise.
11341         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
11343 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11345         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
11346         (stmt_info_for_cost::stmt_info): ...this new field.
11347         (add_stmt_costs): Update accordingly.
11348         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11349         (vect_get_known_peeling_cost): Likewise.
11350         (vect_estimate_min_profitable_iters): Likewise.
11351         * tree-vect-stmts.c (record_stmt_cost): Likewise.
11353 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11355         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
11356         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11357         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
11358         accordingly.
11359         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
11361 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11363         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
11364         a gimple stmt to a stmt_vec_info.
11365         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
11367 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11369         * tree-vectorizer.h (vec_info::grouped_stores): Change from
11370         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11371         (_loop_vec_info::reduction_chains): Likewise.
11372         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
11373         accordingly.
11374         * tree-vect-slp.c (vect_analyze_slp): Likewise.
11376 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11378         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
11379         a gimple stmt to a stmt_vec_info.
11380         (_stmt_vec_info::next_element): Likewise.
11381         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
11382         (vect_slp_analyze_and_verify_node_alignment)
11383         (vect_analyze_group_access_1, vect_analyze_group_access)
11384         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
11385         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
11386         (vect_supportable_dr_alignment): Update accordingly.
11387         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
11388         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
11389         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
11390         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11391         * tree-vect-slp.c (vect_build_slp_tree_1)
11392         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
11393         (vect_split_slp_store_group, vect_analyze_slp_instance)
11394         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
11395         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
11396         (get_group_load_store_type, get_load_store_type)
11397         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
11398         (vect_transform_stmt, vect_remove_stores): Likewise.
11400 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11402         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
11403         than a gimple stmt.
11404         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11405         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
11406         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
11407         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
11408         (vector_alignment_reachable_p, vect_get_data_access_cost)
11409         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
11410         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
11411         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
11412         (vect_analyze_group_access_1, vect_analyze_group_access)
11413         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
11414         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
11415         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
11416         result of vect_dr_stmt and use the stmt_vec_info instead of
11417         the associated gimple stmt.
11418         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11419         (vect_gen_prolog_loop_niters): Likewise.
11420         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11422 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11424         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
11425         to a vec<stmt_vec_info>.
11426         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
11427         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
11428         vec<stmt_vec_info>.
11429         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
11430         to a vec<stmt_vec_info>.
11431         (bst_traits::value_type, bst_traits::value_type): Likewise.
11432         (bst_traits::hash): Update accordingly.
11433         (vect_get_and_check_slp_defs): Change the stmts parameter from
11434         a vec<gimple *> to a vec<stmt_vec_info>.
11435         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
11436         (vect_build_slp_tree): Likewise.
11437         (vect_build_slp_tree_2): Likewise.  Update uses of
11438         SLP_TREE_SCALAR_STMTS.
11439         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
11440         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
11441         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
11442         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
11443         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
11444         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
11445         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
11446         (vect_get_constant_vectors, vect_get_slp_defs)
11447         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
11448         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
11449         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
11450         instead of gimple stmts.
11451         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
11452         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
11453         (vect_slp_analyze_instance_dependence): Update uses of
11454         SLP_TREE_SCALAR_STMTS.
11455         (vect_slp_analyze_and_verify_node_alignment): Likewise.
11456         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
11457         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
11458         (get_initial_defs_for_reduction): Likewise.
11459         (vect_create_epilog_for_reduction): Likewise.
11460         (vectorize_fold_left_reduction): Likewise.
11461         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
11462         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
11463         (can_vectorize_live_stmts): Likewise.
11465 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11467         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
11468         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11469         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
11470         than gimple stmts.
11471         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
11472         of a gimple stmt.
11473         (gather_scalar_reductions): Update after above interface changes.
11474         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
11475         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
11476         than gimple stmts.
11477         (vect_force_simple_reduction): Likewise.
11478         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
11479         LOOP_VINFO_REDUCTIONS.
11480         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11482 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11484         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
11485         a gimple stmt to a stmt_vec_info.
11486         * tree-vect-loop.c (vect_active_double_reduction_p)
11487         (vect_force_simple_reduction, vectorizable_reduction): Update
11488         accordingly.
11490 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11492         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
11493         vec<gimple *> to a vec<stmt_vec_info>.
11494         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
11495         the reduction_phis argument from a vec<gimple *> to a
11496         vec<stmt_vec_info>.
11497         (vectorizable_reduction): Likewise the phis local variable that
11498         is passed to vect_create_epilog_for_reduction.  Update for new type
11499         of SLP_TREE_VEC_STMTS.
11500         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
11501         (vectorizable_live_operation): Likewise.
11502         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
11503         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
11505 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11507         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
11508         a gimple stmt to a stmt_vec_info.
11509         (vectorizable_condition, vectorizable_live_operation)
11510         (vectorizable_reduction, vectorizable_induction): Pass back the
11511         vectorized statement as a stmt_vec_info.
11512         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
11513         use of STMT_VINFO_VEC_STMT.
11514         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
11515         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
11516         as stmt_vec_infos rather than gimple stmts.
11517         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
11518         to a stmt_vec_info.
11519         (vectorizable_live_operation): Likewise.
11520         (vectorizable_reduction, vectorizable_induction): Likewise,
11521         updating use of STMT_VINFO_VEC_STMT.
11522         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
11523         of STMT_VINFO_VEC_STMT.
11524         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
11525         (vectorizable_simd_clone_call, vectorizable_conversion)
11526         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11527         (vectorizable_store, vectorizable_load, vectorizable_condition)
11528         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
11529         from a gimple stmt to a stmt_vec_info.
11530         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
11531         pointer to a stmt_vec_info to the vectorizable_* routines.
11533 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11535         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
11536         a gimple stmt to a stmt_vec_info.
11537         (is_pattern_stmt_p): Update accordingly.
11538         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
11539         (vect_record_grouped_load_vectors): Likewise.
11540         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
11541         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
11542         (vect_model_reduction_cost): Likewise.
11543         (vect_create_epilog_for_reduction): Likewise.
11544         (vectorizable_reduction, vectorizable_induction): Likewise.
11545         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11546         Return the stmt_vec_info for the pattern statement.
11547         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
11548         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
11549         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11550         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
11551         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
11552         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
11553         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
11554         (free_stmt_vec_info, vect_is_simple_use): Likewise.
11556 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11558         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
11559         (vect_finish_stmt_generation): Likewise.
11560         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11561         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
11562         (vect_build_gather_load_calls): Use the return value of the above
11563         functions instead of a separate call to vinfo_for_stmt.  Use narrow
11564         scopes for the input gimple stmt and wider scopes for the associated
11565         stmt_vec_info.  Use vec_info::lookup_def when setting these
11566         stmt_vec_infos from an SSA_NAME definition.
11567         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11568         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11569         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11570         (vectorizable_store, vectorizable_load, vectorizable_condition)
11571         (vectorizable_comparison): Likewise.
11572         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
11573         (vectorizable_reduction): Likewise.
11575 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11577         * tree-vectorizer.h (vect_is_simple_use): Add an optional
11578         stmt_vec_info * parameter before the optional gimple **.
11579         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11580         (process_use, vect_get_vec_def_for_operand_1): Update callers.
11581         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
11582         * tree-vect-loop.c (vectorizable_reduction): Likewise.
11583         (vectorizable_live_operation): Likewise.
11584         * tree-vect-patterns.c (type_conversion_p): Likewise.
11585         (vect_look_through_possible_promotion): Likewise.
11586         (vect_recog_rotate_pattern): Likewise.
11587         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11589 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11591         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
11592         a typedef to a wrapper class.
11593         (NULL_STMT_VEC_INFO): New macro.
11594         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
11595         (stmt_vec_info::operator*): New function.
11596         (stmt_vec_info::operator gimple *): Likewise.
11597         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
11598         (add_stmt_costs): Likewise.
11599         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
11600         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11601         (vect_get_known_peeling_cost): Likewise.
11602         (vect_estimate_min_profitable_iters): Likewise.
11603         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11604         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
11605         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
11606         (vectorizable_store, free_stmt_vec_infos): Likewise.
11607         (new_stmt_vec_info): Change return type of xcalloc to
11608         _stmt_vec_info *.
11610 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11612         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
11613         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
11614         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
11615         a single_imm_use-based sequence.
11616         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
11618 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11620         * tree-vectorizer.h (vec_info::lookup_def): Declare.
11621         * tree-vectorizer.c (vec_info::lookup_def): New function.
11622         * tree-vect-patterns.c (vect_get_internal_def): Use it.
11623         (vect_widened_op_tree): Likewise.
11624         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11625         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
11626         (vectorizable_reduction): Likewise.
11627         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
11628         of a gimple *.
11629         (vect_is_slp_reduction): Update calls accordingly.  Use
11630         vec_info::lookup_def.
11631         (vect_is_simple_reduction): Likewise
11632         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
11634 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11636         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
11637         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
11638         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
11639         of vinfo_for_stmt.
11640         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
11641         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
11642         (vect_update_vf_for_slp, vect_analyze_loop_operations)
11643         (vect_is_slp_reduction, vectorizable_induction)
11644         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
11645         * tree-vect-patterns.c (vect_init_pattern_stmt):
11646         (vect_determine_min_output_precision_1, vect_determine_precisions)
11647         (vect_pattern_recog): Likewise.
11648         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
11649         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
11650         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
11651         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11652         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
11653         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
11654         info field from a loop to a loop_vec_info.
11656 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11658         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
11659         (vec_info::add_stmt): Declare.
11660         * tree-vectorizer.c (vec_info::add_stmt): New function.
11661         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
11662         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
11663         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11664         (vectorizable_induction): Likewise.
11665         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
11666         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11667         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
11668         (vectorizable_load): Likewise.
11669         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11670         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
11671         (vect_recog_gather_scatter_pattern): Likewise.
11672         (append_pattern_def_seq): Likewise.  Remove a check that is
11673         performed by add_stmt itself.
11675 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11677         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
11678         which make_ssa_name was called with new_stmt before new_stmt
11679         had been created.
11681 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11683         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
11684         split out from...
11685         (vect_is_slp_reduction): ...here...
11686         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
11687         that are already known to be false.
11689 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11691         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
11692         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
11693         STMT_VINFO_NUM_SLP_USES when it's true.
11694         (vect_free_slp_instance): Add a final_p parameter and pass it to
11695         vect_free_slp_tree.
11696         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
11697         (vect_analyze_slp_instance): Likewise.
11698         (vect_slp_analyze_operations): Likewise.
11699         (vect_slp_analyze_bb_1): Likewise.
11700         * tree-vectorizer.c (vec_info): Likewise.
11701         * tree-vect-loop.c (vect_transform_loop): Likewise.
11703 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11705         * tree-vect-loop.c (vectorizable_reduction): Assert that the
11706         function is not called for second and subsequent members of
11707         a reduction group.
11709 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11711         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
11712         cases for nested loops from here to ...
11713         (vect_create_epilog_for_reduction): ...here.  Only call
11714         vect_is_simple_use for inner-loop reductions.
11716 2018-07-31  Martin Liska  <mliska@suse.cz>
11718         PR gcov-profile/85338
11719         PR gcov-profile/85350
11720         PR gcov-profile/85372
11721         * profile.c (struct location_triplet): New.
11722         (struct location_triplet_hash): Likewise.
11723         (output_location): Do not output a BB that
11724         is already recorded for a line.
11725         (branch_prob): Use streamed_locations.
11727 2018-07-31  Martin Liska  <mliska@suse.cz>
11729         PR gcov-profile/85370
11730         * coverage.c (coverage_begin_function): Do not mark target
11731         clones as artificial functions.
11733 2018-07-31  Martin Liska  <mliska@suse.cz>
11735         PR gcov-profile/83813
11736         PR gcov-profile/84758
11737         PR gcov-profile/85217
11738         PR gcov-profile/85332
11739         * profile.c (branch_prob): Do not record GOTO expressions
11740         for GIMPLE statements which locations are already streamed.
11742 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11744         * gcc.c (handle_spec_function): Accept a soft_matched_part
11745         argument, as do_spec_1.  Pass it down to ...
11746         (eval_spec_function): Accept a soft_matched_part argument,
11747         and pass it down to ...
11748         (do_spec_2): Accept a soft_matched_part argument, and pass
11749         it down to do_spec_1.
11750         (do_spec_1): Pass soft_matched_part to handle_spec_function.
11751         (handle_braces): Update call to handle_spec_function.
11752         (driver::set_up_specs): Update calls to do_spec_2.
11753         (compare_debug_dump_opt_spec_function): Likewise.
11754         (compare_debug_self_opt_spec_function): Likewise.
11756 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11758         * common.opt (nolibc): New option.
11759         * doc/invoke.texi (Link Options): Document it.
11760         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
11761         * config/alpha/linux.h: Likewise.
11762         * config/arc/elf.h: Likewise.
11763         * config/arm/uclinux-elf.h: Likewise.
11764         * config/arm/unknown-elf.h: Likewise.
11765         * config/avr/avrlibc.h: Likewise.
11766         * config/bfin/bfin.h: Likewise.
11767         * config/bfin/linux.h: Likewise.
11768         * config/bfin/uclinux.h: Likewise.
11769         * config/darwin.h: Likewise.
11770         * config/darwin10.h: Likewise.
11771         * config/darwin12.h: Likewise.
11772         * config/gnu-user.h: Likewise.
11773         * config/lm32/uclinux-elf.h: Likewise.
11774         * config/pa/pa-hpux11.h: Likewise.
11775         * config/pa/pa64-hpux.h: Likewise.
11776         * config/sparc/sparc.h: Likewise.
11778 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11780         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
11781         undefined variables.
11783 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11785         PR target/86640
11786         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
11787         instead of GEN_INT.
11789 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11791         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
11792         terminated string literal.
11794 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11796         PR rtl-optimization/85160
11797         * combine.c (is_just_move): New function.
11798         (try_combine): Allow combining two instructions into two if neither of
11799         the original instructions was a move.
11801 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
11803         PR target/86673
11804         * doc/extend.texi (Global Register Variables): Discourage use of type
11805         qualifiers.
11806         (Local Register Variables): Likewise.
11808 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
11810         PR tree-optimization/86506
11811         * hwint.h (ceil_log2): Resync with hwint.c implementation.
11813 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
11815         PR target/86547
11816         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
11817         hard_regno, make sure no insn between `from` and `to` clobbers it.
11819 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
11820             Tom de Vries  <tdevries@suse.de>
11822         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
11823         (PTX_DEFAULT_RUNTIME_DIM): ... this.
11824         (nvptx_goacc_validate_dims): Set default worker and gang dims to
11825         PTX_DEFAULT_RUNTIME_DIM.
11826         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
11828 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
11830         * config/pa/pa.c (pa_output_addr_vec): Align address table.
11831         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
11832         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
11834 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
11836         * config/rs6000/constraints.md (wG constraint): Delete, no longer
11837         used.
11838         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
11839         predicate to reflect toc fusion has been deleted.
11840         (toc_fusion_mem_raw): Delete, no longer used.
11841         (toc_fusion_mem_wrapped): Likewise.
11842         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
11843         fusion mask bit.
11844         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
11845         Delete, no longer used.
11846         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
11847         meant to be used for toc fusion.
11848         (rs6000_debug_print_mode): Delete toc fusion debugging.
11849         (rs6000_debug_reg_global): Likewise.
11850         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
11851         fusion and secondary reload support that were never used.
11852         (rs6000_option_override_internal): Delete TOC fusion, that was only
11853         partially defined, and it did not work unless you also used the
11854         -mcmodel= switch.
11855         (rs6000_legitimate_address_p): Delete TOC fusion support.
11856         (rs6000_opt_masks): Likewise.
11857         (fusion_wrap_memory_address): Delete function, no longer used.
11858         (fusion_split_address); Delete TOC fusion support.
11859         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
11860         longer used with toc fusion being deleted.
11861         (TARGET_TOC_FUSION_FP): Likewise.
11862         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
11863         UNSPEC.
11864         (toc fusion spliter): Delete TOC fusion support.
11865         (toc_fusionload_<mode>): Likewise.
11866         (toc_fusionload_di): Likewise.
11867         (fusion_gpr_load_<mode>): Delete generator function, this insn no
11868         longer needs to be named.  Rename predicate to delete TOC fusion.
11869         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11870         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11871         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11872         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11873         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
11875 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11877         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
11878         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
11879         __int128 in built-in function prototypes.
11880         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
11881         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
11883 2018-07-27  Martin Sebor  <msebor@redhat.com>
11885         PR tree-optimization/86696
11886         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
11887         types, including enums.
11888         (handle_char_store): Be prepared for the above function to fail.
11890 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
11892         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
11893         when optimization level is lower than 2 or optimize for size.
11895 2018-07-26  Martin Sebor  <msebor@redhat.com>
11897         PR tree-optimization/86043
11898         PR tree-optimization/86042
11899         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
11900         (get_string_cst_length): Rename...
11901         (get_min_string_length): ...to this.  Add argument.
11902         (handle_char_store): Extend to handle multi-character stores by
11903         MEM_REF.
11904         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
11905         * tree.h (initializer_zerop): Add argument.
11907 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
11909         PR middle-end/86660
11910         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
11911         declare target to variables if they have always,{to,from,tofrom} map
11912         kinds.
11914 2018-07-26  Martin Liska  <mliska@suse.cz>
11916         PR lto/86548
11917         * lto-wrapper.c: Add linker_output as prefix
11918         for ltrans_output_file.
11920 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
11922         PR rtl-optimization/85805
11923         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
11924         value for hard registers if that was written in the same mode.
11926 2018-07-26  Martin Liska  <mliska@suse.cz>
11928         PR gcov-profile/86536
11929         * gcov.c (format_gcov): Use printf format %.*f directly
11930         and do not handle special values.
11932 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11934         * common/config/arc/arc-common.c (arc_option_optimization_table):
11935         Update default optimizations for size.
11937 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11939         * config/arc/arc.md (movsf_insn): Add short instruction selection.
11940         * config/arc/constraints.md (CfZ): New constraint.
11941         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
11942         (subsf3_fpu): Likewise.
11943         (cmpsf_fpu): Likewise.
11944         (cmpsf_fpu_uneq): Likewise.
11946 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11948         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
11949         accesses as well.
11950         (arc_is_uncached_mem_p): uncached applies to both the variable and
11951         the pointer.
11953 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11955         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
11956         register names.
11958 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
11960         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
11961         field "m_scopes" from vec to auto_vec.
11963 2018-07-25  Martin Liska  <mliska@suse.cz>
11965         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
11966         return type.
11968 2018-07-25  Richard Biener  <rguenther@suse.de>
11970         PR debug/86654
11971         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
11972         special wrt context_die late.
11973         (gen_subprogram_die): Re-use DIEs in local scope.
11975 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
11977         PR tree-optimization/86644
11978         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
11980 2018-07-25  Martin Liska  <mliska@suse.cz>
11982         PR middle-end/86645
11983         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
11984         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
11986 2018-07-25  Martin Liska  <mliska@suse.cz>
11988         PR sanitizer/79635
11989         * params.def: Explain ASan abbreviation and provide
11990         a documentation link.
11992 2018-07-24  Martin Sebor  <msebor@redhat.com>
11994         PR tree-optimization/86622
11995         PR tree-optimization/86532
11996         * builtins.h (string_length): Declare.
11997         * builtins.c (c_strlen): Correct handling of non-constant offsets.
11998         (check_access): Be prepared for non-constant length ranges.
11999         (string_length): Make extern.
12000         * expr.c (string_constant): Only handle the minor non-constant
12001         array index.  Use string_constant to compute the length of
12002         a generic string constant.
12004 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
12006         PR tree-optimization/86618
12007         * tree-vect-stmts.c (vectorizable_call): Don't take the address
12008         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
12010 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
12012         PR tree-optimization/86636
12013         * json.cc (json::object::set): Fix comment.  Add assertions.
12014         (json::array::append): Move here from json.h.  Add comment and an
12015         assertion.
12016         (json::string::string): Likewise.
12017         * json.h (json::array::append): Move to json.cc.
12018         (json::string::string): Likewise.
12019         * optinfo-emit-json.cc
12020         (optrecord_json_writer::impl_location_to_json): Assert that we
12021         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
12022         wrapper around it.  Expand the location once, rather than three
12023         times.
12024         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
12025         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
12026         wrappers.
12027         (optrecord_json_writer::optinfo_to_json): Likewise, in four
12028         places.  Fix some overlong lines.
12030 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
12032         * config/aarch64/aarch64-simd.md
12033         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
12034         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
12035         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
12036         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
12037         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
12038         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
12039         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
12040         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
12041         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
12043 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
12045         PR middle-end/86627
12046         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
12047         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
12048         and abs_d == d, do the power of two handling if profitable.
12050 2018-07-24  Richard Biener  <rguenther@suse.de>
12052         * match.pd: Add BIT_FIELD_REF canonicalizations.
12054 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12056         PR c/86617
12057         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
12059 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12061         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
12062         terminated STRING_CST object.
12064 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12066         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
12068 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
12070         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
12071         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
12072         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
12073         the elements into a register.
12074         (rs6000_split_v4si_init_di_reg): Delete.
12075         (rs6000_split_v4si_init): Delete.
12076         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
12077         (vsx_init_v4si): Rewrite as a define_expand.
12079 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
12081         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
12082         zero_extend argument from memory): New.
12084 2018-07-22  Martin Sebor  <msebor@redhat.com>
12086         PR bootstrap/86621
12087         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
12088         diagnosing calls with unknown arguments unless -Walloca-larger-than
12089         is restricted to less than PTRDIFF_MAX bytes.
12091 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
12093         * doc/gcov.texi (Invoking Gcov): Editorial changes.
12095 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
12097         * pretty-print.c (text_info::set_location): Remove redundant
12098         "line_table" parameter from call to rich_location::set_range.
12100 2018-07-20  Martin Sebor  <msebor@redhat.com>
12102         PR middle-end/82063
12103         * builtins.c (expand_builtin_alloca): Adjust.
12104         * calls.c (alloc_max_size): Simplify.
12105         * cgraphunit.c (cgraph_node::expand): Adjust.
12106         * common.opt (larger_than_size, warn_frame_larger_than): Remove
12107         variables.
12108         (frame_larger_than_size): Same.
12109         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
12110         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
12111         Initialize.
12112         * doc/invoke.texi (GCC Command Options): Document option arguments.
12113         Explain byte-size arguments and suffixes.
12114         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
12115         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
12116         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
12117         * doc/options.texi (UInteger): Expand.
12118         (Host_Wide_Int, ByteSize): Document new properties.
12119         * final.c (final_start_function_1): Include sizes in an error message.
12120         * function.c (frame_offset_overflow): Same.
12121         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
12122         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
12123         Diagnose unbounded alloca calls only for limits of less than
12124         PTRDIFF_MAX.
12125         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
12126         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
12127         for alloca(0).
12128         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
12129         only for limits of less than PTRDIFF_MAX.
12130         * langhooks-def.h (lhd_handle_option): Change function argument
12131         to HOST_WIDE_INT.
12132         * langhooks.c (lhd_handle_option): Same.
12133         * langhooks.h (handle_option): Same.
12134         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
12135         ByteSize flags.
12136         (var_type, var_type_struct): Same.
12137         (var_set): Handle ByteSize flag.
12138         * optc-gen.awk: Add comments to output to ease debugging.  Make
12139         use of HOST_WIDE_INT where appropriate.
12140         * opts-gen-save.awk:  Use %lx to format unsigned long.
12141         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
12142         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
12143         arguments.  Parse bytes-size suffixes.
12144         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
12145         (enum_value_to_arg): Same.
12146         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
12147         (handle_option): Adjust.
12148         (generate_option): Change function argument to HOST_WIDE_INT.
12149         (cmdline_handle_error): Adjust.
12150         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
12151         (set_option): Change function argument to HOST_WIDE_INT.
12152         (option_enabled): Handle cl_host_wide_int.
12153         (get_option_state): Handle CLVC_SIZE.
12154         (control_warning_option): Same.
12155         * opts.c (common_handle_option): Change function argument to
12156         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
12157         OPT_Wvla_larger_than_.
12158         * opts.h (enum cl_var_type): Add an enumerator.
12159         * stor-layout.c (layout_decl): Print a more meaningful warning.
12160         * toplev.c (output_stack_usage): Adjust.
12162 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
12164         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
12165         call to inline_expand_builtin_string_cmp.
12166         (expand_builtin_strcmp): Likewise.
12167         (expand_builtin_strncmp): Likewise.
12168         (inline_string_cmp): Delete the last parameter, change char_type_node
12169         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
12170         two operands.
12171         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
12172         the inlining expansion on target where the type of the call has same or
12173         narrower precision than unsigned char.
12175 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
12177         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
12178         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
12179         * common.opt (fsave-optimization-record): New option.
12180         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
12181         * doc/invoke.texi (-fsave-optimization-record): New option.
12182         * dumpfile.c: Include "optinfo-emit-json.h".
12183         (struct kv_pair): Move to coretypes.h.
12184         (optgroup_options): Make non-static.
12185         (dump_context::end_scope): Call
12186         optimization_records_maybe_pop_dump_scope.
12187         * dumpfile.h (optgroup_options): New decl.
12188         * json.cc: New file.
12189         * json.h: New file.
12190         * optinfo-emit-json.cc: New file.
12191         * optinfo-emit-json.h: New file.
12192         * optinfo.cc: Include "optinfo-emit-json.h".
12193         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
12194         (optinfo_enabled_p): Check optimization_records_enabled_p.
12195         (optinfo_wants_inlining_info_p): Likewise.
12196         * optinfo.h: Update comment.
12197         * profile-count.c (profile_quality_as_string): New function.
12198         * profile-count.h (profile_quality_as_string): New decl.
12199         (profile_count::quality): New accessor.
12200         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
12201         and optinfo_emit_json_cc_tests.
12202         * selftest.h (selftest::json_cc_tests): New decl.
12203         (selftest::optinfo_emit_json_cc_tests): New decl.
12204         * toplev.c: Include "optinfo-emit-json.h".
12205         (compile_file): Call optimization_records_finish.
12206         (do_compile): Call optimization_records_start.
12207         * tree-ssa-live.c: Include optinfo.h.
12208         (remove_unused_scope_block_p): Retain inlining information if
12209         optinfo_wants_inlining_info_p returns true.
12211 2018-07-20  Richard Biener  <rguenther@suse.de>
12213         PR debug/86585
12214         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
12215         to cover -flto-partition=none.
12217 2018-07-20  Martin Liska  <mliska@suse.cz>
12219         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
12220         (get_decl_source_range): Remove unused function.
12222 2018-07-20  Richard Biener  <rguenther@suse.de>
12224         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
12225         (struct vn_phi_s): Likewise.
12226         (struct vn_reference_s): Likewise.
12227         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
12228         for searching the slot of an entry known to be in the hash itself.
12229         (vn_phi_hasher::equal): Likewise.
12230         (vn_reference_hasher::equal): Likewise.
12231         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
12232         globals.
12233         (optimistic_info, current_info): Remove, keeping only valid_info.
12234         (vn_reference_lookup_1): Remove fallback lookup.
12235         (vn_reference_lookup_2): Likewise.
12236         (vn_nary_op_lookup_1): Likewise.
12237         (vn_phi_lookup): Likewise.
12238         (vn_nary_build_or_lookup_1): Make sure to not chain the built
12239         hash element.
12240         (vn_reference_insert): Adjust, chain the inserted hash element
12241         at last_inserted_ref.
12242         (vn_reference_insert_pieces): Likewise.
12243         (visit_reference_op_call): Likewise.
12244         (vn_nary_op_insert_into): Chain the inserted hash element at
12245         last_inserted_nary.
12246         (vn_nary_op_insert_pieces): Adjust.
12247         (vn_nary_op_insert): Likewise.
12248         (vn_nary_op_insert_stmt): Likewise.
12249         (vn_phi_insert): Adjust, chain the inserted hash element at
12250         last_inserted_phi.
12251         (process_scc): Remove clearing and copying the optimistic
12252         table.  Instead remove elements inserted during an optimistic
12253         iteration from the single table we maintain.
12254         (init_scc_vn): Adjust.
12255         (free_scc_vn): Likewise.
12256         (sccvn_dom_walker::record_cond): Likewise.
12257         (sccvn_dom_walker::after_dom_children): Likewise.
12259 2018-07-19  Martin Sebor  <msebor@redhat.com>
12261         PR tree-optimization/84047
12262         PR tree-optimization/83776
12263         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
12264         (check_array_bounds): Call it.
12266 2018-07-19  Martin Sebor  <msebor@redhat.com>
12268         * align.h (align_flags): Use member initialization.
12270 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
12272         * Makefile.in (OBJS): Add optinfo.o.
12273         * coretypes.h (class symtab_node): New forward decl.
12274         (struct cgraph_node): New forward decl.
12275         (class varpool_node): New forward decl.
12276         * dump-context.h: New file.
12277         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
12278         "tree-pass.h".
12279         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
12280         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
12281         (set_alt_dump_file): Likewise.
12282         (dump_context::~dump_context): New dtor.
12283         (dump_gimple_stmt): Move implementation to...
12284         (dump_context::dump_gimple_stmt): ...this new member function.
12285         Add the stmt to any pending optinfo, creating one if need be.
12286         (dump_gimple_stmt_loc): Move implementation to...
12287         (dump_context::dump_gimple_stmt_loc): ...this new member function.
12288         Start a new optinfo and add the stmt to it.
12289         (dump_gimple_expr): Move implementation to...
12290         (dump_context::dump_gimple_expr): ...this new member function.
12291         Add the stmt to any pending optinfo, creating one if need be.
12292         (dump_gimple_expr_loc): Move implementation to...
12293         (dump_context::dump_gimple_expr_loc): ...this new member function.
12294         Start a new optinfo and add the stmt to it.
12295         (dump_generic_expr): Move implementation to...
12296         (dump_context::dump_generic_expr): ...this new member function.
12297         Add the tree to any pending optinfo, creating one if need be.
12298         (dump_generic_expr_loc): Move implementation to...
12299         (dump_context::dump_generic_expr_loc): ...this new member
12300         function.  Add the tree to any pending optinfo, creating one if
12301         need be.
12302         (dump_printf): Move implementation to...
12303         (dump_context::dump_printf_va): ...this new member function.  Add
12304         the text to any pending optinfo, creating one if need be.
12305         (dump_printf_loc): Move implementation to...
12306         (dump_context::dump_printf_loc_va): ...this new member function.
12307         Start a new optinfo and add the stmt to it.
12308         (dump_dec): Move implementation to...
12309         (dump_context::dump_dec): ...this new member function.  Add the
12310         value to any pending optinfo, creating one if need be.
12311         (dump_context::dump_symtab_node): New member function.
12312         (dump_context::get_scope_depth): New member function.
12313         (dump_context::begin_scope): New member function.
12314         (dump_context::end_scope): New member function.
12315         (dump_context::ensure_pending_optinfo): New member function.
12316         (dump_context::begin_next_optinfo): New member function.
12317         (dump_context::end_any_optinfo): New member function.
12318         (dump_context::s_current): New global.
12319         (dump_context::s_default): New global.
12320         (dump_scope_depth): Delete global.
12321         (dumpfile_ensure_any_optinfo_are_flushed): New function.
12322         (dump_symtab_node): New function.
12323         (get_dump_scope_depth): Reimplement in terms of dump_context.
12324         (dump_begin_scope): Likewise.
12325         (dump_end_scope): Likewise.
12326         (selftest::temp_dump_context::temp_dump_context): New ctor.
12327         (selftest::temp_dump_context::~temp_dump_context): New dtor.
12328         (selftest::verify_item): New function.
12329         (ASSERT_IS_TEXT): New macro.
12330         (ASSERT_IS_TREE): New macro.
12331         (ASSERT_IS_GIMPLE): New macro.
12332         (selftest::test_capture_of_dump_calls): New test.
12333         (selftest::dumpfile_c_tests): Call it.
12334         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
12335         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
12336         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
12337         descriptive comment.
12338         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
12339         (dump_node, dump_bb): Move these unrelated decls.
12340         (class dump_manager): Add leading comment.
12341         * optinfo.cc: New file.
12342         * optinfo.h: New file.
12344 2018-07-19  Michael Collison  <michael.collison@arm.com>
12345             Richard Henderson <rth@redhat.com>
12347         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
12348         (subti): Handle op1 zero.
12349         (subvti4, usub4ti4): New.
12350         (*sub<GPI>3_compare1_imm): New.
12351         (sub<GPI>3_carryinCV): New.
12352         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
12353         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
12355 2018-07-19  Michael Collison  <michael.collison@arm.com>
12356             Richard Henderson <rth@redhat.com>
12358         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
12359         (addti3): Create simpler code if low part is already known to be 0.
12360         (addvti4, uaddvti4): New.
12361         (*add<GPI>3_compareC_cconly_imm): New.
12362         (*add<GPI>3_compareC_cconly): New.
12363         (*add<GPI>3_compareC_imm): New.
12364         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
12365         handle constants within this pattern..
12366         (*add<GPI>3_compareV_cconly_imm): New.
12367         (*add<GPI>3_compareV_cconly): New.
12368         (*add<GPI>3_compareV_imm): New.
12369         (add<GPI>3_compareV): New.
12370         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
12371         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
12372         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
12373         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
12374         with 'comparison' operator.
12375         (*add<GPI>3_compareV_cconly_imm): Ditto.
12376         (*add<GPI>3_compareV_cconly): Ditto.
12377         (*add<GPI>3_compareV_imm): Ditto.
12378         (add<GPI>3_compareV): Ditto.
12379         (add<mode>3_carryinC): Ditto.
12380         (*add<mode>3_carryinC_zero): Ditto.
12381         (*add<mode>3_carryinC): Ditto.
12382         (add<mode>3_carryinV): Ditto.
12383         (*add<mode>3_carryinV_zero): Ditto.
12384         (*add<mode>3_carryinV): Ditto.
12386 2018-07-19  Michael Collison  <michael.collison@arm.com>
12387             Richard Henderson <rth@redhat.com>
12389         * config/aarch64/aarch64-modes.def (CC_V): New.
12390         * config/aarch64/aarch64-protos.h
12391         (aarch64_addti_scratch_regs): Declare
12392         (aarch64_subvti_scratch_regs): Declare.
12393         (aarch64_expand_subvti): Declare.
12394         (aarch64_gen_unlikely_cbranch): Declare
12395         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
12396         for signed overflow using CC_Vmode.
12397         (aarch64_get_condition_code_1): Handle CC_Vmode.
12398         (aarch64_gen_unlikely_cbranch): New function.
12399         (aarch64_addti_scratch_regs): New function.
12400         (aarch64_subvti_scratch_regs): New function.
12401         (aarch64_expand_subvti): New function.
12403 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12405         * config/aarch64/aarch64-option-extensions.def: New entry for profile
12406         extension.
12407         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
12408         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
12409         extension.
12411 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12413         PR target/83009
12414         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12415         address check not strict.
12417 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12419         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
12420         Umq with Umn.
12421         (store_pair_lanes<mode>): Likewise.
12422         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
12423         enum value 'ADDR_QUERY_LDP_STP_N'.
12424         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
12425         (aarch64_print_address_internal): Add declaration.
12426         (aarch64_print_ldpstp_address): Remove.
12427         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
12428         (aarch64_print_operand): Change printing of 'y'.
12429         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
12430         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
12431         'true' rather than '1'.
12432         * config/aarch64/constraints.md (Uml): Likewise.
12433         (Uml): Rename to Umn.
12434         (Umq): Remove.
12436 2018-07-19  Richard Biener  <rguenther@suse.de>
12438         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
12439         a trailing array.
12440         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
12441         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
12442         (vn_reference_hasher): Likewise.
12443         (struct vn_tables_s): Remove obstack and alloc-pool members.
12444         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
12445         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
12446         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
12447         (vn_reference_insert_pieces): Likewise.
12448         (alloc_vn_nary_op_noinit): Adjust.
12449         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
12450         (vn_phi_eq): Adjust.
12451         (shared_lookup_phiargs): Remove.
12452         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
12453         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
12454         (visit_reference_op_call): Likewise.
12455         (copy_nary, copy_phi, copy_reference): Remove.
12456         (process_scc): Rewind the obstack when iterating.  Do not
12457         copy the elements to valid_info but just move them from one
12458         hashtable to the other.
12459         (allocate_vn_table): Adjust.
12460         (free_vn_table): Likewise.
12461         (init_scc_vn): Likewise.
12462         (free_scc_vn): Likewise.
12464 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
12466         PR target/86560
12467         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
12468         indirect_return as function type attribute.
12469         (ix86_attribute_table): Change indirect_return to function
12470         type attribute.
12471         * doc/extend.texi: Update indirect_return attribute.
12473 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
12475         * wide-int.h (widest2_int): New.
12476         * gimple-fold.c (arith_overflowed_p): Use it.
12477         * tree.h (widest2_int_cst): New.
12478         * tree-vrp.c (wide_int_binop_overflow): Rename from
12479         vrp_int_const_binop.
12480         Rewrite to work on trees.
12481         (extract_range_from_multiplicative_op_1): Abstract code to...
12482         (wide_int_range_min_max): ...here.
12483         (wide_int_range_cross_product): ...and here.
12484         (extract_range_from_binary_expr_1): Abstract overflow code to...
12485         (wide_int_range_mult_wrapping): ...here.
12486         * tree-vrp.h (wide_int_range_cross_product): New.
12487         (wide_int_range_mult_wrapping): New.
12489 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
12490             Julia Koval  <julia.koval@intel.com>
12492         * config/i386/x86-tune-costs.h (skylake_memcpy,
12493         skylake_memset): Replace rep_prefix with unrolling for size 512.
12495 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
12497         PR middle-end/86544
12498         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
12499         comparision with EQ_EXPR in last stmt.
12501 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12503         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
12504         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
12505         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
12506         previously known as "PowerPC AltiVec Built-in Functions".  Move
12507         some material to new subsubsections "PowerPC AltiVec Built-in
12508         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
12509         ISA 2.07".
12510         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
12511         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
12512         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
12513         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
12515 2018-07-18  Richard Biener  <rguenther@suse.de>
12517         PR tree-optimization/86557
12518         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
12519         EXACT_DIV_EXPR.
12521 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
12523         * config/s390/s390.c (s390_function_profiler): Generate CFI.
12525 2018-07-17  Jeff Law  <law@redhat.com>
12527         * config/arm/arm.c (get_label_padding): Update for recent
12528         changes to label_to_alignment.
12530         PR tree-optimization/86010
12531         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
12533         * config/mips/mips.c (vr4130_align_insns): Update for recent
12534         changes to label_to_alignment.
12536         * config/frv/frv.c (frv_label_align): Update for recent changes
12537         to label_to_alignment.
12539         * config/nios2/nios2.c (nios2_label_align): Update for recent
12540         changes which dropped ALIGN_LABELS_LOG.
12542 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
12544         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
12545         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
12547 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
12549         * config/arc/arc.c (arc_label_align): Use align_labels instead of
12550         deprecated align_labels_log.
12552 2018-07-17  Richard Biener  <rguenther@suse.de>
12554         PR lto/86456
12555         * dwarf2out.c (init_sections_and_labels): Always generate
12556         a debug_line_str_section for early LTO debug.
12557         (dwarf2out_finish): Reset debug_line_str_hash output early.
12558         Bump counter for extra dwarf5 .debug_loc labels to not conflict
12559         with fat LTO part.
12560         (dwarf2out_early_finish): Output debug_line_str.
12562 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12564         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
12565         index register on z196 or later.
12567 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12569         * config/s390/s390.c (s390_default_align): Set default function
12570         alignment to 16.
12571         (s390_override_options_after_change): Call s390_default align.
12572         (s390_option_override_internal): Call s390_default_align.
12573         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
12575 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12577         PR middle-end/86542
12578         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
12579         to _looptemp_ clauses, other than the first two.
12581 2018-07-17  Martin Liska  <mliska@suse.cz>
12583         * opts.c: Do not enable OPT_falign_* for -Os.
12585 2018-07-17  Martin Liska  <mliska@suse.cz>
12587         * align.h (MAX_CODE_ALIGN): New.
12588         (MAX_CODE_ALIGN_VALUE): New.
12589         * common/config/i386/i386-common.c (ix86_handle_option):
12590         (MAX_CODE_ALIGN): Moved to align.h.
12591         * final.c (MAX_CODE_ALIGN): Likewise.
12592         * opts.c (parse_and_check_align_values):
12593         (MAX_CODE_ALIGN): Likewise.
12594         (MAX_CODE_ALIGN_VALUE): Likewise.
12596 2018-07-17  Martin Liska  <mliska@suse.cz>
12598         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
12599         in order to fulfil coding style.
12600         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
12601         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12602         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12603         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
12604         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
12605         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
12606         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
12607         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
12608         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12610 2018-07-17  Martin Liska  <mliska@suse.cz>
12612         * align.h: New file.
12613         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
12614         directly.
12615         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
12616         align_flags of label_to_alignment.
12617         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
12618         align_flags class.
12619         * config/m68k/m68k.c: Do not use removed align_labels_value and
12620         align_loops_value.
12621         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
12622         (LOOP_ALIGN): Likewise.
12623         (LABEL_ALIGN): Likewise.
12624         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12625         Remove not used macro.
12626         (rs6000_loop_align): Change return type to align_flags.
12627         (rs6000_loop_align_max_skip): Remove.
12628         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
12629         Change return type to align_flags.
12630         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12631         Remove not used macro.
12632         (rs6000_loop_align):  Change return type to align_flags.
12633         (rs6000_loop_align_max_skip): Remove.
12634         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
12635         * config/rx/rx-protos.h (rx_align_for_label): Make it
12636         static function.
12637         * config/rx/rx.c (rx_align_for_label): Change return type
12638         to align_flags.
12639         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
12640         macro definitions.
12641         into align_flags class.
12642         (LABEL_ALIGN): Likewise.
12643         (LOOP_ALIGN): Likewise.
12644         * config/s390/s390.c (s390_label_align): Use align_flags
12645         class member.
12646         (s390_asm_output_function_label): Likewise.
12647         * config/sh/sh.c (sh_override_options_after_change):
12648         Use align_flags class directly without macros.
12649         (find_barrier): Likewise.
12650         (barrier_align): Likewise.
12651         (sh_loop_align): Likewise.
12652         * config/spu/spu.c (spu_option_override):
12653         Use align_flags_tuple::get_value instead of removed macros.
12654         (spu_sched_init): Likewise.
12655         * config/spu/spu.h (GTY): Likewise.
12656         * config/visium/visium.c (visium_option_override):
12657         Set "8" as default secondary alignment.
12658         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
12659         in order to guarantee secondary alignment of 8.
12660         * coretypes.h: Include align.h header file.
12661         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
12662         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
12663         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
12664         * doc/tm.texi.in: Likewise.
12665         * final.c (struct label_alignment): Remove not used structure.
12666         (LABEL_ALIGN): Change type to align_flags.
12667         (LOOP_ALIGN): Likewise.
12668         (JUMP_ALIGN): Likewise.
12669         (default_loop_align_max_skip): Remove.
12670         (default_label_align_max_skip): Likewise.
12671         (default_jump_align_max_skip): Likewise.
12672         (default_label_align_after_barrier_max_skip):
12673         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
12674         (LABEL_TO_MAX_SKIP): Remove.
12675         (label_to_alignment): Return align_flags type instead of integer.
12676         (label_to_max_skip): Remove.
12677         (align_fuzz): Use align_flags type.
12678         (compute_alignments): Use align_flags type and use align_flags::max
12679         to combine multiple alignments.
12680         (grow_label_align): Grow vec instead of C array.
12681         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
12682         (shorten_branches):  Use align_flags type and use align_flags::max
12683         to combine multiple alignments.
12684         (final_scan_insn_1): Remove usage of secondary alignment that comes
12685         from label alignment, but instead use proper secondary alignment
12686         which is computed in grow_label_align.
12687         * flags.h (struct align_flags_tuple): Move to align.h.
12688         (struct align_flags): Likewise.
12689         (state_align_loops): Rename to align_loops.
12690         (state_align_jumps): Rename to align_jumps.
12691         (state_align_labels): Rename to align_labels.
12692         (state_align_functions): Rename to align_functions.
12693         (align_loops_log): Remove.
12694         (align_jumps_log): Remove.
12695         (align_labels_log): Remove.
12696         (align_functions_log): Remove.
12697         (align_loops_max_skip): Remove.
12698         (align_jumps_max_skip): Remove.
12699         (align_labels_max_skip): Remove.
12700         (align_functions_max_skip): Remove.
12701         (align_loops_value): Remove.
12702         (align_jumps_value): Remove.
12703         (align_labels_value): Remove.
12704         (align_functions_value): Remove.
12705         * output.h (label_to_alignment): Change return type to align_flags.
12706         (label_to_max_skip): Remove.
12707         * target.def: Remove loop_align_max_skip, label_align_max_skip,
12708         jump_align_max_skip macros.
12709         * targhooks.h (default_loop_align_max_skip): Remove.
12710         (default_label_align_max_skip): Likewise.
12711         (default_jump_align_max_skip): Likewise.
12712         (default_label_align_after_barrier_max_skip): Remove.
12713         * toplev.c (read_log_maxskip): Use ::normalize function.
12714         (parse_N_M): Remove not used argument and also call ::normalize.
12715         (parse_alignment_opts): Do not pass unused arguments.
12716         * varasm.c (assemble_start_function): Use directly align_functions
12717         instead of removed macros.
12718         * system.h: Do not poison removed macros.
12720 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12722         PR middle-end/86539
12723         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
12724         and cond temporaries don't have reference type if iterator has
12725         pointer type.  For init use &for_pre_body instead of pre_p if
12726         for_pre_body is non-empty.
12728 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12730         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
12731         double-double modes to SFmode directly directly.
12732         (trunc<mode>sf2_fprs): Delete.
12734 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12736         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
12737         for conversions between IFmode and the decimal floating point modes.
12738         (init_float128_ieee): Use the correct names for conversions between
12739         KFmode and the decimal floating point modes.
12741 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12743         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
12744         for the conversions between TDmode and IFmode.
12745         (init_float128_ieee): Use more correct names for the conversions
12746         between TDmode and KFmode.
12748 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
12750         PR tree-optimization/86526
12751         * builtins.c (expand_builtin_memcmp): Formatting fixes.
12752         (inline_expand_builtin_string_cmp): Likewise.
12753         (inline_string_cmp): Likewise.  Use c_readstr instead of
12754         builtin_memcpy_read_str.  Add unit_mode temporary.
12756 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12758         PR middle-end/86528
12759         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
12760         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
12762 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12764         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
12765         Alphabetize prototypes of built-in functions, separating out
12766         built-in functions that are listed in this section but should be
12767         described elsewhere.
12769 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
12771         PR target/86511
12772         * expmed.c (emit_store_flag): Do not emit setcc followed by a
12773         conditional move when trapping comparison was split to a
12774         non-trapping one (and vice versa).
12776 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12778         * config/s390/s390.c (s390_function_profiler): Generate nops
12779         instead of profiler call sequences.
12780         * config/s390/s390.opt: Add the new option.
12782 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12784         * config/s390/s390.c (s390_function_profiler): Generate
12785         __mcount_loc section.
12786         * config/s390/s390.opt: Add the new option.
12788 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12790         * common.opt: Add the new warning.
12791         * config/s390/s390.c (s390_function_profiler): Emit "brasl
12792         %r0,__fentry__" when -mfentry is specified.
12793         (s390_option_override_internal): Disallow -mfentry for 31-bit
12794         CPUs.
12795         * config/s390/s390.opt: Add the new option.
12797 2018-07-16  Richard Biener  <rguenther@suse.de>
12799         PR lto/86523
12800         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
12801         for function-local FUNCTION_DECL and RESULT_DECL immediately.
12803 2018-07-16  Martin Liska  <mliska@suse.cz>
12805         PR ipa/86529
12806         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
12807         to ::get_create.
12809 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12811         * config/arc/arcHS.md: Update ARCHS scheduling rules.
12813 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12815         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
12816         for ARCHS4x.
12817         * config/arc/arc-cpus.def (hs4x): New cpu.
12818         (hs4xd): Likewise.
12819         * config/arc/arc-tables.opt: Regenerate.
12820         * config/arc/arc.c (arc_sched_issue_rate): New function.
12821         (TARGET_SCHED_ISSUE_RATE): Define.
12822         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
12823         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
12824         fpu_cvt.
12825         (attr tune): Add ARCHS4x tune values.
12826         (attr tune_dspmpy): Define.
12827         (*tst): Correct instruction type.
12828         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
12829         * config/arc/arcHS4x.md: New file.
12830         * config/arc/fpu.md: Update instruction type attributes.
12831         * config/arc/t-multilib: Regenerate.
12833 2018-07-16  Tom de Vries  <tdevries@suse.de>
12835         PR debug/86455
12836         * var-tracking.c (vt_initialize): Fix pre_dec handling.
12838 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
12840         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
12841         early clobber.
12843 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
12845         PR tree-optimization/86514
12846         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
12847         conversion to a boolean type from a type with greater precision.
12849 2018-07-16  Tom de Vries  <tdevries@suse.de>
12851         * var-tracking.c (vt_initialize): Print adjusted insn slim if
12852         dump_flags request TDF_SLIM.
12854 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
12856         * fold-const.c (int_const_binop_1): Abstract...
12857         (wide_int_binop): ...wide int code here.
12858         (poly_int_binop): ...poly int code here.
12859         Abstract the rest of int_const_binop_1 into int_const_binop.
12860         * fold-const.h (wide_int_binop): New.
12861         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
12862         Remove useless PLUS/MINUS_EXPR case.
12863         (zero_nonzero_bits_from_vr): Move wide int code...
12864         (zero_nonzero_bits_from_bounds): ...here.
12865         (extract_range_from_binary_expr_1): Move mask optimization code...
12866         (range_easy_mask_min_max): ...here.
12867         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
12868         (range_easy_mask_min_max): New.
12870 2018-07-15  Jeff Law  <law@redhat.com>
12872         PR target/85993
12873         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
12874         block.
12876 2018-07-14  Jim Wilson  <jimw@sifive.com>
12878         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
12880 2018-07-14  Paul Koning  <ni1d@arrl.net>
12882         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
12884 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
12886         * lto-streamer-out.c (copy_function_or_variable): Dump info about
12887         copying section.
12889 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
12890             Steve Munroe  <munroesj52@gmail.com>
12892         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
12893         (_mm_andnot_si128): Likewise.
12894         (_mm_or_si128): Likewise.
12895         (_mm_xor_si128): Likewise.
12897 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
12899         PR middle-end/78809
12900         * builtins.c (expand_builtin_memcmp): Inline the calls first
12901         when result_eq is false.
12902         (expand_builtin_strcmp): Inline the calls first.
12903         (expand_builtin_strncmp): Likewise.
12904         (inline_string_cmp): New routine. Expand a string compare
12905         call by using a sequence of char comparison.
12906         (inline_expand_builtin_string_cmp): New routine. Inline expansion
12907         a call to str(n)cmp/memcmp.
12908         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
12909         New option.
12910         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
12912 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
12914         * config/arm/driver-arm.c: Include arm-native.h.
12915         (host_detect_local_cpu): Use auto-generated data tables.
12916         (vendors, arm_cpu_table): Delete.  Move part information to ...
12917         * config/arm/arm-cpus.in: ... here.
12918         * config/arm/parsecpu.awk (gen_native): New function.
12919         (vendor, part): New CPU fields.
12920         (END): Add support for building the native CPU detection tables.
12921         * config/arm/t-arm (arm-native.h): Add build rule.
12922         (driver-arm.o): Add dependency on arm-native.h.
12924 2018-07-13  Richard Biener  <rguenther@suse.de>
12926         PR middle-end/85974
12927         * match.pd (addr1 - addr2): Allow either of the operand to
12928         have a conversion.
12930 2018-07-13  Tom de Vries  <tdevries@suse.de>
12932         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
12933         in remap_ssa_name.
12935 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
12937         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
12938         arrays instead of numbered variables.
12940 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
12942         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
12943         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
12945 2018-07-13  Richard Biener  <rguenther@suse.de>
12947         PR debug/86452
12948         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
12949         instead of get_context_die.
12951 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
12952             Richard Biener  <rguenther@suse.de>
12954         PR middle-end/86489
12955         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
12956         that the loop latch destination where phi is defined.
12958 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
12960         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
12961         (riscv_expand_epilogue): Add assertion to check interrupt mode.
12962         (riscv_set_current_function): Extract getting interrupt type to new
12963         function.
12964         (riscv_get_interrupt_type): New function.
12965         (riscv_merge_decl_attributes): New function, checking interrupt type is
12966         same.
12967         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
12969 2018-07-12  Paul Koning  <ni1d@arrl.net>
12971         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
12972         directive.
12974 2018-07-12  Paul Koning  <ni1d@arrl.net>
12976         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
12977         zero reference, add doloop_end instead.
12978         * doc/md.texi (decrement_and_branch_until_zero): Remove.
12979         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
12980         detail for doloop_end.
12982 2018-07-12  Martin Sebor  <msebor@redhat.com>
12984         PR c/86453
12985         * attribs.c (decl_attributes): Reject conflicting attributes before
12986         calling attribute handlers.
12988 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12990         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
12991          parameter.
12992         (gcc::dump_manager::get_dump_file_name): likewise.
12993         (dump_begin): Likewise.
12994         * dumpfile.h (dump_begin): Update prototype.
12995         (gcc::dump_manager::get_dump_file_name,
12996         gcc::dump_manager::get_dump_file_name): Update prototype.
12998 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13000         * internal-fn.h (vectorizable_internal_fn_p): New function.
13001         * tree-vect-slp.c (compatible_calls_p): Likewise.
13002         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
13003         to internal functions.
13004         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
13006 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13008         * fold-const.h (inverse_conditions_p): Declare.
13009         * fold-const.c (inverse_conditions_p): New function.
13010         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
13011         that test the inverse condition of a conditional internal function.
13012         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
13013         * internal-fn.c (internal_fn_mask_index): Handle conditional
13014         internal functions.
13015         (vectorized_internal_fn_supported_p): New function.
13016         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
13017         (any_pred_load_store): Replace with...
13018         (need_to_predicate): ...this new variable.
13019         (redundant_ssa_names): New variable.
13020         (ifcvt_can_use_mask_load_store): Move initial checks to...
13021         (ifcvt_can_predicate): ...this new function.  Handle tree codes
13022         for which a conditional internal function exists.
13023         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
13024         instead of ifcvt_can_use_mask_load_store.  Update after variable
13025         name change.
13026         (predicate_load_or_store): New function, split out from
13027         predicate_mem_writes.
13028         (check_redundant_cond_expr): New function.
13029         (value_available_p): Likewise.
13030         (predicate_rhs_code): Likewise.
13031         (predicate_mem_writes): Rename to...
13032         (predicate_statements): ...this.  Use predicate_load_or_store
13033         and predicate_rhs_code.
13034         (combine_blocks, tree_if_conversion): Update after above name changes.
13035         (ifcvt_local_dce): Handle redundant_ssa_names.
13036         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
13037         general conditional functions.
13038         * tree-vect-stmts.c (vectorizable_call): Likewise.
13040 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13041             Alan Hayward  <alan.hayward@arm.com>
13042             David Sherwood  <david.sherwood@arm.com>
13044         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
13045         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
13046         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
13047         plus and minus and convert them into IFN_COND_FMA-based sequences.
13048         (convert_mult_to_fma): Handle conditional plus and minus.
13050 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13052         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
13053         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
13054         (cond_fnms_optab): New optabs.
13055         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
13056         internal functions.
13057         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
13058         * internal-fn.h (get_conditional_internal_fn): Declare.
13059         (get_unconditional_internal_fn): Likewise.
13060         * internal-fn.c (cond_ternary_direct): New macro.
13061         (expand_cond_ternary_optab_fn): Likewise.
13062         (direct_cond_ternary_optab_supported_p): Likewise.
13063         (FOR_EACH_COND_FN_PAIR): Likewise.
13064         (get_conditional_internal_fn): New function.
13065         (get_unconditional_internal_fn): Likewise.
13066         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
13067         (gimple_match_op::gimple_match_op): Add a new overload for 5
13068         operands.
13069         (gimple_match_op::set_op): Likewise.
13070         (gimple_resimplify5): Declare.
13071         * genmatch.c (decision_tree::gen): Generate simplifications for
13072         5 operands.
13073         * gimple-match-head.c (gimple_simplify): Define an overload for
13074         5 operands.  Handle calls with 5 arguments in the top-level overload.
13075         (convert_conditional_op): Handle conversions from unconditional
13076         internal functions to conditional ones.
13077         (gimple_resimplify5): New function.
13078         (build_call_internal): Pass a fifth operand.
13079         (maybe_push_res_to_seq): Likewise.
13080         (try_conditional_simplification): Try converting conditional
13081         internal functions to unconditional internal functions.
13082         Handle 3-operand unconditional forms.
13083         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
13084         Define ternary equivalents of the current rules for binary conditional
13085         internal functions.
13086         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
13087         ternary operations.
13088         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
13089         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
13090         (optab): Handle them.
13091         (SVE_COND_FP_TERNARY): New int iterator.
13092         (sve_fmla_op, sve_fmad_op): New int attributes.
13093         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
13094         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
13095         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
13097 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13099         * target.def (preferred_else_value): New target hook.
13100         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
13101         * doc/tm.texi: Regenerate.
13102         * targhooks.h (default_preferred_else_value): Declare.
13103         * targhooks.c (default_preferred_else_value): New function.
13104         * internal-fn.h (conditional_internal_fn_code): Declare.
13105         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
13106         (get_conditional_internal_fn): Use it.
13107         (conditional_internal_fn_code): New function.
13108         * gimple-match.h (gimple_match_cond): New struct.
13109         (gimple_match_op): Add a cond member function.
13110         (gimple_match_op::gimple_match_op): Update all forms to take a
13111         gimple_match_cond.
13112         * genmatch.c (expr::gen_transform): Use the same condition as res_op
13113         for the suboperation, but don't specify a particular else_value.
13114         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
13115         (visit_nary_op, visit_reference_op_load): Pass
13116         gimple_match_cond::UNCOND to the gimple_match_op constructor.
13117         * gimple-match-head.c: Include tree-eh.h
13118         (convert_conditional_op): New function.
13119         (maybe_resimplify_conditional_op): Likewise.
13120         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
13121         (gimple_resimplify2): Likewise.
13122         (gimple_resimplify3): Likewise.
13123         (gimple_resimplify4): Likewise.
13124         (maybe_push_res_to_seq): Return null for conditional operations.
13125         (try_conditional_simplification): New function.
13126         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
13127         constructor.
13128         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
13129         IFN_COND_* call.
13130         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
13131         function.
13132         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
13134 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
13136         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
13137         DECL_FCONTEXT
13138         (hash_tree): Do not hash DECL_FCONTEXT
13139         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
13140         Do not stream DECL_FCONTEXT.
13141         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
13142         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
13144 2018-07-12  Richard Biener  <rguenther@suse.de>
13146         PR debug/86462
13147         * dwarf2out.c (gen_block_die): Only output blocks when they have
13148         at least one !DECL_IGNORED_P variable.
13150 2018-07-12  Richard Biener  <rguenther@suse.de>
13152         PR target/84829
13153         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
13154         Remove -mieee-fp handling.
13156 2018-07-12  Richard Biener  <rguenther@suse.de>
13158         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
13159         left-over from last patch.
13161 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
13163         PR tree-optimization/86492
13164         * gimple-ssa-store-merging.c
13165         (imm_store_chain_info::coalesce_immediate_stores): Call
13166         check_no_overlap even for the merge_overlapping case.  Formatting fix.
13168 2018-07-12  Richard Biener  <rguenther@suse.de>
13170         PR middle-end/86479
13171         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
13172         move possibly trapping operations into the conditional.
13174 2018-07-12  Richard Biener  <rguenther@suse.de>
13176         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
13177         (vn_lookup_simplify_result): Remove recursion limit applied
13178         here.
13179         (vn_nary_build_or_lookup_1): Adjust.
13180         (try_to_simplify): Likewise.
13181         * gimple-match-head.c (gimple_resimplify1): Instead apply one
13182         here.
13183         (gimple_resimplify2): Likewise.
13184         (gimple_resimplify3): Likewise.
13185         (gimple_resimplify4): Likewise.
13187 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
13189         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
13190         Use __mmask64 type instead of __mmask8 for __M argument.
13191         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
13192         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
13193         __U argument.
13194         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
13195         __mmask16 for __M argument.
13196         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
13197         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
13198         to __mmask16 instead of __mmask8.
13199         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
13200         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
13201         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
13202         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
13203         instead of __mmask16 for __U argument.
13204         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
13205         __mmask16 instead of __mmask8 for __U argument.
13206         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
13207         __U argument.
13208         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
13209         __mmask16.
13210         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
13211         argument.
13212         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
13213         __U argument.
13214         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
13215         __mmask16.
13216         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
13217         of __mmask16.
13218         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
13219         __U argument.
13220         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
13221         __U argument.
13222         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
13223         __U argument.
13224         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
13225         __U argument.
13226         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
13227         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
13228         return type as well as __M argument type and all casts from __mmask8
13229         to __mmask32.
13230         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
13231         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
13232         return type as well as __M argument type and all casts from __mmask8
13233         to __mmask16.
13234         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
13235         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
13236         return type as well as __M argument type and all casts from __mmask8
13237         to __mmask32.
13238         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
13239         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
13240         return type as well as __M argument type and all casts from __mmask8
13241         to __mmask16.
13242         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
13243         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
13244         __mmask16.
13246 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
13248         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
13249         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
13250         for __U argument.
13252 2018-07-11  Paul Koning  <ni1d@arrl.net>
13254         * doc/md.texi (define_subst): Document how multiple occurrences of
13255         the same argument in the replacement pattern are handled.
13257 2018-07-11  Paul Koning  <ni1d@arrl.net>
13259         * doc/extend.texi (Common Variable Attributes): Move "mode" into
13260         alphabetical order.
13261         (Common Type Attributes): Add "mode" attribute.
13263 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
13265         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
13266         stream DECL_ORIGINAL_TYPE.
13267         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
13268         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
13269         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
13270         Do not walk original type.
13271         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
13272         external decls.
13273         (write_ts_decl_non_common_tree_pointers): Do not stream
13274         DECL_ORIGINAL_TYPE
13275         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
13276         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
13278 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
13280         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
13281         thread twice from the same starting edge.
13283 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
13285         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
13286         * gimple.c (gimple_call_nonnull_result_p): ...here...
13287         (gimple_call_nonnull_arg): ...and here.
13288         * gimple.h (gimple_call_nonnull_result_p): New.
13289         (gimple_call_nonnull_arg): New.
13291 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
13293         * config/arm/arm-cpus.in: Move information from fpu field of each
13294         cpu definition to the isa field.
13295         * config/arm/parsecpu.awk (fpu): Delete match rule.
13296         (gen_comm_data): Don't add bits from the CPU's FPU entry.
13298 2018-07-11  Richard Biener  <rguenther@suse.de>
13300         PR debug/86457
13301         * dwarf2out.c (init_sections_and_labels): Use
13302         output_asm_line_debug_info consistently.
13303         (dwarf2out_early_finish): Likewise.
13304         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
13305         type units.
13307 2018-07-11  Richard Biener  <rguenther@suse.de>
13309         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
13310         Rework father_bb setting in a way to avoid propagating constants
13311         multiple times on a loop body.
13313 2018-07-10  Mark Wielaard  <mark@klomp.org>
13315         PR debug/86459
13316         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
13318 2018-07-10  Richard Biener  <rguenther@suse.de>
13320         * hash-map.h (hash_map::iterator::operator*): Return
13321         references to key and value.
13323 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
13325         PR c++/86443
13326         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
13327         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
13328         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
13329         (gimplify_omp_for): For composite loops, move outer
13330         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
13331         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
13332         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
13333         TREE_LIST for both the original class iterator and the "last" helper
13334         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
13335         loop, remember has_decl_expr from outer composite loops for the
13336         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
13338 2018-07-09  Martin Sebor  <msebor@redhat.com>
13340         PR middle-end/77357
13341         PR middle-end/86428
13342         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
13343         accessing implicitly initialized array elements.
13344         * expr.c (string_constant): Handle string initializers of
13345         character arrays within aggregates.
13346         * gimple-fold.c (fold_array_ctor_reference): Add argument.
13347         Store element offset.  As a special case, handle zero size.
13348         (fold_nonarray_ctor_reference): Same.
13349         (fold_ctor_reference): Add argument.  Store subobject offset.
13350         * gimple-fold.h (fold_ctor_reference): Add argument.
13352 2018-07-09  Paul Koning  <ni1d@arrl.net>
13354         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
13355         (pdp11_insn_cost): New function.
13356         (pdp11_md_asm_adjust): New function.
13357         (TARGET_INVALID_WITHIN_DOLOOP): Define.
13358         (pdp11_rtx_costs): Update to match machine better.
13359         (output_addr_const_pdp11): Correct format mismatch warnings.
13360         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
13361         * config/pdp11/pdp11.md: General change to add base_cost and/or
13362         length attributes for use by new pdp11_insn_cost function.
13363         (MIN_BRANCH): Correct definition.
13364         (MIN_SOB): Ditto.
13365         (doloop_end): Use standard pattern name for looping pattern.
13366         (doloop_end_nocc): New.
13367         (movsf): Add another constraint alternative.
13368         (zero_extendqihi2): Add constraint alternatives for not in place
13369         extend.
13370         (zero_extendhisi2): Remove.
13371         (shift patterns): Add CC handling variants.
13372         (bswaphi2): New.
13373         (bswapsi2): New.
13374         (rothi3): New.
13375         (define_peephole2): New peephole to recognize mov that sets CC for
13376         subsequent test.
13378 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13380         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
13381         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
13382         wi::add.
13384 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
13386         PR c/86420
13387         * real.c (real_nextafter): Return true if result is denormal.
13389 2018-07-09  Martin Liska  <mliska@suse.cz>
13391         * common.opt: Add back wrongly removed attribute.
13393 2018-07-09  Richard Biener  <rguenther@suse.de>
13395         PR debug/86413
13396         * dwarf2out.c (gen_block_die): For an early generated DIE
13397         always output high/low PC attributes.
13399 2018-07-09  Tom de Vries  <tdevries@suse.de>
13401         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
13402         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
13403         onto VAR_DECL with abstract origin.
13405 2018-07-07  Jim Wilson  <jimw@sifive.com>
13407         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
13409 2018-07-07  Tom de Vries  <tdevries@suse.de>
13411         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
13413 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
13415         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
13416         overflow_type.
13417         (combine_bound): Use wide-int overflow calculation instead of
13418         rolling our own.
13419         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
13420         overflow_type.
13421         * fold-const.c (int_const_binop_2): Same.
13422         (extract_muldiv_1): Same.
13423         (fold_div_compare): Same.
13424         (fold_abs_const): Same.
13425         * match.pd: Same.
13426         * poly-int.h (add): Same.
13427         (sub): Same.
13428         (neg): Same.
13429         (mul): Same.
13430         * predict.c (predict_iv_comparison): Same.
13431         * profile-count.c (slow_safe_scale_64bit): Same.
13432         * simplify-rtx.c (simplify_const_binary_operation): Same.
13433         * tree-chrec.c (tree_fold_binomial): Same.
13434         * tree-data-ref.c (split_constant_offset_1): Same.
13435         * tree-if-conv.c (idx_within_array_bound): Same.
13436         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
13437         * tree-ssa-phiopt.c (minmax_replacement): Same.
13438         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
13439         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
13440         * vr-values.c (vr_values::adjust_range_with_scev): Same.
13441         * wide-int.cc (wi::add_large): Same.
13442         (wi::mul_internal): Same.
13443         (wi::sub_large): Same.
13444         (wi::divmod_internal): Same.
13445         * wide-int.h: Change overflow type to overflow_type for neg, add,
13446         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
13447         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
13448         mul_internal, divmod_internal.
13449         (overflow_type): New enum.
13450         (accumulate_overflow): New.
13452 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
13454         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
13455         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
13457 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
13459         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
13460         argument is checked for zero before entering loop, avoid checking again.
13462 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
13464         * gimplify.h (generic_expr_could_trap_p): Set as global function.
13465         * gimplify.h (generic_expr_could_trap_p): Likwise.
13466         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
13468 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
13470         PR tree-optimization/86401
13471         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
13472         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
13473         (fold_bit_and_mask): ... here.  New helper function for match.pd.
13474         * fold-const.h (fold_bit_and_mask): Declare.
13475         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
13477 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
13479         PR target/86324
13480         * target.def (translate_mode_attribute): New hook.
13481         * targhooks.h (default_translate_mode_attribute): Declare.
13482         * targhooks.c (default_translate_mode_attribute): New function.
13483         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
13484         * doc/tm.texi: Regenerate.
13485         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
13486         (rs6000_translate_mode_attribute): New function.
13488 2018-07-06  Paul Koning  <ni1d@arrl.net>
13490         * doc/md.texi (define_split): Document DONE and FAIL.
13491         (define_peephole2): Ditto.
13493 2018-07-05  Jeff Law  <law@redhat.com>
13495         PR tree-optimization/86010
13496         * tree-ssa-dse.c (compute_trims): More aggressively trim at
13497         both the head and tail of mem* and str* calls.
13499 2018-07-05  Jim Wilson  <jimw@sifive.com>
13501         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
13503 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
13505         * config/aarch64/aarch64-simd.md: correct flags text for
13506         MIN_EXPR replacement.
13508 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
13510         * configure: Regenerated.
13512 2018-07-05  Carl Love  <cel@us.ibm.com>
13514         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
13515         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
13516         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
13517         VSX_BUILTIN_DOUBLEL_V4SF.
13519 2018-07-05  Martin Sebor  <msebor@redhat.com>
13521         PR c++/86400
13522         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
13523         than its domain to compute its the upper bound of a char array.
13525 2018-07-05  Nathan Sidwell  <nathan@acm.org>
13527         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
13528         * doc/cpp.texi: Update comment.
13529         * doc/tm.texi: Rebuilt.
13530         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
13531         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
13532         * doc/extend.texi (Backwards Compatibility): Clarify it is system
13533         headers affected by extern "C".
13534         * system.h: Poison NO_IMPLICIT_EXTERN_C.
13535         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
13536         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
13537         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
13538         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
13539         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
13540         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
13541         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
13542         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
13543         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
13544         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
13545         config/sparc/sp64-elf.h, config/spu/spu.h,
13546         config/stormy16/stormy16.h, config/v850/v850.h,
13547         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
13548         define NO_IMPLICIT_EXTERN_C.
13549         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
13551 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13553         PR target/84711
13554         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
13555         instead of GET_MODE_SIZE when comparing Units.
13557 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13559         PR target/84711
13560         * rtlanal.c (set_noop_p): Constrain on mode change,
13561         include hard-reg-set.h
13563 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13565         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
13567 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
13569         Revert
13570         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13572         PR sanitizer/84250
13573         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13574         libasan.
13575         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13577 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13579         PR sanitizer/84250
13580         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13581         libasan.
13582         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13584 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
13586         PR middle-end/86380
13587         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
13589 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
13591         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
13592         neg_*_op* variables.
13594 2018-07-04  Martin Liska  <mliska@suse.cz>
13596         * tree-switch-conversion.c: Define
13597         max_ratio_for_speed and max_ratio_for_size constants.
13599 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13600             Martin Liska  <mliska@suse.cz>
13602         PR middle-end/66240
13603         PR target/45996
13604         PR c/84100
13605         * common.opt: Rename align options with 'str_' prefix.
13606         * common/config/i386/i386-common.c (set_malign_value): New
13607         function.
13608         (ix86_handle_option): Use it to set -falign-* options/
13609         * config/aarch64/aarch64-protos.h (struct tune_params): Change
13610         type from int to string.
13611         * config/aarch64/aarch64.c: Update default values from int
13612         to string.
13613         * config/alpha/alpha.c (alpha_override_options_after_change):
13614         Likewise.
13615         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
13616         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13617         max skip conditionally.
13618         * config/i386/freebsd.h (SUBALIGN_LOG): New.
13619         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13620         max skip conditionally.
13621         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13622         max skip conditionally.
13623         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
13624         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13625         max skip conditionally.
13626         * config/i386/i386.c (struct ptt): Change type from int to
13627         string.
13628         (ix86_default_align): Set default values.
13629         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
13630         max skip conditionally.
13631         * config/i386/iamcu.h (SUBALIGN_LOG): New.
13632         (ASM_OUTPUT_MAX_SKIP_ALIGN):
13633         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
13634         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13635         max skip conditionally.
13636         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
13637         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
13638         * config/i386/x86-64.h (SUBALIGN_LOG): New.
13639         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13640         max skip conditionally.
13641         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13642         * config/ia64/ia64.c (ia64_option_override): Set default values
13643         for alignment options.
13644         * config/m68k/m68k.c: Handle new str_align_* options.
13645         * config/mips/mips.c (mips_set_compression_mode): Change
13646         type of constants.
13647         (mips_option_override): Set default values for options.
13648         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
13649         Likewise.
13650         * config/rs6000/rs6000.c (rs6000_option_override_internal):
13651         Likewise.
13652         * config/rx/rx.c (rx_option_override): Likewise.
13653         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
13654         (LABEL_ALIGN): Use align_labels_log.
13655         (LOOP_ALIGN): Use align_loops_align.
13656         * config/s390/s390.c (s390_asm_output_function_label): Use new
13657         macros.
13658         * config/sh/sh.c (sh_override_options_after_change):
13659         Change type of constants.
13660         * config/spu/spu.c (spu_sched_init): Likewise.
13661         * config/sparc/sparc.c (sparc_option_override): Set default
13662         values for options.
13663         * config/visium/visium.c (visium_option_override): Likewise.
13664         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
13665         emit p2align format with last argument if it's not needed.
13666         * doc/invoke.texi: Document extended format of -falign-*.
13667         * final.c: Use align_labels alignment.
13668         * flags.h (struct target_flag_state): Change type to use
13669         align_flags.
13670         (struct align_flags_tuple): New.
13671         (struct align_flags): Likewise.
13672         (align_loops_log): Redefine macro to use new types.
13673         (align_loops_max_skip): Redefine macro to use new types.
13674         (align_jumps_log): Redefine macro to use new types.
13675         (align_jumps_max_skip): Redefine macro to use new types.
13676         (align_labels_log): Redefine macro to use new types.
13677         (align_labels_max_skip): Redefine macro to use new types.
13678         (align_functions_log): Redefine macro to use new types.
13679         (align_loops): Redefine macro to use new types.
13680         (align_jumps): Redefine macro to use new types.
13681         (align_labels): Redefine macro to use new types.
13682         (align_functions): Redefine macro to use new types.
13683         (align_functions_max_skip): Redefine macro to use new types.
13684         (align_loops_value): New macro.
13685         (align_jumps_value): New macro.
13686         (align_labels_value): New macro.
13687         (align_functions_value): New macro.
13688         * function.c (invoke_set_current_function_hook): Propagate
13689         alignment values from flags to global variables default in
13690         topleev.h.
13691         * ipa-icf.c (sem_function::equals_wpa): Use
13692         cl_optimization_option_eq instead of memcmp.
13693         * lto-streamer.h (cl_optimization_stream_out): Support streaming
13694         of string types.
13695         (cl_optimization_stream_in): Likewise.
13696         * optc-save-gen.awk: Support strings in cl_optimization.
13697         * opth-gen.awk: Likewise.
13698         * opts.c (finish_options): Remove error checking of invalid
13699         value ranges.
13700         (MAX_CODE_ALIGN): Remove.
13701         (MAX_CODE_ALIGN_VALUE): Likewise.
13702         (parse_and_check_align_values): New function.
13703         (check_alignment_argument): Likewise.
13704         (common_handle_option): Use check_alignment_argument.
13705         * opts.h (parse_and_check_align_values): Declare.
13706         * toplev.c (init_alignments): Remove.
13707         (read_log_maxskip): New.
13708         (parse_N_M): Likewise.
13709         (parse_alignment_opts): Likewise.
13710         (backend_init_target): Remove usage of init_alignments.
13711         * toplev.h (parse_alignment_opts): Declare.
13712         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
13713         argument.
13714         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
13715         * tree.c (cl_option_hasher::equal): New.
13716         * varasm.c: Use new global macros.
13718 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13720         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
13721         Use a simpler align directive also if MAXSKIP = ALIGN-1.
13722         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13723         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13724         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13725         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13726         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
13727         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
13728         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
13729         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13730         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13731         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13732         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13734 2018-07-04  Martin Liska  <mliska@suse.cz>
13735             Jonathan Wakely  <jwakely@redhat.com>
13737         * coverage.c: Use correct type.
13738         * doc/invoke.texi: Language correction.
13740 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
13742         PR target/85620
13743         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
13744         ENDBRANCH for non-tail call which may return via indirect branch.
13745         * doc/extend.texi: Document indirect_return attribute.
13747 2018-07-03  Martin Sebor  <msebor@redhat.com>
13749         PR tree-optimization/86274
13750         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
13751         precondition.
13752         (format_floating): Correct handling of infinities and NaNs.
13754 2018-07-03  Martin Sebor  <msebor@redhat.com>
13756         * print-tree.c (print_real_cst): New function.
13757         (print_node_brief): Call it.
13758         (print_node): Ditto.
13760 2018-07-03  Jeff Law  <law@redhat.com>
13762         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
13763         into a single pattern.
13765         * config/h8300/h8300.md (ors code_iterator): New.
13766         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
13767         a single pattern and single splitter.
13768         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
13769         (iorqi3_1, xorqi3_1): Likewise.
13770         (iorqi3, xorqi3 expanders): Similarly.
13772         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
13773         (movmd_internal) into a single pattern using the P mode iterator.
13774         (movmd splitters): Similarly.
13775         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
13776         (movsd splitters): Similarly.
13778         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
13779         ADDB, ADDW and ADDL into a single ADD attribute which selects the
13780         right table based on the size of the operand.
13781         * config/h8300/h8300.md (length_table): Corresponding changes. All
13782         references to "addb", "addw" and "addl" changed to "add".
13783         (btst patterns): Merge two variants into a single pattern.
13784         (tstqi, tsthi): Likewise.
13785         (addhi3_incdec, addsi3_incdec): Likewise.
13786         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
13787         (mulhi3, mulsi3): Likewise.
13788         (udivhi3, udivsi3): Likewise.
13789         (divhi3, divsi3): Likewise.
13790         (andorqi3, andorhi3, andorsi3): Likewise.
13792 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
13794         PR target/85694
13795         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
13796         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
13798 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13800         PR tree-optimization/85694
13801         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
13802         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
13803         UNSPEC_URHADD.
13804         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
13805         (<u>avg<mode>3_ceil): New patterns.
13807 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
13809         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
13810         scan-tree-dump directive.
13811         * gcc.dg/vect/slp-perm-2.c: Likewise.
13812         * gcc.dg/vect/slp-perm-3.c: Likewise.
13813         * gcc.dg/vect/slp-perm-5.c: Likewise.
13814         * gcc.dg/vect/slp-perm-6.c: Likewise.
13815         * gcc.dg/vect/slp-perm-7.c: Likewise.
13816         * gcc.dg/vect/slp-perm-8.c: Likewise.
13818 2018-07-03  Marek Polacek  <polacek@redhat.com>
13820         PR middle-end/86202
13821         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
13823 2018-07-03  Richard Biener  <rguenther@suse.de>
13825         PR ipa/86389
13826         * tree-ssa-structalias.c (find_func_clobbers): Properly
13827         handle indirect calls.
13829 2018-07-03  Jeff Law  <law@redhat.com>
13831         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
13832         (shifts): New code iterator.
13833         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
13834         expander.  Fix HImode handling on H8/SX.
13835         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
13836         (subqi3, subhi3, subsi3 expanders): Likewise.
13837         (andqi3, andhi3, andsi3 expanders): Likewise.
13838         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
13839         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
13840         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
13841         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
13842         (zero_extendqihi2, zero_extendqisi2): Likewise.
13843         (extendqihi2, extendqisi2): Likewise.
13844         (rotlqi3, rotlhi3, rotlsi3): Likewise.
13845         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
13846         (rotlqi3_1, rotlhi3_1): Likewise.
13847         (logicalhi3_sn, logicalsi3_sn): Likewise.
13848         (logicalhi3, logicalsi3): Likewise.
13850 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13852         * tree-vect-patterns.c (vect_recog_rotate_pattern)
13853         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13854         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
13855         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
13856         type to append_pattern_def_seq instead of creating a stmt_vec_info
13857         directly.
13858         (build_mask_conversion): Likewise.  Remove vinfo argument.
13859         (vect_add_conversion_to_patterm): Likewise, renaming to...
13860         (vect_add_conversion_to_pattern): ...this.
13861         (vect_recog_mask_conversion_pattern): Update call to
13862         build_mask_conversion.  Pass the vector type to
13863         append_pattern_def_seq here too.
13864         (vect_recog_gather_scatter_pattern): Update call to
13865         vect_add_conversion_to_pattern.
13867 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13869         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
13870         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13871         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
13872         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
13873         STMT_VINFO_PATTERN_DEF_SEQ to null here.
13874         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
13875         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
13876         append_pattern_def_seq instead of new_pattern_def_seq.
13877         (vect_recog_divmod_pattern): Do both of the above.
13878         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
13879         is null.
13881 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13883         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
13884         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
13885         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
13886         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
13887         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
13888         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13889         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
13890         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13891         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13892         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
13893         parameter with a single stmt_vec_info.
13894         (vect_recog_func_ptr): Likewise.
13895         (vect_recog_gather_scatter_pattern): Likewise, folding in...
13896         (vect_try_gather_scatter_pattern): ...this.
13897         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
13898         the stmt_vec_info of the statement to be matched.  Don't clear
13899         STMT_VINFO_RELATED_STMT.
13900         (vect_pattern_recog): Update call accordingly.
13902 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13904         PR tree-optimization/85694
13905         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
13906         (uavgM3_ceil): Document new optabs.
13907         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
13908         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
13909         functions.
13910         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
13911         (savg_ceil_optab): New optabs.
13912         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
13913         (vect_vect_recog_func_ptrs): Add it.
13914         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
13915         constant directly from the associated lhs.
13917 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13919         * tree-vect-patterns.c (vect_split_statement): New function.
13920         (vect_convert_input): Use it to try to split an existing cast.
13922 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13924         * poly-int.h (print_hex): New function.
13925         * dumpfile.h (dump_dec, dump_hex): Declare.
13926         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
13927         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
13928         min_input_precision, operation_precision and operation_sign.
13929         * tree-vect-patterns.c (vect_get_range_info): New function.
13930         (vect_same_loop_or_bb_p, vect_single_imm_use)
13931         (vect_operation_fits_smaller_type): Delete.
13932         (vect_look_through_possible_promotion): Add an optional
13933         single_use_p parameter.
13934         (vect_recog_over_widening_pattern): Rewrite to use new
13935         stmt_vec_info infomration.  Handle one operation at a time.
13936         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
13937         (vect_truncatable_operation_p, vect_set_operation_type)
13938         (vect_set_min_input_precision): New functions.
13939         (vect_determine_min_output_precision_1): Likewise.
13940         (vect_determine_min_output_precision): Likewise.
13941         (vect_determine_precisions_from_range): Likewise.
13942         (vect_determine_precisions_from_users): Likewise.
13943         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
13944         (vect_vect_recog_func_ptrs): Put over_widening first.
13945         Add cast_forwprop.
13946         (vect_pattern_recog): Call vect_determine_precisions.
13948 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13950         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
13951         statements that have been replaced by further pattern statements.
13952         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
13954 2018-07-03  Richard Biener  <rguenther@suse.de>
13956         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
13957         always set *dt.  Dump vectype in vectype overload.
13958         * dumpfile.h (dump_gimple_expr): New function.
13959         (dump_gimple_expr_loc): Likewise.
13960         * dumpfile.c (dump_gimple_expr): New function.
13961         (dump_gimple_expr_loc): Likewise.
13963 2018-07-02  Jeff Law  <law@redhat.com>
13965         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
13966         the H8/300, H8/300H and H8/S variants into a single pattern.
13967         (movhi_h8300, movqi_h8300hs): Similarly.
13968         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
13969         (QHI mode iterator): New.
13971         * config/h8300/h8300.md: Remove trailing whitespace.
13973 2018-07-02  Jim Wilson  <jimw@sifive.com>
13975         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
13976         instead of emit_insn for interrupt returns.
13977         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
13978         (riscv_sret, riscv_uret): Likewise.
13980 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13982         * pretty-print.c (selftest::test_pp_format): Move save and restore
13983         of quotes to class auto_fix_quotes, and add an instance.
13984         * selftest.c: Include "intl.h".
13985         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
13986         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
13987         * selftest.h (selftest::auto_fix_quotes): New class.
13989 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
13991         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
13992         (aarch64_sve_prepare_conditional_op): Remove.
13993         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
13994         Allow aarch64_simd_reg_or_zero as select operand; remove
13995         the aarch64_sve_prepare_conditional_op call.
13996         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
13997         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13998         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
13999         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
14000         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
14001         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
14002         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
14003         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
14004         and a splitters to match all of the *_any patterns.
14005         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
14007         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
14008         (SVE_COND_FP_BINARY_REV): Remove.
14009         (sve_int_op_rev, sve_fp_op_rev): New.
14010         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
14011         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
14012         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
14013         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
14014         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
14015         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
14016         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
14017         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
14018         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
14020         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
14021         Remove match_dup 1 from the inner unspec.
14022         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
14024         * config/aarch64/aarch64.md (movprfx): New attr.
14025         (length): Default movprfx to 8.
14026         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
14027         (*madd<SVE_I>, *msub<SVE_I): Likewise.
14028         (*<su>mul<SVE_I>3_highpart): Likewise.
14029         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
14030         (*v<ASHIFT><SVE_I>3): Likewise.
14031         (*<su><MAXMIN><SVE_I>3): Likewise.
14032         (*<su><MAXMIN><SVE_F>3): Likewise.
14033         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
14034         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
14035         (*div<SVE_F>4): Likewise.
14037 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
14039         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
14040         in dump string.
14042 2018-07-02  Richard Biener  <rguenther@suse.de>
14044         PR tree-optimization/86363
14045         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
14046         memset argument refers to a non-variable address.
14048 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
14050         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
14051         {PLUS,MINUS}_EXPR code to...
14052         (adjust_symbolic_bound): ...here,
14053         (combine_bound): ...here,
14054         (set_value_range_with_overflow): ...and here.
14056 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
14058         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
14059         code...
14060         (extract_range_from_abs_expr): ...here.
14062 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
14064         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
14065         -fno-omit-frame-pointer when not optimizing.
14067 2018-07-02  Martin Liska  <mliska@suse.cz>
14069         PR ipa/86279
14070         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
14071         (propagate_nothrow): Likewise.
14073 2018-07-02  Martin Liska  <mliska@suse.cz>
14075         PR ipa/86323
14076         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
14078 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
14080         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
14081         function in r262149, changing "loc" param from source_location to
14082         const dump_location_t &.
14083         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
14084         declaration, as above.
14086 2018-07-01  Paul Koning  <ni1d@arrl.net>
14088         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14089         -munit-asm, -mgnu-asm, -mdec-asm.
14090         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
14091         (pdp11_output_labelref): New.
14092         (pdp11_output_def): New.
14093         (pdp11_output_addr_vec_elt): New.
14094         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
14095         %# and %@ format codes.
14096         (pdp11_option_override): New.
14097         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
14098         (pdp11_output_ident): New.
14099         (pdp11_asm_named_section): New.
14100         (pdp11_asm_init_sections): New.
14101         (pdp11_file_start): New.
14102         (pdp11_file_end): New.
14103         (output_ascii): Use .ascii/.asciz for -mdec-asm.
14104         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
14105         %o, like %c but octal.
14106         (pdp11_option_override): New.
14107         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
14108         -mdec-asm.
14109         (DATA_SECTION_ASM_OP): Ditto.
14110         (READONLY_DATA_SECTION_ASM_OP): New.
14111         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
14112         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
14113         (ASM_OUTPUT_LABELREF): Ditto.
14114         (ASM_OUTPUT_DEF): Ditto.
14115         (ASM_OUTPUT_EXTERNAL): New.
14116         (ASM_OUTPUT_SOURCE_FILENAME): New.
14117         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
14118         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
14119         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
14120         %# and %@ format codes.
14121         * config/pdp11/pdp11.opt (mgnu-asm): New.
14122         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
14123         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
14124         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
14126 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
14128         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
14129         dereferencing path[] beyond its length.
14130         (debug_path): New.
14131         (debug_all_paths): New.
14132         (rewire_first_differing_edge): New.
14133         (adjust_paths_after_duplication): New.
14134         (duplicate_thread_path): Call adjust_paths_after_duplication.
14135         Add new argument.
14136         (thread_through_all_blocks): Add new argument to
14137         duplicate_thread_path.
14139 2018-06-30  Jim Wilson  <jimw@sifive.com>
14141         * config/riscv/predicates.md (p2m1_shift_operand): New.
14142         (high_mask_shift_operand): New.
14143         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
14144         pattern using p2m1_shift_operand.
14145         (lshsi3_zero_extend_3+2): New combiner pattern using
14146         high_mask_shift_operand.
14148 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14150         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
14151         split out from...
14152         (vect_recog_rotate_pattern): ...here.
14153         (vect_convert_input): Try to insert casts of invariants in the
14154         preheader.
14155         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
14156         preheader to be empty.
14158 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14160         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
14161         vector type.  If given, install it in the new statement's
14162         STMT_VINFO_VECTYPE.
14163         (vect_element_precision): New function.
14164         (vect_unpromoted_value): New struct.
14165         (vect_unpromoted_value::vect_unpromoted_value): New function.
14166         (vect_unpromoted_value::set_op): Likewise.
14167         (vect_look_through_possible_promotion): Likewise.
14168         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
14169         (vect_widened_op_tree, vect_convert_input): Likewise.
14170         (vect_convert_inputs, vect_convert_output): Likewise.
14171         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
14172         to handle the optional cast of the multiplication result and
14173         vect_widened_op_tree to detect the widened multiplication itself.
14174         Do not require the input and output of promotion casts to have
14175         the same sign, but base the signedness of the operation on the
14176         input rather than the result.  If the pattern includes two
14177         promotions, check that those promotions have the same sign.
14178         Do not restrict the MULT_EXPR handling to a double-width result;
14179         handle quadruple-width results and wider.  Use vect_convert_inputs
14180         to convert the inputs to the common type.
14181         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
14182         to handle the optional cast of the ABS result.  Also allow a sign
14183         change or a sign extension between the ABS and MINUS.
14184         Use vect_widened_op_tree to detect the widened subtraction and use
14185         vect_convert_inputs to convert the inputs to the common type.
14186         (vect_handle_widen_op_by_const): Delete.
14187         (vect_recog_widen_op_pattern): New function.
14188         (vect_recog_widen_mult_pattern): Use it.
14189         (vect_recog_widen_shift_pattern): Likewise.
14190         (vect_recog_widen_sum_pattern): Use
14191         vect_look_through_possible_promotion to handle the promoted
14192         PLUS_EXPR operand.
14194 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14196         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
14197         the containing gimple_seq *.
14198         * gimple-iterator.h (gsi_for_stmt): Declare it.
14199         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
14200         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
14201         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
14202         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
14203         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
14204         checks.
14205         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
14206         split out from...
14207         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
14208         statement being replaced is part of an existing pattern
14209         definition sequence, inserting the new pattern statements before
14210         the original one.
14211         (vect_pattern_recog_1): Don't return a bool.  If the statement
14212         is already part of a pattern, instead apply pattern matching
14213         to the pattern definition statements.  Don't clear the
14214         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
14215         (vect_pattern_recog): Don't break after the first match;
14216         continue processing the pattern definition statements instead.
14217         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
14219 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14221         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14222         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14223         (vect_recog_widen_sum_pattern): Use it.
14225 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14227         * tree-vect-loop.c (vectorizable_reduction): Assert that the
14228         phi is not a pattern statement and has not been replaced by
14229         a pattern statement.
14230         * tree-vect-patterns.c (type_conversion_p): Don't check
14231         STMT_VINFO_IN_PATTERN_P.
14232         (vect_recog_vector_vector_shift_pattern): Likewise.
14233         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
14234         the pattern statement rather than the original statement; check
14235         directly for a WIDEN_MULT_EXPR here.
14236         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
14237         vect_is_simple_use to return the pattern statement rather
14238         than the original statement; use is_pattern_stmt_p to check
14239         for such a pattern statement.
14240         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
14241         to return the pattern statement rather than the original statement;
14242         don't do the same transformation here.
14243         (vect_is_simple_use): If the defining statement has been replaced
14244         by a pattern statement, return the pattern statement instead.
14245         Remove the corresponding (local) transformation from the vectype
14246         overload.
14248 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14250         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
14251         end and default to null.
14252         * tree-vect-loop.c (vect_create_epilog_for_reduction)
14253         (vectorizable_reduction): Update calls accordingly, dropping the
14254         gimple ** argument if the passed-back statement isn't needed.
14255         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
14256         (vect_recog_rotate_pattern): Likewise.
14257         (vect_recog_mask_conversion_pattern): Likewise.
14258         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
14259         (vect_mask_constant_operand_p): Likewise.
14260         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
14261         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
14262         (get_group_load_store_type, get_load_store_type): Likewise.
14263         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
14264         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
14265         (vectorizable_conversion, vectorizable_assignment): Likewise.
14266         (vectorizable_shift, vectorizable_operation): Likewise.
14267         (vectorizable_store, vect_is_simple_cond): Likewise.
14268         (vectorizable_condition, vectorizable_comparison): Likewise.
14269         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
14270         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
14271         and move it to the end.  Cope with null def_stmt_outs.
14273 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14275         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
14277 2018-06-29  Jeff Law  <law@redhat.com>
14279         * config/v850/v850.c (v850_legitimate_address_p): Handle large
14280         displacements for TARGET_V850E2V3 and newer.
14281         (TARGET_LRA_P): Remove.  Defaults to LRA now.
14282         * config/v850/v850.md (sign23byte_load): Remove.
14283         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
14284         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
14286 2018-06-29  Martin Liska  <mliska@suse.cz>
14288         PR lto/85759
14289         * coverage.c (coverage_init): Mangle full path name.
14290         * doc/invoke.texi: Document the change.
14291         * gcov-io.c (mangle_path): New.
14292         * gcov-io.h (mangle_path): Likewise.
14293         * gcov.c (mangle_name): Use mangle_path for path mangling.
14295 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14297         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
14298         if starting source register is not even.
14300 2018-06-29  Martin Liska  <mliska@suse.cz>
14302         PR tree-optimization/86263
14303         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
14304         Make edge redirection.
14306 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
14308         * dumpfile.c (dump_loc): Add indentation based on scope depth.
14309         (dump_scope_depth): New variable.
14310         (get_dump_scope_depth): New function.
14311         (dump_begin_scope): New function.
14312         (dump_end_scope): New function.
14313         * dumpfile.h (get_dump_scope_depth): New declaration.
14314         (dump_begin_scope): New declaration.
14315         (dump_end_scope): New declaration.
14316         (class auto_dump_scope): New class.
14317         (AUTO_DUMP_SCOPE): New macro.
14318         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
14319         AUTO_DUMP_SCOPE.
14321 2018-06-29  Richard Biener  <rguenther@suse.de>
14323         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
14324         compute_all_dependences succeeds.
14325         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
14326         exceed --param loop-max-datarefs-for-datadeps.
14328 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
14330         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
14332 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
14334         PR target/86348
14335         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
14336         alternative 0 in preferred_for_speed attribute.
14338 2018-06-28  Paul Koning  <ni1d@arrl.net>
14340         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
14341         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
14342         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
14343         * config/pdp11/pdp11.md: Correct "length" attribute calculation
14344         for shift insn patterns.
14346 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
14348         * cgraph.c (cgraph_node::get_body): Replace assignments to
14349         "dump_file" with calls to set_dump_file.
14350         * dumpfile.c (alt_dump_file): Make static, and group with...
14351         (alt_flags): ...this definition.
14352         (dumps_are_enabled): New variable.
14353         (refresh_dumps_are_enabled): New function.
14354         (set_dump_file): New function.
14355         (set_alt_dump_file): New function.
14356         (gcc::dump_manager::dump_start): Replace assignments to
14357         "dump_file" and "alt_dump_file" with calls to set_dump_file and
14358         set_alt_dump_file.
14359         (gcc::dump_manager::dump_finish): Likewise.
14360         * dumpfile.h (alt_dump_file): Delete decl.
14361         (dumps_are_enabled): New variable decl.
14362         (set_dump_file): New function decl.
14363         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
14364         global.
14365         * tree-nested.c (lower_nested_functions): Replace assignments to
14366         "dump_file" with calls to set_dump_file.
14368 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14370         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
14371         goto_locus of each outgoing edge of each basic block.
14373 2018-06-28  Richard Biener  <rguenther@suse.de>
14375         * dwarf2out.c (decl_scope_table): Remove.
14376         (push_decl_scope): Likewise.
14377         (pop_decl_scope): Likewise.
14378         (gen_type_die_for_member): Do not call push/pop_decl_scope.
14379         (gen_struct_or_union_type_die): Likewise.
14380         (gen_tagged_type_die): Likewise.
14381         (dwarf2out_init): Do not initialize decl_scope_table.
14382         (dwarf2out_c_finalize): Do not free it.
14384 2018-06-28  Richard Biener  <rguenther@suse.de>
14386         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
14387         deciding whether to not re-use a DIE.
14389 2018-06-28  Richard Biener  <rguenther@suse.de>
14391         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
14392         DW_AT_abstract_origin attribute.
14394 2018-06-28  Martin Liska  <mliska@suse.cz>
14396         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
14397         Use newly introduced constants.
14398         * tree-switch-conversion.h (struct jump_table_cluster):
14399         Define max_ratio_for_size and max_ratio_for_speed.
14401 2018-06-28  Martin Liska  <mliska@suse.cz>
14403         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14404         Add new checking assert to catch invalid state.
14405         (jump_table_cluster::can_be_handled): Handle single case
14406         clusters.
14407         (jump_table_cluster::is_beneficial): Bail out for such case.
14408         (bit_test_cluster::find_bit_tests):
14409         Add new checking assert to catch invalid state.
14410         (bit_test_cluster::can_be_handled): Handle single case
14411         clusters.
14412         (bit_test_cluster::is_beneficial): Bail out for such case.
14413         (switch_decision_tree::analyze_switch_statement):
14414         Fix comment.
14416 2018-06-28  Martin Liska  <mliska@suse.cz>
14418         * common.opt: Introduce -completion option.
14419         * gcc.c (driver_handle_option): Handle it.
14420         (driver::main): Print completions if completion
14421         is set.
14422         * opt-suggestions.c (option_proposer::get_completions):
14423         New function.
14424         (option_proposer::suggest_completion): Likewise.
14425         (option_proposer::find_param_completions): Likewise.
14426         (verify_autocompletions): Likewise.
14427         (test_completion_valid_options): Likewise.
14428         (test_completion_valid_params): Likewise.
14429         (in_completion_p): Likewise.
14430         (empty_completion_p): Likewise.
14431         (test_completion_partial_match): Likewise.
14432         (test_completion_garbage): Likewise.
14433         (opt_proposer_c_tests): Likewise.
14434         * opt-suggestions.h: Declare new functions.
14435         * opts.c (common_handle_option): Handle OPT__completion_.
14436         * selftest-run-tests.c (selftest::run_tests): Add
14437         opt_proposer_c_tests.
14438         * selftest.c (assert_str_startswith): New.
14439         * selftest.h (assert_str_startswith): Likewise.
14440         (opt_proposer_c_tests): New.
14441         (ASSERT_STR_STARTSWITH): Likewise.
14443 2018-06-28  Martin Liska  <mliska@suse.cz>
14445         * Makefile.in: Add opt-suggestions.o.
14446         * gcc-main.c: Include opt-suggestions.h.
14447         * gcc.c (driver::driver): Likewise.
14448         (driver::~driver): Remove m_option_suggestions.
14449         (driver::build_option_suggestions): Moved to option_proposer.
14450         (driver::suggest_option): Likewise.
14451         (driver::handle_unrecognized_options): Use option_proposer.
14452         * gcc.h (class driver): Add new memver m_option_proposer.
14453         * opt-suggestions.c: New file.
14454         * opt-suggestions.h: New file.
14456 2018-06-28  Martin Liska  <mliska@suse.cz>
14458         * vec.h (class auto_string_vec): New (moved from auto_argvec).
14459         (auto_string_vec::~auto_string_vec): Likewise.
14461 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14463         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
14464         prevent_decl_creation_for_types fields up and add reset_location field.
14465         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
14466         statement if id->reset_location is true.
14467         (copy_edges_for_bb): Do not set goto_locus on the new edges if
14468         id->reset_location is true.
14469         (copy_phis_for_bb): Force input_location on the arguments if
14470         id->reset_location is true.
14471         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
14472         is set on the function to be inlined.
14474 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
14476         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
14478 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
14480         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
14481         registers for Pmode.
14482         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
14483         hard registers for the clobbered pseudo.
14485 2018-06-27  Paul Koning  <ni1d@arrl.net>
14487         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14488         mutually exclusive options.
14489         * config/pdp11/constraints.md (h): New constraint.
14490         (O): Update definition to match shift code generation.
14491         (D): New constraint.
14492         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
14493         (CCFP): Remove.
14494         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
14495         function.
14496         (output_jump): Change arguments.
14497         (pdp11_fixed_cc_regs): New function.
14498         (pdp11_cc_mode): Ditto.
14499         (pdp11_expand_shift): Ditto.
14500         (pdp11_assemble_shift): Ditto.
14501         (pdp11_small_shift): Ditto.
14502         (pdp11_branch_cost): Remove.
14503         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
14504         from output.
14505         (pdp11_register_move_cost): Update for CC registers.
14506         (pdp11_rtx_costs): Add case for LSHIFTRT.
14507         (pdp11_output_jump): Add CCNZ mode conditional branches.
14508         (notice_update_cc_on_set): Remove.
14509         (pdp11_cc_mode): New function.
14510         (simple_memory_operand): Correct pre/post decrement case.
14511         (no_side_effect_operand): New function.
14512         (pdp11_regno_reg_class): Add CC_REGS class.
14513         (pdp11_fixed_cc_regs): New function.
14514         (pdp11_small_shift): New function.
14515         (pdp11_expand_shift): New function to expand shift insns.
14516         (pdp11_assemble_shift): New function to output shifts.
14517         (pdp11_branch_cost): Remove.
14518         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
14519         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
14520         (WCHAR_TYPE): Ditto.
14521         (PTRDIFF_TYPE): Ditto.
14522         (ADJUST_INSN_LENGTH): New macro.
14523         (FIXED_REGISTERS): Add CC registers.
14524         (CALL_USED_REGISTERS): Ditto.
14525         (reg_class): Ditto.
14526         (REG_CLASS_NAMES): Ditto.
14527         (REG_CLASS_CONTENTS): Ditto.
14528         (SELECT_CC_MODE): Use new function.
14529         (TARGET_FLAGS_REGNUM): New macro.
14530         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
14531         (cc0_reg_rtx): Remove.
14532         (CC_STATUS_MDEP): Remove.
14533         (CC_STATUS_MDEFP_INIT): Remove.
14534         (CC_IN_FPU): Remove.
14535         (NOTICE_UPDATE_CC): Remove.
14536         (REGISTER_NAMES): Add CC registers.
14537         (BRANCH_COST): Change to constant 1.
14538         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
14539         handling.
14540         * config/pdp11/pdp11.opt (mbcopy): Remove.
14541         (mbcopy-builtin): Remove.
14542         (mbranch-cheap): Remove.
14543         (mbranch-expensive): Remove.
14544         * config/pdp11/predicates.md (expand_shift_operand): Update to
14545         match shift code generation.
14546         (ccnz_operator): New predicate.
14547         * doc/invoke.texi (PDP-11 Options): Remove deleted options
14548         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
14549         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
14550         exclusive options.
14551         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
14552         description of O constraint.
14554 2018-06-27  Jeff Law  <law@redhat.com>
14555             Austin Law  <austinklaw@gmail.com>
14557         * config/v850/v850.md (addsi3_set_flags): New pattern.
14558         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
14559         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
14560         (zero_extendhisi2_v850_set_flags): Likewise.
14561         (zero_extendqisi2_v850_set_flags): Likewise.
14562         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
14563         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
14564         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
14566         * config/v850/v850-protos.h (notice_update_cc): Remove.
14567         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
14568         (v850_print_operand): Handle 'D' and "d".
14569         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
14570         Add handling of arithmetic/logical operations compared against zero.
14571         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
14572         Do not look at v850_compare_op, instead get mode from last argument.
14573         (v850_gen_compare): Remove
14574         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
14575         after reload for prologue insns.
14576         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
14577         patterns.
14578         (construct_save_jarl): Likewise.
14579         (TARGET_FLAGS_REGNUM): Define.
14580         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
14581         (NOTICE_UPDATE_CC): Remove.
14582         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
14583         than cc0.  Conditionalize on reload_completed.
14584         (cmpsi_insn, setfcc_insn): Likewise.
14585         (tst1 splitter): Turn into define_and_split which sets the flags
14586         after reload.
14587         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
14588         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
14589         (cstoresf4, cstoredf4): Clobber the flags.
14590         (cmpsi, cmpsf, cmpdf): Remove expanders.
14591         (setf_insn): Remove pattern.
14592         (addsi3): Turn into define_and_split which clobbers the flags after
14593         reload and a suitable pattern (addsi3_clobber_flags) for use after
14594         reload.
14595         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
14596         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
14597         (ashrsi3, ashrsi3_v850e2): Likewise.
14598         (bins): Clobber the flags.
14599         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
14600         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
14601         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
14602         (call_value_internal_short, call_value_internal_long): Likewise.
14603         (callt_save_interrupt, callt_return_interrupt): Likewise.
14604         (save_interrupt, return_interrupt): Likewise.
14605         (callt_save_all_interrupt, save_all_interrupt): Likewise.
14606         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
14607         (restore_all_interrupt, _restore_all_interrupt): Likewise.
14608         (All FP comparisons): Only allow after reload has completed.
14609         (trfsr): Likewise.
14610         (divh, divhu): Tweak output template.
14611         (branch_z_normal, branch_z_invert): Remove
14612         (branch_nz_normal, branch_nz_invert): Likewise.
14613         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
14615         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
14616         * config/v850/v850.c (notice_update_cc): Remove.
14617         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
14618         (CC_NO_CARRY): Likewise.
14619         (NOTICE_UPDATE_CC): Define to nothing.
14620         * config/v850/v850.md: Remove block comment on cc0 handling
14621         Remove "cc" attribute from all patterns.  Remove cc_status handling
14622         from all patterns.  Minor formatting fixes.
14624 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14626         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
14627         (cortex-a76.cortex-a55): Likewise.
14628         * config/aarch64/aarch64-tune.md: Regenerate.
14629         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
14630         cortex-a76.cortex-a55.
14632 2018-06-27  Jeff Law  <law@redhat.com>
14634         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
14635         (MULTILIB_DIRNAMES): Similarly.
14637 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
14639         * gimple.h (gimple_return_retbnd): Delete.
14640         (gimple_return_set_retbnd): Likewise.
14641         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
14642         gimple_return_set_retbnd.
14643         * gimple-pretty-print.c (dump_gimple_return): Remove call to
14644         gimple_return_retbnd and adjust.
14645         * tree-inline.h (struct copy_body_data): Remove retbnd field.
14646         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
14647         Explicitly return NULL in a couple more cases.  Move assertion
14648         on debug statements and remove unreachable code.
14649         (reset_debug_binding): Do not test id->retbnd.
14650         (expand_call_inline): Do not set it.
14652 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
14654         * configure.ac: Add --disable-gcov option.
14655         * configure: Regenerate.
14656         * Makefile.in: Honour @enable_gcov@.
14657         * doc/install.texi: Document --disable-gcov.
14659 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14661         * config/arm/arm-cpus.in (cortex-a76): New entry.
14662         (cortex-a76.cortex-a55): Likewise.
14663         * config/arm/arm-tables.opt: Regenerate.
14664         * config/arm/arm-tune.md: Likewise.
14665         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
14666         * doc/invoke.texi (ARM Options): Document cortex-a76 and
14667         cortex-a76.cortex-a55.
14669 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
14671         PR target/85769
14672         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
14674 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
14676         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
14677         comment.
14678         (EPILOGUE_USES): Likewise.
14680 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
14682         * tree-inline.c (remap_location): New function extracted from...
14683         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
14684         (copy_phis_for_bb): ...here.  Call remap_location.
14685         (copy_cfg_body): Adjust call to copy_edges_for_bb.
14687 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
14689         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
14690         unaligned vsx for 16B memset.
14692 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
14694         PR target/86285
14695         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
14696         ieee128_float_type_node to long_double_type_node unless
14697         TARGET_LONG_DOUBLE_128 is set.
14699 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
14701         * cfgloop.c (get_loop_location): Convert return type from
14702         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
14703         by implicit construction from rtx_insn *, and using
14704         dump_user_location_t::from_function_decl for the fallback case.
14705         * cfgloop.h (get_loop_location): Convert return type from
14706         location_t to dump_user_location_t.
14707         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
14708         dump_printf_loc to pass in a dump_location_t rather than a
14709         location_t, via the gimple stmt.
14710         * coverage.c (get_coverage_counts): Update calls to
14711         dump_printf_loc to pass in dump_location_t rather than a
14712         location_t.
14713         * doc/optinfo.texi (Dump types): Convert example of
14714         dump_printf_loc from taking "locus" to taking "insn".  Update
14715         description of the "_loc" calls to cover dump_location_t.
14716         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
14717         "selftest.h".
14718         (dump_user_location_t::dump_user_location_t): New constructors,
14719         from gimple *stmt and rtx_insn *.
14720         (dump_user_location_t::from_function_decl): New function.
14721         (dump_loc): Make static.
14722         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
14723         const dump_location_t &.
14724         (dump_generic_expr_loc): Delete.
14725         (dump_printf_loc): Convert param "loc" from location_t to
14726         const dump_location_t &.
14727         (selftest::test_impl_location): New function.
14728         (selftest::dumpfile_c_tests): New function.
14729         * dumpfile.h: Include "profile-count.h".
14730         (class dump_user_location_t): New class.
14731         (struct dump_impl_location_t): New struct.
14732         (class dump_location_t): New class.
14733         (dump_printf_loc): Convert 2nd param from source_location to
14734         const dump_location_t &.
14735         (dump_generic_expr_loc): Delete.
14736         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
14737         const dump_location_t &.
14738         * gimple-fold.c (fold_gimple_assign): Update call to
14739         dump_printf_loc to pass in a dump_location_t rather than a
14740         location_t, via the gimple stmt.
14741         (gimple_fold_call): Likewise.
14742         * gimple-loop-interchange.cc
14743         (loop_cand::analyze_iloop_reduction_var): Update for change to
14744         check_reduction_path.
14745         (tree_loop_interchange::interchange): Update for change to
14746         find_loop_location.
14747         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
14748         change in return-type of find_loop_location.
14749         (graphite_regenerate_ast_isl): Likewise.
14750         * graphite-optimize-isl.c (optimize_isl): Likewise.
14751         * graphite.c (graphite_transform_loops): Likewise.
14752         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
14753         pass in a dump_location_t rather than a location_t, via the
14754         gimple stmt.
14755         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14756         * ipa.c (walk_polymorphic_call_targets): Likewise.
14757         * loop-unroll.c (report_unroll): Convert "locus" param from
14758         location_t to dump_location_t.
14759         (decide_unrolling): Update for change to get_loop_location's
14760         return type.
14761         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
14762         location_t to dump_user_location_t.
14763         (grid_find_single_omp_among_assignments_1): Updates calls to
14764         dump_printf_loc to pass in a dump_location_t rather than a
14765         location_t, via the gimple stmt.
14766         (grid_parallel_clauses_gridifiable): Convert "tloc" from
14767         location_t to dump_location_t.  Updates calls to dump_printf_loc
14768         to pass in a dump_location_t rather than a location_t, via the
14769         gimple stmt.
14770         (grid_inner_loop_gridifiable_p): Likewise.
14771         (grid_dist_follows_simple_pattern): Likewise.
14772         (grid_gfor_follows_tiling_pattern): Likewise.
14773         (grid_target_follows_gridifiable_pattern): Likewise.
14774         (grid_attempt_target_gridification): Convert initialization
14775         of local "grid" from memset to zero-initialization; FIXME: does
14776         this require C++11?  Update call to dump_printf_loc to pass in a
14777         optinfo_location rather than a location_t, via the gimple stmt.
14778         * profile.c (read_profile_edge_counts): Updates call to
14779         dump_printf_loc to pass in a dump_location_t rather than a
14780         location_t
14781         (compute_branch_probabilities): Likewise.
14782         * selftest-run-tests.c (selftest::run_tests): Call
14783         dumpfile_c_tests.
14784         * selftest.h (dumpfile_c_tests): New decl.
14785         * tree-loop-distribution.c (pass_loop_distribution::execute):
14786         Update for change in return type of find_loop_location.
14787         * tree-parloops.c (parallelize_loops): Likewise.
14788         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
14789         "locus" from location_t to dump_user_location_t.
14790         (canonicalize_loop_induction_variables): Likewise.
14791         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
14792         for change in return type of find_loop_location.
14793         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
14794         to dump_printf_loc to pass in a dump_location_t rather than a
14795         location_t, via the stmt.
14796         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
14797         Likewise.
14798         * tree-vect-loop-manip.c (find_loop_location): Convert return
14799         type from source_location to dump_user_location_t.
14800         (vect_do_peeling): Update for above change.
14801         (vect_loop_versioning): Update for change in type of
14802         vect_location.
14803         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
14804         from location_t to dump_user_location_t.
14805         (vect_estimate_min_profitable_iters): Update for change in type
14806         of vect_location.
14807         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
14808         location_t to dump_location_t.
14809         (vect_slp_bb): Update for change in type of vect_location.
14810         * tree-vectorizer.c (vect_location): Convert from source_location
14811         to dump_user_location_t.
14812         (try_vectorize_loop_1): Update for change in vect_location's type.
14813         (vectorize_loops): Likewise.
14814         (increase_alignment): Likewise.
14815         * tree-vectorizer.h (vect_location): Convert from source_location
14816         to dump_user_location_t.
14817         (find_loop_location): Convert return type from source_location to
14818         dump_user_location_t.
14819         (check_reduction_path): Convert 1st param from location_t to
14820         dump_user_location_t.
14821         * value-prof.c (check_counter): Update call to dump_printf_loc to
14822         pass in a dump_user_location_t rather than a location_t; update
14823         call to error_at for change in type of "locus".
14824         (check_ic_target): Update call to dump_printf_loc to
14825         pass in a dump_user_location_t rather than a location_t, via the
14826         call_stmt.
14828 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
14830         * config/s390/s390.h (enum processor_flags): Do not use
14831         default tune parameter when -march was specified.
14833 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
14835         PR target/86314
14836         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
14837         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
14838         operands.
14840 2018-06-26  Richard Biener  <rguenther@suse.de>
14842         PR tree-optimization/86287
14843         PR bootstrap/86316
14844         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
14845         (vect_analyze_loop): Initialize n_stmts.
14847 2018-06-26  Richard Biener  <rguenther@suse.de>
14849         PR middle-end/86271
14850         * fold-const.c (fold_convertible_p): Pointer extension
14851         isn't valid.
14853 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
14855         PR debug/86064
14856         * dwarf2out.c (loc_list_has_views): Adjust comments.
14857         (dw_loc_list): Split single cross-partition range with
14858         nonzero locview.
14860 2018-06-25  Jeff Law  <law@redhat.com>
14862         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
14863         on -mbig-switch by default.
14865         * config/v850/predicates.md (const_float_1_operand): Fix match_code
14866         test.
14867         (const_float_0_operand): Remove unused predicate.
14868         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
14869         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
14870         (recipsf2): New expander.  Original pattern now called
14871         (recipsf2_insn).
14872         (recipdf2, recipdf2_insn): Similarly.
14873         (rsqrtsf2, rsqrtsf2_insn): Similarly
14874         (rsqrtdf2, rsqrtdf2_insn): Similarly
14876 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
14878         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
14879         Simplify logic for FreeBSD (twice).
14881 2018-06-25  Martin Sebor  <msebor@redhat.com>
14883         PR tree-optimization/86204
14884         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
14885         a strnlen result if it's less than the length of the string.
14887 2018-06-25  Martin Sebor  <msebor@redhat.com>
14889         PR tree-optimization/85700
14890         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
14891         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
14892         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
14894 2018-06-25  Martin Sebor  <msebor@redhat.com>
14896         * doc/extend.texi (Zero-length arrays): Update and clarify.
14898 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
14900         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
14901         added IEEE/IBM long double multilib support on PowerPC little
14902         endian Linux systems.
14903         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
14904         (MULTILIB_DEFAULTS): Likewise.
14905         * config/rs6000/rs6000.c (rs6000_option_override_internal):
14906         Likewise.
14907         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
14908         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
14909         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
14911 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
14913         PR middle-end/86311
14914         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
14915         (REORDER_45): Likewise.
14917 2018-06-25  Jeff Law  <law@redhat.com>
14919         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
14920         dividend to 32 bits.  Adjust length.
14921         (udivmodhi4): Cleanup output template.  Fix length.
14923 2018-06-25  Carl Love  <cel@us.ibm.com>
14925         * config/rs6000/vsx.md: Change word selector to prefered location.
14927 2018-06-25  Richard Biener  <rguenther@suse.de>
14929         PR tree-optimization/86304
14930         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
14931         epilogue-if-converted loops as well.
14933 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
14935         * lto-section-out.c (lto_begin_section): Do not print section
14936         name for noaddr and unnumbered dumps.
14938 2018-06-25  Richard Biener  <rguenther@suse.de>
14940         * tree-vectorizer.h (struct vec_info_shared): New structure
14941         with parts split out from struct vec_info and loop_nest from
14942         struct _loop_vec_info.
14943         (struct vec_info): Adjust accordingly.
14944         (struct _loop_vec_info): Likewise.
14945         (LOOP_VINFO_LOOP_NEST): Adjust.
14946         (LOOP_VINFO_DATAREFS): Likewise.
14947         (LOOP_VINFO_DDRS): Likewise.
14948         (struct _bb_vec_info): Likewise.
14949         (BB_VINFO_DATAREFS): Likewise.
14950         (BB_VINFO_DDRS): Likewise.
14951         (struct _stmt_vec_info): Add dr_aux member.
14952         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
14953         (DR_MISALIGNMENT_UNINITIALIZED): New.
14954         (set_dr_misalignment): Adjust.
14955         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
14956         (vect_analyze_loop): Adjust prototype.
14957         (vect_analyze_loop_form): Likewise.
14958         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14959         Compute dependences lazily.
14960         (vect_record_base_alignments): Use shared datarefs/ddrs.
14961         (vect_verify_datarefs_alignment): Likewise.
14962         (vect_analyze_data_refs_alignment): Likewise.
14963         (vect_analyze_data_ref_accesses): Likewise.
14964         (vect_analyze_data_refs): Likewise.
14965         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
14966         constructor parameter for shared part.
14967         (vect_analyze_loop_form): Pass in shared part and adjust.
14968         (vect_analyze_loop_2): Pass in storage for the number of
14969         stmts.  Move loop nest finding to the caller.  Compute
14970         datarefs lazily.
14971         (vect_analyze_loop): Pass in shared part.
14972         (vect_transform_loop): Verify shared datarefs are unchanged.
14973         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
14974         constructor parameter for shared part.
14975         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
14976         (vect_slp_bb): Verify shared datarefs are unchanged before
14977         transform.
14978         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
14979         change.
14980         (new_stmt_vec_info): Initialize DR_AUX misalignment to
14981         DR_MISALIGNMENT_UNINITIALIZED.
14982         * tree-vectorizer.c (vec_info::vec_info): Add constructor
14983         parameter for shared part.
14984         (vec_info::~vec_info): Adjust.
14985         (vec_info_shared::vec_info_shared): New.
14986         (vec_info_shared::~vec_info_shared): Likewise.
14987         (vec_info_shared::save_datarefs): Likewise.
14988         (vec_info_shared::check_datarefs): Likewise.
14989         (try_vectorize_loop_1): Construct shared part live for analyses
14990         of a single loop for multiple vector sizes.
14991         * tree-parloops.c (gather_scalar_reductions): Adjust.
14993 2018-06-25  Richard Biener  <rguenther@suse.de>
14995         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
14996         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
14997         (vect_analyze_data_refs): Remove similar code from here and
14998         simplify accordingly.
15000 2018-06-25  Richard Biener  <rguenther@suse.de>
15002         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
15003         for reverse storage order accesses rather than asserting
15004         they cannot happen here.
15006 2018-06-25  Tom de Vries  <tdevries@suse.de>
15008         PR debug/86257
15009         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
15010         Use data16 instead of .byte for insn prefix.
15012 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
15014         PR C++/86082
15015         * parser.c (make_char_string_pack): Pass this literal chars
15016         through cpp_interpret_string.
15017         (cp_parser_userdef_numeric_literal): Check the result of
15018         make_char_string_pack.
15020 2018-06-24  Maya Rashish  <coypu@sdf.org>
15022         * ginclude/stddef.h: Simplify conditions around avoiding
15023         re-definition of __size_t.
15025 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
15027         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
15028         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
15030 2018-06-22  Maya Rashish  <coypu@sdf.org>
15032         * doc/invoke.texi (mno-fancy-math-387): Update for changes
15033         made to OpenBSD and NetBSD through the years.
15035 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15037         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
15038         behavior of vec_pack (vector double, vector double) to match
15039         behavior of vec_float2 (vector double, vector double).
15041 2018-06-22  Olivier Hainque  <hainque@adacore.com>
15043         * gimplify.c (gimplify_function_tree): Prevent creation
15044         of a trampoline for the address of the current function
15045         passed to entry/exit instrumentation hooks.
15047 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
15049         PR target/86222
15050         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
15051         correctly.
15053 2018-06-22  Martin Liska  <mliska@suse.cz>
15055         PR tree-optimization/86263
15056         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15057         Bail out if is_enabled is false.
15058         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
15059         New declaration.
15060         (jump_table_cluster::is_enabled): New function.
15062 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
15064         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
15065         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
15066         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
15067         (lto_input_ts_binfo_tree_pointers): Likewise.
15068         * tree-streamer-out.c (streamer_write_tree_bitfields,
15069         write_ts_binfo_tree_pointers): Likewise.
15070         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
15072 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
15074         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
15076 2018-06-22  Martin Liska  <mliska@suse.cz>
15078         * symbol-summary.h (get): Make it pure and inline move
15079         functionality from ::get function.
15080         (get): Remove and inline into ::get and ::get_create.
15081         (get_create): Move code from ::get function.
15083 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15085         PR target/85994
15086         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
15087         -x assembler-with-cpp.
15089 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15091         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
15092         _FILE_OFFSET_BITS=64 for C++.
15094 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
15096         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
15097         conversion insn that shows up when pr85657-3.c is compiled using
15098         IEEE 128-bit long double.
15099         (neg<mode>2_internal): Use the correct mode to check whether the
15100         mode is IBM extended.
15101         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
15102         multiply and divide external functions from being created more
15103         than once.
15105 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15107         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
15108         functions.
15109         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
15110         the edge can be forwarded.
15111         (cfg_layout_merge_blocks): Likewise.
15113 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15115         * except.c (finish_eh_generation): Commit edge insertions only after
15116         the EH edges have been redirected from post-landing to landing pads.
15118 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15120         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
15121         create_tmp_var_for to create the FRAME decl.
15122         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
15124 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15126         * tree-inline.c (copy_edges_for_bb): Minor tweak.
15127         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
15128         debug statement when resetting its value.
15129         (expand_call_inline): Copy the locus of the call onto the assignment
15130         of the return value, if any.  Use local variable in more cases.
15132 2018-06-21  Martin Liska  <mliska@suse.cz>
15134         * ipa-pure-const.c (propagate_nothrow): Use
15135         funct_state_summaries->get.
15136         (dump_malloc_lattice): Likewise.
15137         (propagate_malloc): Likewise.
15139 2018-06-21  Richard Biener  <rguenther@suse.de>
15141         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
15142         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
15143         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
15144         comment.
15145         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
15146         BLOCK_ABSTRACT_ORIGIN unconditionally.
15148 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
15150         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
15151         deleting it.
15152         * ipa-reference.c (ipa_reference_c_finalize): Delete
15153         ipa_ref_opt_sum_summaries and set it to NULL.
15155 2018-06-21  Tom de Vries  <tdevries@suse.de>
15157         PR tree-optimization/85859
15158         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
15159         test with comment from bb_no_side_effects_p.
15161 2018-06-21  Richard Biener  <rguenther@suse.de>
15163         PR tree-optimization/86232
15164         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
15165         max for constant niter.
15167 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15169         * config/aarch64/aarch64-simd.md
15170         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
15172 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15174         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
15175         Make opernads of the unspec commutative.
15177 2018-06-21  Richard Biener  <rguenther@suse.de>
15179         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
15180         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
15181         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
15182         (vect_analyze_data_ref_dependence): Re-order checks to deal with
15183         NULL DR_STEP.
15184         (vect_record_base_alignments): Do not record base alignment
15185         for gathers or scatters.
15186         (vect_compute_data_ref_alignment): Drop return value that is always
15187         true.  Bail out early for gathers or scatters.
15188         (vect_enhance_data_refs_alignment): Bail out early for gathers
15189         or scatters.
15190         (vect_find_same_alignment_drs): Likewise.
15191         (vect_analyze_data_refs_alignment): Remove dead code.
15192         (vect_slp_analyze_and_verify_node_alignment): Likewise.
15193         (vect_analyze_data_refs): For possible gathers or scatters do
15194         not create an alternate DR, just check their possible validity
15195         and mark them.  Adjust DECL_NONALIASED handling to not rely
15196         on DR_BASE_ADDRESS.
15197         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
15198         update inits of gathers or scatters.
15199         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
15200         Also copy gather/scatter flag to pattern vinfo.
15202 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15204         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
15205         behavior of vec_packsu (vector unsigned long long, vector unsigned
15206         long long) to match behavior of vec_packs with same signature.
15208 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
15209             Thomas Schwinge <thomas@codesourcery.com>
15210             Cesar Philippidis  <cesar@codesourcery.com>
15212         * gimplify.c (gimplify_scan_omp_clauses): Add support for
15213         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15214         (gimplify_adjust_omp_clauses): Likewise.
15215         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
15216         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
15217         (gimplify_omp_target_update): Update handling of acc update and
15218         enter/exit data.
15219         * omp-low.c (install_var_field): Remove unused parameter
15220         base_pointers_restrict.
15221         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
15222         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
15223         FINALIZE}
15224         (omp_target_base_pointers_restrict_p): Delete.
15225         (scan_omp_target): Update call to scan_sharing_clauses.
15226         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
15227         FINALIZE}.
15228         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
15229         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15230         (convert_local_omp_clauses): Likewise.
15231         * tree-pretty-print.c (dump_omp_clause): Likewise.
15232         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
15233         FINALIZE}.
15234         (omp_clause_code_name): Likewise.
15236 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
15238         PR debug/86194
15239         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
15240         be narrowed.
15242         PR tree-optimization/86231
15243         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
15244         anti-range don't overwrite *vr0min before using it to compute *vr0max.
15246 2018-06-20  Tom de Vries  <tdevries@suse.de>
15248         PR tree-optimization/86097
15249         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
15250         iv type if signedness of iv type is not the same as that of *nit.
15252 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
15254         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
15255         EDGE_EH edges, verify they are all EDGE_EH.
15257 2018-06-20  Maya Rashish  <coypu@sdf.org>
15259         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
15261 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15263         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
15264         * config/aarch64/aarch64.c (xgene1_tunings): Add
15265         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
15266         (aarch64_mode_valid_for_sched_fusion_p):
15267         Allow 16-byte modes.
15268         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
15269         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
15270         128-bit modes.
15271         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
15272         New pattern.
15273         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
15274         * config/aarch64/iterators.md (VQ2): New mode iterator.
15276 2018-06-20  Martin Liska  <mliska@suse.cz>
15278         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
15279         Change default ratio from 10 to 8.
15281 2018-06-20  Martin Liska  <mliska@suse.cz>
15283         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15284         New.
15285         (bit_test_cluster::find_bit_tests): Likewise.
15286         (switch_decision_tree::analyze_switch_statement): Find clusters.
15287         * tree-switch-conversion.h (struct jump_table_cluster): Document
15288         hierarchy.
15290 2018-06-20  Martin Liska  <mliska@suse.cz>
15292         * tree-switch-conversion.c (switch_conversion::collect):
15293         Record m_uniq property.
15294         (switch_conversion::expand): Bail out for special conditions.
15295         (group_cluster::~group_cluster): New.
15296         (group_cluster::group_cluster): Likewise.
15297         (group_cluster::dump): Likewise.
15298         (jump_table_cluster::emit): New.
15299         (switch_decision_tree::fix_phi_operands_for_edges): New.
15300         (struct case_node): Remove struct.
15301         (jump_table_cluster::can_be_handled): New.
15302         (case_values_threshold): Moved to header.
15303         (reset_out_edges_aux): Likewise.
15304         (jump_table_cluster::is_beneficial): New.
15305         (bit_test_cluster::can_be_handled): Likewise.
15306         (add_case_node): Remove.
15307         (bit_test_cluster::is_beneficial): New.
15308         (case_bit_test::cmp): New.
15309         (bit_test_cluster::emit): New.
15310         (expand_switch_as_decision_tree_p): Remove.
15311         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
15312         (fix_phi_operands_for_edge): Likewise.
15313         (switch_decision_tree::analyze_switch_statement): New.
15314         (compute_cases_per_edge): Move ...
15315         (switch_decision_tree::compute_cases_per_edge): ... here.
15316         (try_switch_expansion): Likewise.
15317         (switch_decision_tree::try_switch_expansion): Likewise.
15318         (record_phi_operand_mapping): Likewise.
15319         (switch_decision_tree::record_phi_operand_mapping): Likewise.
15320         (emit_case_decision_tree): Likewise.
15321         (switch_decision_tree::emit): Likewise.
15322         (balance_case_nodes): Likewise.
15323         (switch_decision_tree::balance_case_nodes): Likewise.
15324         (dump_case_nodes): Likewise.
15325         (switch_decision_tree::dump_case_nodes): Likewise.
15326         (emit_jump): Likewise.
15327         (switch_decision_tree::emit_jump): Likewise.
15328         (emit_cmp_and_jump_insns): Likewise.
15329         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
15330         (emit_case_nodes): Likewise.
15331         (switch_decision_tree::emit_case_nodes): Likewise.
15332         (conditional_probability): Remove.
15333         * tree-switch-conversion.h (enum cluster_type): New.
15334         (PRINT_CASE): New.
15335         (struct cluster): Likewise.
15336         (cluster::cluster): Likewise.
15337         (struct simple_cluster): Likewise.
15338         (simple_cluster::simple_cluster): Likewise.
15339         (struct group_cluster): Likewise.
15340         (struct jump_table_cluster): Likewise.
15341         (struct bit_test_cluster): Likewise.
15342         (struct min_cluster_item): Likewise.
15343         (struct case_tree_node): Likewise.
15344         (case_tree_node::case_tree_node): Likewise.
15345         (jump_table_cluster::case_values_threshold): Likewise.
15346         (struct case_bit_test): Likewise.
15347         (struct switch_decision_tree): Likewise.
15348         (struct switch_conversion): Likewise.
15349         (switch_decision_tree::reset_out_edges_aux): Likewise.
15351 2018-06-20  Martin Liska  <mliska@suse.cz>
15353         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
15354         (hoist_edge_and_branch_if_true): Likewise.
15355         (expand_switch_using_bit_tests_p): Likewise.
15356         (struct case_bit_test): Likewise.
15357         (case_bit_test_cmp): Likewise.
15358         (emit_case_bit_tests): Likewise.
15359         (switch_conversion::switch_conversion): New class.
15360         (struct switch_conv_info): Remove old struct.
15361         (collect_switch_conv_info): More to ...
15362         (switch_conversion::collect): ... this.
15363         (check_range): Likewise.
15364         (switch_conversion::check_range): Likewise.
15365         (check_all_empty_except_final): Likewise.
15366         (switch_conversion::check_all_empty_except_final): Likewise.
15367         (check_final_bb): Likewise.
15368         (switch_conversion::check_final_bb): Likewise.
15369         (create_temp_arrays): Likewise.
15370         (switch_conversion::create_temp_arrays): Likewise.
15371         (free_temp_arrays): Likewise.
15372         (gather_default_values): Likewise.
15373         (switch_conversion::gather_default_values): Likewise.
15374         (build_constructors): Likewise.
15375         (switch_conversion::build_constructors): Likewise.
15376         (constructor_contains_same_values_p): Likewise.
15377         (switch_conversion::contains_same_values_p): Likewise.
15378         (array_value_type): Likewise.
15379         (switch_conversion::array_value_type): Likewise.
15380         (build_one_array): Likewise.
15381         (switch_conversion::build_one_array): Likewise.
15382         (build_arrays): Likewise.
15383         (switch_conversion::build_arrays): Likewise.
15384         (gen_def_assigns): Likewise.
15385         (switch_conversion::gen_def_assigns): Likewise.
15386         (prune_bbs): Likewise.
15387         (switch_conversion::prune_bbs): Likewise.
15388         (fix_phi_nodes): Likewise.
15389         (switch_conversion::fix_phi_nodes): Likewise.
15390         (gen_inbound_check): Likewise.
15391         (switch_conversion::gen_inbound_check): Likewise.
15392         (process_switch): Use the newly created class.
15393         (switch_conversion::expand): New.
15394         (switch_conversion::~switch_conversion): New.
15395         * tree-switch-conversion.h: New file.
15397 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15399         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
15400         tree-vect-patterns.c.
15401         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
15402         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
15403         (vect_recog_sad_pattern): Likewise.
15404         (vect_recog_widen_sum_pattern): Likewise.
15405         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
15406         (vect_recog_widen_shift_pattern): Remove the type_in argument.
15407         (vect_recog_rotate_pattern): Likewise.
15408         (vect_recog_mult_pattern): Likewise.
15409         (vect_recog_vector_vector_shift_pattern): Likewise.
15410         (vect_recog_divmod_pattern): Likewise.
15411         (vect_recog_mixed_size_cond_pattern): Likewise.
15412         (vect_recog_bool_pattern): Likewise.
15413         (vect_recog_mask_conversion_pattern): Likewise.
15414         (vect_try_gather_scatter_pattern): Likewise.
15415         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
15416         (vect_recog_over_widening_pattern): Likewise.
15417         (vect_recog_gather_scatter_pattern): Likewise.
15418         (vect_recog_func_ptr): Move from tree-vectorizer.h
15419         (vect_vect_recog_func_ptrs): Move further down the file.
15420         (vect_recog_func): Likewise.  Remove the third argument.
15421         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
15422         (vect_pattern_recog_1): Expect the pattern function to do any
15423         necessary target tests.  Also expect it to provide a vector type.
15424         Remove the type_in handling.
15426 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15428         * tree-vect-patterns.c (vect_pattern_detected): New function.
15429         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
15430         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
15431         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
15432         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
15433         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
15434         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
15435         (vect_recog_mask_conversion_pattern)
15436         (vect_try_gather_scatter_pattern): Likewise.
15438 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15440         * tree-vect-patterns.c (vect_get_internal_def): New function.
15441         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
15442         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
15443         (search_type_for_mask_1): Use it.
15445 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15447         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15448         redundant WIDEN_SUM_EXPR handling.
15449         (vect_recog_sad_pattern): Likewise.
15451 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15453         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15454         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
15455         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
15456         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
15457         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
15459 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15461         * tree-vect-stmts.c (vectorizable_call): Make sure that we
15462         use the stmt_vec_info of the original bb statement for the
15463         new zero assignment, even if the call is part of a pattern.
15465 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15467         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
15468         that the sequence is attached to the original statement rather
15469         than the pattern statement.
15470         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
15471         PATTERN_DEF_SEQ from the original statement rather than
15472         the main pattern statement.
15473         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
15474         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
15475         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
15477 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15479         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
15480         definition statements before the early exit for statements that aren't
15481         live or relevant.
15482         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
15483         split out from...
15484         (vect_transform_loop): ...here.  Process pattern definition
15485         statements without first checking whether the main pattern
15486         statement is live or relevant.
15488 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15490         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
15491         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
15493 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
15495         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
15496         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
15497         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
15498         (expand_block_compare): Change select_block_compare_mode call.
15499         (expand_strncmp_align_check): Use new functions, fix comment.
15500         (emit_final_str_compare_gpr): New function.
15501         (expand_strn_compare): Refactor and clean up code.
15502         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
15504 2018-06-19  Tony Reix  <tony.reix@atos.com>
15505             Damien Bergamini  <damien.bergamini@atos.com>
15506             David Edelsohn  <dje.gcc@gmail.com>
15508         * collect2.c (static_obj): New variable.
15509         (static_libs): New variable.
15510         (is_in_list): Uncomment declaration.
15511         (main): Track AIX libraries linked statically.
15512         (is_in_list): Uncomment definition.
15513         (scan_prog_file): Don't add AIX shared libraries initializer
15514         to constructor list if linking statically.
15516 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
15518         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
15519         constant.
15520         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
15522 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
15524         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
15525         blocks.
15527 2018-06-19  Martin Liska  <mliska@suse.cz>
15529         * config/i386/i386.c (ix86_can_inline_p): Do not use
15530         ipa_fn_summaries::get_create.
15531         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
15532         get.
15533         (devirtualization_time_bonus): Likewise.
15534         (ipcp_propagate_stage): Likewise.
15535         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
15536         (edge_set_predicate): Likewise.
15537         (evaluate_conditions_for_known_args): Likewise.
15538         (evaluate_properties_for_edge): Likewise.
15539         (ipa_call_summary::reset): Tranform to ...
15540         (ipa_call_summary::~ipa_call_summary): ... this.
15541         (ipa_fn_summary::reset): Transform to ...
15542         (ipa_fn_summary::~ipa_fn_summary): ... this.
15543         (ipa_fn_summary_t::remove): Rename to ...
15544         (ipa_fn_summary_t::remove_callees): ... this.
15545         (ipa_fn_summary_t::duplicate): Use placement new
15546         instead of memory copy.
15547         (ipa_call_summary_t::duplicate): Likewise.
15548         (ipa_call_summary_t::remove): Remove.
15549         (dump_ipa_call_summary): Change get_create to get.
15550         (ipa_dump_fn_summary): Dump only when summary exists.
15551         (analyze_function_body): Use symbol_summary::get instead
15552         of get_create.
15553         (compute_fn_summary): Likewise.
15554         (estimate_edge_devirt_benefit): Likewise.
15555         (estimate_edge_size_and_time): Likewise.
15556         (inline_update_callee_summaries): Likewise.
15557         (remap_edge_change_prob): Likewise.
15558         (remap_edge_summaries): Likewise.
15559         (ipa_merge_fn_summary_after_inlining): Likewise.
15560         (write_ipa_call_summary): Likewise.
15561         (ipa_fn_summary_write): Likewise.
15562         (ipa_free_fn_summary): Likewise.
15563         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
15564         (struct ipa_call_summary): Likewise.
15565         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
15566         of get_create.
15567         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
15568         (estimate_size_after_inlining): Likewise.
15569         (estimate_growth): Likewise.
15570         (growth_likely_positive): Likewise.
15571         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15572         (inline_call): Likewise.
15573         * ipa-inline.c (caller_growth_limits): Likewise.
15574         (can_inline_edge_p): Likewise.
15575         (can_inline_edge_by_limits_p): Likewise.
15576         (compute_uninlined_call_time): Likewise.
15577         (compute_inlined_call_time): Likewise.
15578         (want_inline_small_function_p): Likewise.
15579         (edge_badness): Likewise.
15580         (update_caller_keys): Likewise.
15581         (update_callee_keys): Likewise.
15582         (inline_small_functions): Likewise.
15583         (inline_to_all_callers_1): Likewise.
15584         (dump_overall_stats): Likewise.
15585         (early_inline_small_functions): Likewise.
15586         (early_inliner): Likewise.
15587         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15588         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15589         * ipa-pure-const.c (malloc_candidate_p): Likewise.
15590         * ipa-split.c (execute_split_functions): Likewise.
15591         * symbol-summary.h: Likewise.
15592         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15594 2018-06-19  Richard Biener  <rguenther@suse.de>
15596         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
15597         (vectorize_loops): ... here.  Fix dbgcnt handling.
15598         (try_vectorize_loop): Wrap try_vectorize_loop_1.
15600 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
15602         PR target/86197
15603         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
15604         ieee128 argument takes up only one (vector) register, not two (floating
15605         point) registers.
15607 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15609         * gimplify.c (gimplify_init_constructor): Really never clear for an
15610         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
15612 2018-06-19  Richard Biener  <rguenther@suse.de>
15614         PR tree-optimization/86179
15615         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
15616         after failed recognition.
15618 2018-06-18  Martin Sebor  <msebor@redhat.com>
15620         PR middle-end/85602
15621         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
15622         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
15623         Handle integer subtraction.
15624         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
15625         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
15627 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15629         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
15630         param from rtx to rtx_insn *.
15631         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
15632         param.
15633         (frv_ifcvt_modify_insn): Likwise.
15634         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
15635         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
15636         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
15637         as_a <rtx_insn *> cast to local "unprotected_region" once
15638         it's been established that it's not NULL or pc_rtx.
15639         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
15640         param "sethi" from rtx to rtx_insn *.
15641         (nds32_group_float_insns): Likewise for param "insn".
15642         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
15643         param.
15644         (vax_output_int_subtract): Likewise.
15645         * config/vax/vax.c (vax_output_int_add): Likewise for param
15646         "insn".
15647         (vax_output_int_subtract): Likewise.
15648         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
15649         (emit_pattern_after): Likewise for param "after".
15650         (emit_insn_after): Likewise.
15651         (emit_jump_insn_after): Likewise.
15652         (emit_call_insn_after): Likewise.
15653         (emit_debug_insn_after): Likewise.
15654         (emit_pattern_before): Likewise for param "before".
15655         (emit_insn_before): Likewise.
15656         (emit_jump_insn_before): Likewise.
15657         * final.c (get_insn_template): Likewise for param "insn", removing
15658         a cast.
15659         * output.h (get_insn_template): Likewise for 2nd param.
15660         * rtl.h (emit_insn_before): Likewise.
15661         (emit_jump_insn_before): Likewise.
15662         (emit_debug_insn_before_noloc): Likewise.
15663         (emit_insn_after): Likewise.
15664         (emit_jump_insn_after): Likewise.
15665         (emit_call_insn_after): Likewise.
15666         (emit_debug_insn_after): Likewise.
15667         (set_insn_deleted): Likewise for param.
15669 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
15671         PR target/85358
15672         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
15673         floating point modes, so that IFmode is numerically greater than
15674         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
15675         to declare the ordering.  This prevents IFmode from being
15676         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
15677         machine.  Include rs6000-modes.h to share the fractional values
15678         between genmodes* and the rest of the compiler.
15679         (IFmode): Likewise.
15680         (KFmode): Likewise.
15681         (TFmode): Likewise.
15682         * config/rs6000/rs6000-modes.h: New file.
15683         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
15684         meaning of rs6000_long_double_size so that 126..128 selects an
15685         appropriate 128-bit floating point type.
15686         (rs6000_option_override_internal): Likewise.
15687         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
15688         (TARGET_LONG_DOUBLE_128): Change the meaning of
15689         rs6000_long_double_size so that 126..128 selects an appropriate
15690         128-bit floating point type.
15691         (LONG_DOUBLE_TYPE_SIZE): Update comment.
15692         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
15693         source and destination to match the standard usage.
15694         (truncifkf2): Likewise.
15695         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
15696         ISA 2.07 to use an explicit clobber, instead of passing in a
15697         temporary.
15698         (copysign<mode>3_soft): Likewise.
15700 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15702         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
15703         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
15704         (vect_slp_analyze_instance_dependence): Likewise.
15705         (vect_enhance_data_refs_alignment): Likewise.
15706         (vect_analyze_data_refs_alignment): Likewise.
15707         (vect_slp_analyze_and_verify_instance_alignment
15708         (vect_analyze_data_ref_accesses): Likewise.
15709         (vect_prune_runtime_alias_test_list): Likewise.
15710         (vect_analyze_data_refs): Likewise.
15711         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
15712         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15713         (vect_analyze_scalar_cycles_1): Likewise.
15714         (vect_get_loop_niters): Likewise.
15715         (vect_analyze_loop_form_1): Likewise.
15716         (vect_update_vf_for_slp): Likewise.
15717         (vect_analyze_loop_operations): Likewise.
15718         (vect_analyze_loop): Likewise.
15719         (vectorizable_induction): Likewise.
15720         (vect_transform_loop): Likewise.
15721         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
15722         * tree-vect-slp.c (vect_analyze_slp): Likewise.
15723         (vect_make_slp_decision): Likewise.
15724         (vect_detect_hybrid_slp): Likewise.
15725         (vect_slp_analyze_operations): Likewise.
15726         (vect_slp_bb): Likewise.
15727         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
15728         (vectorizable_bswap): Likewise.
15729         (vectorizable_call): Likewise.
15730         (vectorizable_simd_clone_call): Likewise.
15731         (vectorizable_conversion): Likewise.
15732         (vectorizable_assignment): Likewise.
15733         (vectorizable_shift): Likewise.
15734         (vectorizable_operation): Likewise.
15735         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
15737 2018-06-18  Martin Sebor  <msebor@redhat.com>
15739         PR tree-optimization/81384
15740         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
15741         * builtins.c (expand_builtin_strnlen): New function.
15742         (expand_builtin): Call it.
15743         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
15744         * builtins.def (BUILT_IN_STRNLEN): New.
15745         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
15746         Warn for bounds in excess of maximum object size.
15747         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
15748         single-value ranges.  Handle strnlen.
15749         (handle_builtin_strlen): Handle strnlen.
15750         (strlen_check_and_optimize_stmt): Same.
15751         * doc/extend.texi (Other Builtins): Document strnlen.
15753 2018-06-18  Maya Rashish  <coypu@sdf.org>
15755         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
15756         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
15757         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
15759         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
15760         here to ...
15761         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
15763 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15765         * tree.c (escaped_string::escape): Replace cast to char * by
15766         const_cast<char *> (unescaped).
15768 2018-06-18  Nick Clifton  <nickc@redhat.com>
15770         PR 84195
15771         * tree.c (escaped_string): New class.  Converts an unescaped
15772         string into its escaped equivalent.
15773         (warn_deprecated_use): Use the new class to convert the
15774         deprecation message, if present.
15775         (test_escaped_strings): New self test.
15776         (test_c_tests): Add test_escaped_strings.
15777         * doc/extend.texi (deprecated): Add a note that the
15778         deprecation message is affected by the -fmessage-length
15779         option, and that control characters will be escaped.
15780         (#pragma GCC error): Document this pragma.
15781         (#pragma GCC warning): Likewise.
15782         * doc/invoke.texi (-fmessage-length): Document this option's
15783         effect on the #warning and #error preprocessor directives and
15784         the deprecated attribute.
15786 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
15788         * tree.c (decl_value_expr_lookup): Revert latest change.
15789         (decl_value_expr_insert): Likewise.
15791 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
15793         * gimplify.c (nonlocal_vlas): Delete.
15794         (nonlocal_vla_vars): Likewise.
15795         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
15796         referenced VLAs.
15797         (gimplify_body): Do not create and destroy nonlocal_vlas.
15798         * tree-nested.c: Include diagnostic.h.
15799         (use_pointer_in_frame): Tweak.
15800         (lookup_field_for_decl): Add assertion and declare the transformation.
15801         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
15802         internal error when the reference is in a wrong context.  Do not
15803         create a debug decl by default.
15804         (note_nonlocal_block_vlas): Delete.
15805         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
15806         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
15807         create a debug decl by default.
15808         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
15809         call is in a wrong context.
15810         (fixup_vla_decls): New function.
15811         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
15812         debug variables were created.
15813         * tree.c (decl_value_expr_lookup): Add checking assertion.
15814         (decl_value_expr_insert): Likewise.
15816 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15818         PR middle-end/82479
15819         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
15820         * tree-scalar-evolution.c (interpret_expr): Likewise.
15821         (expression_expensive_p): Likewise.
15822         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
15823         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
15824         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
15825         (ssa_defined_by_minus_one_stmt_p): New.
15827 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15829         PR middle-end/64946
15830         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
15831         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15832         * dojump.c (do_jump): Likewise.
15833         * expr.c (expand_expr_real_2): Check operand type's sign.
15834         * fold-const.c (const_unop): Handle ABSU_EXPR.
15835         (fold_abs_const): Likewise.
15836         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
15837         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
15838         (strip_sign_op_1): Likesise.
15839         * match.pd: Add new pattern to generate ABSU_EXPR.
15840         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
15841         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15842         * tree-eh.c (operation_could_trap_helper_p): Likewise.
15843         * tree-inline.c (estimate_operator_cost): Likewise.
15844         * tree-pretty-print.c (dump_generic_node): Likewise.
15845         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
15846         * tree.def (ABSU_EXPR): New.
15848 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
15850         PR middle-end/86095
15851         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
15852         documented as preserved for backward compatibility only.
15853         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
15855         PR rtl-optimization/86108
15856         * bb-reorder.c (create_forwarder_block): Renamed to ...
15857         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
15858         jump from new landing pad to the second part.
15859         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
15860         Adjust callers.
15862 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
15864         PR middle-end/85878
15865         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
15866         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
15867         Only call store_expr for halves if the mode is the same.
15869         PR middle-end/86123
15870         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
15871         Fix up comment formatting.
15873 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15875         * typed-splay-tree.h (typed_splay_tree::remove): New function.
15876         (typed_splay_tree::closure,
15877         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
15878         (typed_splay_tree::typed_splay_tree,
15879         typed_splay_tree::operator =): Declared private.
15880         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
15881         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
15882         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
15883         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
15884         typed_splay_tree::splay_tree_splay,
15885         typed_splay_tree::splay_tree_foreach_helper,
15886         typed_splay_tree::splay_tree_insert,
15887         typed_splay_tree::splay_tree_remove,
15888         typed_splay_tree::splay_tree_lookup,
15889         typed_splay_tree::splay_tree_predecessor,
15890         typed_splay_tree::splay_tree_successor,
15891         typed_splay_tree::splay_tree_min,
15892         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
15893         (typed_splay_tree::root, typed_splay_tree::comp,
15894         typed_splay_tree::delete_key,
15895         typed_splay_tree::delete_value): New data members.
15896         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
15897         typed_splay_tree::remove.
15899 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
15901         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
15902         -mginv and -mno-ginv to the assembler.
15903         * config/mips/mips.opt (-mcrc): New option.
15904         (-mginv): Likewise.
15905         * doc/invoke.text (-mcrc): Document.
15906         (-mginv): Likewise.
15908 2018-06-15  Nick Clifton  <nickc@redhat.com>
15910         PR 84195
15911         * tree.c (escaped_string): New class.  Converts an unescaped
15912         string into its escaped equivalent.
15913         (warn_deprecated_use): Use the new class to convert the
15914         deprecation message, if present.
15915         (test_escaped_strings): New self test.
15916         (test_c_tests): Add test_escaped_strings.
15917         * doc/extend.texi (deprecated): Add a note that the
15918         deprecation message is affected by the -fmessage-length
15919         option, and that control characters will be escaped.
15920         (#pragma GCC error): Document this pragma.
15921         (#pragma GCC warning): Likewise.
15922         * doc/invoke.texi (-fmessage-length): Document this option's
15923         effect on the #warning and #error preprocessor directives and
15924         the deprecated attribute.
15926 2018-06-15  Richard Biener  <rguenther@suse.de>
15928         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
15929         here, also noting vector size used.
15930         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
15931         size used in MSG_OPTIMIZED_LOCATIONS dump.
15932         (pass_slp_vectorize::execute): Adjust.
15934 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
15936         PR target/85968
15937         * config/arc/arc.c (arc_return_address_register): Fix
15938         if-condition.
15940 2018-06-15  Richard Biener  <rguenther@suse.de>
15942         PR middle-end/86159
15943         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
15944         leave useless conversion stripping to force_gimple_operand_gsi.
15945         (gimplify_build2): Likewise.
15946         (gimplify_build1): Likewise.
15948 2018-06-15  Richard Biener  <rguenther@suse.de>
15950         PR middle-end/86076
15951         * tree-cfg.c (move_stmt_op): unshare invariant addresses
15952         before adjusting their block.
15954 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15956         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
15957         multilibs for *-*-rtems*.
15958         * config/riscv/t-rtems: New file.
15960 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
15962         PR middle-end/86122
15963         * match.pd ((A +- CST1) +- CST2): Punt if last resort
15964         unsigned_type_for returns NULL.
15966         PR target/85945
15967         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
15968         subregs of multi-word pseudos unless the float mode has word size.
15970 2018-06-14  Richard Biener  <rguenther@suse.de>
15972         PR middle-end/86139
15973         * tree-vect-generic.c (build_word_mode_vector_type): Remove
15974         duplicate and harmful type_hash_canon.
15975         * tree.c (type_hash_canon): Assert we didn't find ourselves.
15977 2018-06-14  Richard Biener  <rguenther@suse.de>
15979         PR ipa/86124
15980         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
15981         NULL cgraph_node.
15983 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15985         * config/rtems.h (STDINT_LONG32): Define.
15987 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
15988             Prachi Godbole  <prachi.godbole@imgtec.com>
15990         * config/mips/mips-cpus.def: Define P6600.
15991         * config/mips/mips-tables.opt: Regenerate.
15992         * config/mips/mips.c (mips_ucbranch_type): New enum.
15993         (mips_rtx_cost_data): Add support for P6600.
15994         (mips_issue_rate): Likewise.
15995         (mips_multipass_dfa_lookahead): Likewise.
15996         (mips_avoid_hazard): Likewise.
15997         (mips_reorg_process_insns): Likewise.
15998         (mips_classify_branch_p6600): New function.
15999         * config/mips/mips.h (TUNE_P6600): New define.
16000         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
16001         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
16002         * config/mips/mips.md: Include p6600.md.
16003         (processor): Add p6600.
16004         * config/mips/p6600.md: New file.
16005         * doc/invoke.texi: Add p6600 to supported architectures.
16007 2018-06-13  Martin Sebor  <msebor@redhat.com>
16009         PR tree-optimization/86114
16010         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
16011         of integer types.
16012         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
16014 2018-06-13  Richard Biener  <rguenther@suse.de>
16016         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
16017         Properly set vector type of the intermediate stmt.
16018         * tree-vect-stmts.c (vectorizable_operation): The destination
16019         var always has vectype_out type.
16021 2018-06-13  Jeff Law  <law@redhat.com>
16023         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
16024         integer 0 for argument to print_rtl_with_bb.
16025         (rl78_reorg): Likewise.
16027 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
16029         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
16030         from rtx to rtx_insn *.
16031         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
16032         "label".
16033         (add_sched_insns_for_speculation): Likewise for local "target",
16034         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
16035         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
16036         from rtx_insn ** to rtx_code_label **.
16037         (reorg_emit_nops): Likewise.
16038         (c6x_reorg): Likewise for local "call_labels".
16039         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
16040         rtx to rtx_insn *.
16041         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
16042         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
16043         the loops over LABEL_REFS.
16044         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
16045         braf_label.
16046         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
16047         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
16048         (split_branches): Strengthen local "olabel" from rtx to
16049         rtx_insn *, adding a safe_as_a cast.
16050         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
16051         to "rtx_insn *".
16052         (add_insn_after): Likewise for first two params.
16053         (add_insn_before): Likewise.
16054         (remove_insn): Likewise for param.
16055         (emit_pattern_before_noloc): Likewise for second and third params.
16056         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
16057         (emit_call_insn_before_noloc): Likewise.
16058         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
16059         to "rtx_insn *".
16060         (emit_barrier_before): Likewise.
16061         (emit_label_before): Strengthen "label" param from "rtx" to
16062         "rtx_code_label *".  Strengthen "before" param from "rtx" to
16063         "rtx_insn *".
16064         (emit_insn_after_1): Strengthen "after" param from "rtx" to
16065         "rtx_insn *".
16066         (emit_pattern_after_noloc): Likewise.
16067         (emit_insn_after_noloc): Likewise.
16068         (emit_jump_insn_after_noloc): Likewise.
16069         (emit_call_insn_after_noloc): Likewise.
16070         (emit_debug_insn_after_noloc): Likewise.
16071         (emit_barrier_after): Likewise.
16072         (emit_label_after): Likewise for both params.
16073         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
16074         "loc" param from "int" to "location_t".
16075         (emit_insn_after_setloc): Likewise.
16076         (emit_jump_insn_after_setloc): Likewise.
16077         (emit_call_insn_after_setloc): Likewise.
16078         (emit_debug_insn_after_setloc): Likewise.
16079         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
16080         "loc" param from "int" to "location_t".
16081         (emit_pattern_before): Convert NULL_RTX to NULL.
16082         (emit_insn_before_setloc): Convert "loc" param from "int" to
16083         "location_t".
16084         (emit_jump_insn_before_setloc): Likewise.
16085         (emit_call_insn_before_setloc): Likewise.
16086         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
16087         rtx_insn *.  Convert "loc" param from "int" to "location_t".
16088         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
16089         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
16090         Convert 3rd param from "int" to "location_t".
16091         (emit_barrier_before, emit_barrier_after, next_real_insn):
16092         Strengthen param from rtx to rtx_insn *.
16093         (emit_label_before): Strengthen 1st param from "rtx" to
16094         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
16095         "rtx_insn *".
16096         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
16097         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
16098         Strengthen 2nd param from "rtx" to "rtx_insn *".
16099         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
16100         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
16101         Likewise. Convert 3rd param from "int" to "location_t".
16102         (emit_label_after): Strengthen 1st param from "rtx" to
16103         "rtx_code_label *".
16104         (next_real_insn, remove_insn): Strengthen param from "rtx" to
16105         "rtx_insn *".
16106         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
16107         from "rtx" to "rtx_insn *".
16109 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
16111         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
16112         bodies streamed in with -Q.
16113         * dumpfile.c (dump_files): Add lto-stream-out dump file.
16114         * dumpfile.h (tree_dump_index): Add lto_stream_out.
16115         * gimple-streamer-out.c: Include gimple-pretty-print.h
16116         (output_bb): Dump stmts streamed.
16117         * lto-section-out.c: Include print-tree.h
16118         (lto_begin_section): Dump sections created.
16119         (lto_output_decl_index): Dump decl encoded.
16120         * lto-streamer-out.c: Include print-tree.h
16121         (create_output_block): Dump output block created.
16122         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
16123         (output_function): Dump function output.
16124         (output_constructor): Dump constructor streamed.
16125         (write_global_stream): Output indexes encoded.
16126         (produce_asm_for_decls): Dump streams encoded.
16127         * lto-streamer.c (streamer_dump_file): New global var.
16128         * lto-streamer.h (streamer_dump_file): Declare.
16129         * passes.c (ipa_write_summaries): Initialize streamer dump.
16130         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
16131         in.
16133 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
16135         PR target/86048
16136         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
16137         offsets for register save directives.  Emit a second batch of save
16138         directives, if need be, when the function accesses prior frames.
16140 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16142         * config/arc/fpu.md (fmasf4): Force operand to register.
16143         (fnmasf4): Likewise.
16145 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16147         * config/arc/arc-protos.h (arc_pad_return): Remove.
16148         * config/arc/arc.c (machine_function): Remove force_short_suffix
16149         and size_reason.
16150         (arc_print_operand): Adjust printing of '&'.
16151         (arc_verify_short): Remove conditional printing of short suffix.
16152         (arc_final_prescan_insn): Remove reference to size_reason.
16153         (pad_return): New function.
16154         (arc_reorg): Call pad_return.
16155         (arc_pad_return): Remove.
16156         (arc_init_machine_status): Remove reference to force_short_suffix.
16157         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
16158         (attr length): When attribute iscompact is true force to 2
16159         regardless; in the case of maybe check if we want to force the
16160         instruction to have 4 bytes length.
16161         (nopv): Change it to generate 4 byte long nop as well.
16162         (blockage): New pattern.
16163         (simple_return): Remove call to arc_pad_return.
16164         (p_return_i): Likewise.
16166 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16168         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
16170 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16172         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
16173         ARC cores.
16175 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16177         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
16178         for ARC700 and ARCv2.
16180 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
16182         PR target/86076
16183         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
16184         operands[2] instead of operands[1].
16187 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
16189         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
16190         case, check whether the outer register overlaps an unallocatable
16191         register, not just whether it fits the required class.
16193 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
16195         * poly-int.h (can_div_trunc_p): Add new overload in which all values
16196         are poly_ints.
16197         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
16198         (memrefs_conflict_p): Likewise.
16199         (init_alias_analysis): Likewise.
16200         * cfgexpand.c (expand_debug_expr): Likewise.
16201         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
16202         * cse.c (fold_rtx): Likewise.
16203         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
16204         * expr.c (emit_block_move_hints): Likewise.
16205         (clear_storage_hints, push_block, emit_push_insn): Likewise.
16206         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
16207         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
16208         (emit_group_store): Likewise.
16209         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
16210         to read the PRE/POST_MODIFY increment.
16211         * calls.c (store_one_arg): Use strip_offset.
16212         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
16213         poly_int_rtx_p.
16214         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
16215         by a VEC_SELECT.
16216         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
16217         (simplify_binary_operation_1): Extend CONST_INT handling to
16218         poly_int_rtx_p.
16219         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
16220         than a HOST_WIDE_INT.
16221         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
16222         poly_int64.
16223         (adjust_mems, add_stores): Update accodingly.
16224         (vt_canonicalize_addr): Track polynomial offsets.
16225         (emit_note_insn_var_location): Likewise.
16226         (vt_add_function_parameter): Likewise.
16227         (vt_initialize): Likewise.
16229 2018-06-12  Jeff Law  <law@redhat.com>
16231         * config.gcc (alpha*-*-freebsd*): Remove.
16232         * config/alpha/freebsd.h: Remove.
16234 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
16236         PR other/69968
16237         * spellcheck-tree.c (levenshtein_distance): Rename to...
16238         (get_edit_distance): ...this, and update for underlying renaming.
16239         * spellcheck-tree.h (levenshtein_distance): Rename to...
16240         (get_edit_distance): ...this.
16241         * spellcheck.c (levenshtein_distance): Rename to...
16242         (get_edit_distance): ...this.  Convert from Levenshtein distance
16243         to Damerau-Levenshtein distance by supporting transpositions of
16244         adjacent characters.  Rename "v1" to "v_next" and "v0" to
16245         "v_one_ago".
16246         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
16247         (selftest::test_edit_distance_unit_test_oneway): ...this, and
16248         update for underlying renaming.
16249         (selftest::levenshtein_distance_unit_test): Rename to...
16250         (selftest::test_get_edit_distance_unit): ...this, and update for
16251         underlying renaming.
16252         (selftest::test_find_closest_string): Add example from PR 69968
16253         where transposition helps
16254         (selftest::test_metric_conditions): Update for renaming.
16255         (selftest::test_metric_conditions): Likewise.
16256         (selftest::spellcheck_c_tests): Likewise.
16257         * spellcheck.h (levenshtein_distance): Rename both overloads to...
16258         (get_edit_distance): ...this.
16259         (best_match::consider): Update for renaming.
16261 2018-06-12  Martin Sebor  <msebor@redhat.com>
16263         PR tree-optimization/85259
16264         * builtins.c (compute_objsize): Handle constant offsets.
16265         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
16266         true iff a warning has been issued.
16267         * gimple.h (gimple_nonartificial_location): New function.
16268         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
16269         gimple_nonartificial_location and handle -Wno-system-headers.
16270         (handle_builtin_stxncpy): Same.
16272 2018-06-12  Martin Sebor  <msebor@redhat.com>
16274         PR c/85931
16275         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
16277 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
16279         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16280         BUILTIN_VEC_XST entries for pointer to double and long long.
16282 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
16284         PR target/85990
16285         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16286         Update comments.
16287         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16288         Likewise.
16290 2018-06-12  Martin Liska  <mliska@suse.cz>
16292         * doc/options.texi: Document IntegerRange.
16294 2018-06-12  Martin Liska  <mliska@suse.cz>
16296         * config/i386/i386.opt: Make MPX-related options as Deprecated.
16297         * opt-functions.awk: Handle Deprecated flag.
16298         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
16299         and report error.
16300         (read_cmdline_option): Report warning for a deprecated option.
16301         * opts.h (struct cl_option): Add new field cl_deprecated.
16302         (CL_ERR_DEPRECATED): New.
16304 2018-06-12  Martin Liska  <mliska@suse.cz>
16306         * doc/options.texi: Document Deprecated option flag.
16308 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16310         * config/arc/arc-arch.h (arc_extras): New enum.
16311         (arc_cpu_t):Add field extra.
16312         (arc_cpu_types): Consider the extras.
16313         * config/arc/arc-cpus.def: Add extras info.
16314         * config/arc/arc-opts.h (processor_type): Consider extra field.
16315         * config/arc/arc.c (arc_override_options): Handle extra field.
16317 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16319         * config/arc/arc-arch.h: Update ARC_OPTX macro.
16320         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
16321         field.
16322         * config/arc/arc.c (arc_init): Update pic warning.
16323         (irq_range): Update irq range parsing warnings.
16324         (arc_override_options): Update various warning messages.
16325         (arc_handle_aux_attribute): Likewise.
16327 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
16329         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
16331 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16333         * doc/sourcebuild.texi: Document usage of line number 0 in verify
16334         compiler messages directives.
16336 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
16338         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
16339         * config/mips/mips-tables.opt: Regenerate.
16340         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
16341         mips64r6.
16342         * doc/invoke.texi: Document -march=i6500.
16344 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
16346         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
16347         (i6400_gpmul): Add cpu_unit.
16348         (i6400_gpdiv): Likewise.
16349         (i6400_msa_add_d): Update reservations.
16350         (i6400_msa_int_add) Likewise.
16351         (i6400_msa_short_logic3) Likewise.
16352         (i6400_msa_short_logic2) Likewise.
16353         (i6400_msa_short_logic) Likewise.
16354         (i6400_msa_move) Likewise.
16355         (i6400_msa_cmp) Likewise.
16356         (i6400_msa_short_float2) Likewise.
16357         (i6400_msa_div_d) Likewise.
16358         (i6400_msa_long_logic1) Likewise.
16359         (i6400_msa_long_logic2) Likewise.
16360         (i6400_msa_mult) Likewise.
16361         (i6400_msa_long_float2) Likewise.
16362         (i6400_msa_long_float4) Likewise.
16363         (i6400_msa_long_float5) Likewise.
16364         (i6400_msa_long_float8) Likewise.
16365         (i6400_fpu_fadd): Include frint type.
16366         (i6400_fpu_store): New define_insn_reservation.
16367         (i6400_fpu_load): Likewise.
16368         (i6400_fpu_move): Likewise.
16369         (i6400_fpu_fcmp): Likewise.
16370         (i6400_fpu_fmadd): Likewise.
16371         (i6400_int_mult): Include imul3nc type and update reservation.
16372         (i6400_int_div): Include idiv3 type and update reservation.
16373         (i6400_int_load): Update to check type not move_type.
16374         (i6400_int_store): Likewise.
16375         (i6400_int_prefetch): Set zero latency.
16377 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
16379         * gcc.c: Document new %@{...} sequence.
16380         (LINK_COMMAND_SPEC): Use it for the -L switches.
16381         (cpp_unique_options): Use it for the -I switches.
16382         (at_file_argbuf): New global variable.
16383         (in_at_file): Likewise.
16384         (alloc_args): Create at_file_argbuf.
16385         (clear_args): Truncate at_file_argbuf.
16386         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
16387         (open_at_file): New function.
16388         (close_at_file): Likewise.
16389         (create_at_file): Delete.
16390         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
16391         <'o'>: Likewise.
16392         <'@'>: New case.
16393         (validate_switches_from_spec): Deal with %@{...} sequence.
16394         (validate_switches): Likewise.
16395         (driver::finalize): Call clear_args.
16397 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
16399         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
16401 2018-06-11  Martin Sebor  <msebor@redhat.com>
16403         * doc/invoke.texi (-Wall): List -Wc++17-compat.
16404         (Wno-class-memaccess): Add @opindex.
16405         (Wno-templates, Wno-multiple-inheritance): Same.
16406         (Wno-virtual-inheritance, Wno-namespaces): Same.
16407         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
16408         (Wno-format-overflow, Wno-format-truncation): Same.
16409         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
16410         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
16411         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
16412         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
16413         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
16414         (Wno-misspelled-isr): Same.
16416 2018-06-11  Martin Sebor  <msebor@redhat.com>
16418         * PR tree-optimization/86083
16419         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
16421 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
16423         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
16425 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
16427         PR target/85755
16428         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
16429         on the correct operand.
16430         (*movdi_internal64): Ditto.
16432 2018-06-11  Martin Liska  <mliska@suse.cz>
16434         PR tree-optimization/86089
16435         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
16437 2018-06-11  Julia Koval  <julia.koval@intel.com>
16439         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
16440         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
16441         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
16443 2018-06-11  Olivier Hainque  <hainque@adacore.com>
16445         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
16446         for Ada with strict dwarf2.
16448 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
16450         PR target/85755
16451         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
16452         addresses.
16454 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
16456         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
16458 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
16460         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
16461         TARGET_ELF.
16463 2018-06-08  Martin Liska  <mliska@suse.cz>
16465         * tree-cfg.h (debug_function): Fix argument type to match
16466         implementation.
16468 2018-06-08  Martin Liska  <mliska@suse.cz>
16470         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
16471         Remove usage of MPX-related (and removed) fields.
16472         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
16474 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
16476         * cfg.c (debug): Use TDF_NONE rather than 0.
16477         * cfghooks.c (debug): Likewise.
16478         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
16479         (struct dump_option_value_info): Convert to...
16480         (struct kv_pair): ...this template type.
16481         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
16482         rather than 0.
16483         (optinfo_verbosity_options): Likewise.
16484         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
16485         OPTGROUP_NONE.
16486         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
16487         than int for "optgroup_flags" param.
16488         (dump_generic_expr_loc): Use dump_flags_t rather than int for
16489         "dump_kind" param.
16490         (dump_dec): Likewise.
16491         (dump_finish): Use TDF_NONE rather than 0.
16492         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
16493         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
16494         than 0.  Update for change to option_ptr.
16495         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
16496         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
16497         0.  Update for changes to optinfo_verbosity_options and
16498         optgroup_options.
16499         (opt_info_switch_p): Convert optgroup_flags from int to
16500         optgroup_flags_t.
16501         (dump_basic_block): Use dump_flags_t rather than int
16502         for "dump_kind" param.
16503         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
16504         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
16505         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
16506         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
16507         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
16508         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
16509         TDF_NONE): Convert from macros to...
16510         (enum dump_flag): ...this new enum.
16511         (dump_flags_t): Update to use enum.
16512         (operator|, operator&, operator~, operator|=, operator&=):
16513         Implement for dump_flags_t.
16514         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
16515         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
16516         Convert from macros to...
16517         (enum optgroup_flag): ...this new enum.
16518         (optgroup_flags_t): New typedef.
16519         (operator|, operator|=): Implement for optgroup_flags_t.
16520         (struct dump_file_info): Convert field "alt_flags" to
16521         dump_flags_t.  Convert field "optgroup_flags" to
16522         optgroup_flags_t.
16523         (dump_basic_block): Use dump_flags_t rather than int for param.
16524         (dump_generic_expr_loc): Likewise.
16525         (dump_dec): Likewise.
16526         (dump_register): Convert param "optgroup_flags" to
16527         optgroup_flags_t.
16528         (opt_info_enable_passes): Likewise.
16529         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
16530         than 0.
16531         * gimple-pretty-print.c (debug): Likewise.
16532         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
16533         (merged_store_group::apply_stores): Likewise.
16534         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
16535         * gimple.c (verify_gimple_pp): Likewise.
16536         * graphite-poly.c (print_pbb_body): Likewise.
16537         * passes.c (pass_manager::register_one_dump_file): Convert
16538         local "optgroup_flags" to optgroup_flags_t.
16539         * print-tree.c (print_node): Use TDF_NONE rather than 0.
16540         (debug): Likewise.
16541         (debug_body): Likewise.
16542         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
16543         to optgroup_flags_t.
16544         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
16545         than 0.
16546         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
16547         (convert_mult_to_fma): Likewise.
16548         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
16549         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
16550         * tree-vect-data-refs.c (dump_lower_bound): Convert param
16551         "dump_kind" to dump_flags_t.
16553 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16555         * config/rs6000/rs6000.c (min, max): Delete.
16557 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16559         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
16560         -mabi=spe and -mabi=no-spe.
16562 2018-06-08  Martin Liska  <mliska@suse.cz>
16564         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
16565         where we expect an existing summary.
16567 2018-06-08  Martin Liska  <mliska@suse.cz>
16569         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
16570         * ipa-inline.h (estimate_edge_growth): Likewise.
16572 2018-06-08  Martin Liska  <mliska@suse.cz>
16574         * cgraph.c (function_version_hasher::hash): Use
16575         cgraph_node::get_uid ().
16576         (function_version_hasher::equal):
16577         * cgraph.h (cgraph_node::get_uid): New method.
16578         * ipa-inline.c (update_caller_keys): Use
16579         cgraph_node::get_uid ().
16580         (update_callee_keys): Likewise.
16581         * ipa-utils.c (searchc): Likewise.
16582         (ipa_reduced_postorder): Likewise.
16583         * lto-cgraph.c (input_node): Likewise.
16584         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
16585         * symbol-summary.h (symtab_insertion): Likewise.
16586         (symtab_removal): Likewise.
16587         (symtab_duplication): Likewise.
16588         * tree-pretty-print.c (dump_function_header): Likewise.
16589         * tree-sra.c (convert_callers_for_node): Likewise.
16591 2018-06-08  Martin Liska  <mliska@suse.cz>
16593         * cgraph.c (symbol_table::create_edge): Always assign a new
16594         unique number.
16595         (symbol_table::free_edge): Do not recycle numbers.
16596         * cgraph.h (cgraph_edge::get): New method.
16597         * symbol-summary.h (symtab_removal): Use it.
16598         (symtab_duplication): Likewise.
16599         (call_summary::hashable_uid): Remove.
16601 2018-06-08  Martin Liska  <mliska@suse.cz>
16603         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
16604         (initialize_growth_caches): Remove.
16605         (free_growth_caches): Likewise.
16606         (do_estimate_edge_time): Use edge_growth_cache.
16607         (do_estimate_edge_size): Likewise.
16608         (do_estimate_edge_hints): Likewise.
16609         * ipa-inline.c (reset_edge_caches): Likewise.
16610         (recursive_inlining): Likewise.
16611         (inline_small_functions): Likewise.
16612         * ipa-inline.h (initialize_growth_caches): Remove.
16613         (estimate_edge_size): Likewise.
16614         (estimate_edge_time): Likewise.
16615         (estimate_edge_hints): Likewise.
16616         (reset_edge_growth_cache): Likewise.
16617         * symbol-summary.h (call_summary::remove): New method.
16619 2018-06-08  Martin Liska  <mliska@suse.cz>
16621         * ipa-cp.c (class edge_clone_summary): New summary.
16622         (grow_edge_clone_vectors): Remove.
16623         (ipcp_edge_duplication_hook): Remove.
16624         (class edge_clone_summary_t): New call_summary class.
16625         (ipcp_edge_removal_hook): Remove.
16626         (edge_clone_summary_t::duplicate): New function.
16627         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
16628         (create_specialized_node): Likewise.
16629         (ipcp_driver): Initialize edge_clone_summaries and do not
16630         register hooks.
16632 2018-06-08  Martin Liska  <mliska@suse.cz>
16634         * symbol-summary.h (get): New function.
16635         (call_summary::m_initialize_when_cloning): New class member.
16637 2018-06-08  Martin Liska  <mliska@suse.cz>
16639         * cgraph.c (cgraph_node::remove): Do not recycle uid.
16640         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
16641         (symbol_table::allocate_cgraph_symbol): Do not set uid.
16642         * passes.c (uid_hash_t): Record removed_nodes by their uids.
16643         (remove_cgraph_node_from_order): Use the removed_nodes set.
16644         (do_per_function_toporder): Likwise.
16645         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
16646         instead of summary_uid.
16647         (symtab_removal): Likewise.
16648         (symtab_duplication): Likewise.
16650 2018-06-08  Martin Liska  <mliska@suse.cz>
16652         * ipa-cp.c (ipcp_store_bits_results): Use
16653         ipcp_transformation_sum.
16654         (ipcp_store_vr_results): Likewise.
16655         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
16656         to ...
16657         (ipcp_transformation_initialize): ... this.
16658         (ipa_set_node_agg_value_chain):
16659         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
16660         (write_ipcp_transformation_info): Likewise.
16661         (read_ipcp_transformation_info): Likewise.
16662         (ipcp_update_bits): Likewise.
16663         (ipcp_update_vr): Likewise.
16664         (ipcp_transform_function): Likewise.
16665         * ipa-prop.h: Rename ipcp_transformation_summary to
16666         ipcp_transformation.
16667         (class ipcp_transformation_t): New function summary.
16668         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
16669         (ipa_get_agg_replacements_for_node): Likewise.
16671 2018-06-08  Martin Liska  <mliska@suse.cz>
16673         * ipa-pure-const.c (struct funct_state_d): Do it class instead
16674         of struct.
16675         (class funct_state_summary_t): New function_summary class.
16676         (has_function_state): Remove.
16677         (get_function_state): Likewise.
16678         (set_function_state): Likewise.
16679         (add_new_function): Likewise.
16680         (funct_state_summary_t::insert): New function.
16681         (duplicate_node_data): Remove.
16682         (remove_node_data): Remove.
16683         (funct_state_summary_t::duplicate): New function.
16684         (register_hooks): Create new funct_state_summaries.
16685         (pure_const_generate_summary): Use it.
16686         (pure_const_write_summary): Likewise.
16687         (pure_const_read_summary): Likewise.
16688         (propagate_pure_const): Likewise.
16689         (propagate_nothrow): Likewise.
16690         (dump_malloc_lattice): Likewise.
16691         (propagate_malloc): Likewise.
16692         (execute): Do not register hooks, just remove summary
16693         instead.
16694         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
16695         constructor.
16697 2018-06-08  Martin Liska  <mliska@suse.cz>
16699         * ipa-reference.c (remove_node_data): Remove.
16700         (duplicate_node_data): Likewise.
16701         (class ipa_ref_var_info_summary_t): New class.
16702         (class ipa_ref_opt_summary_t): Likewise.
16703         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
16704         (get_reference_optimization_summary): Use
16705         ipa_ref_opt_sum_summaries.
16706         (set_reference_vars_info): Remove.
16707         (set_reference_optimization_summary): Likewise.
16708         (ipa_init): Create summaries.
16709         (init_function_info): Use function summary.
16710         (ipa_ref_opt_summary_t::duplicate): New function.
16711         (ipa_ref_opt_summary_t::remove): New function.
16712         (get_read_write_all_from_node): Fix GNU coding style.
16713         (propagate): Use function summary.
16714         (write_node_summary_p): Fix GNU coding style.
16715         (stream_out_bitmap): Likewise.
16716         (ipa_reference_read_optimization_summary): Use function summary.
16717         (ipa_reference_c_finalize): Do not release hooks.
16719 2018-06-08  Martin Liska  <mliska@suse.cz>
16721         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
16722         (analyze_function_body): Extract multiple calls of get_create.
16723         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16724         * ipa-inline.c (recursive_inlining): Use ::get method.
16725         * ipa-inline.h (estimate_edge_growth): Likewise.
16727 2018-06-08  Martin Liska  <mliska@suse.cz>
16729         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
16730         HSA_INVALID.
16731         (hsa_function_summary::hsa_function_summary): Use the new enum
16732         value.
16733         (hsa_gpu_implementation_p): Use hsa_summaries::get.
16734         * hsa-gen.c (hsa_get_host_function): Likewise.
16735         (get_brig_function_name): Likewise.
16736         * ipa-hsa.c (process_hsa_functions): Likewise.
16737         (ipa_hsa_write_summary): Likewise.
16738         * symbol-summary.h (symtab_duplication): Use ::get function/
16739         (get): New function.
16741 2018-06-08  Martin Liska  <mliska@suse.cz>
16743         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
16744         of get.
16745         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
16746         (hsa_register_kernel): Likewise.
16747         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
16748         * hsa-gen.c (hsa_get_host_function): Likewise.
16749         (get_brig_function_name): Likewise.
16750         (generate_hsa): Likewise.
16751         (pass_gen_hsail::execute): Likewise.
16752         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
16753         (devirtualization_time_bonus): Likewise.
16754         (ipcp_propagate_stage): Likewise.
16755         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
16756         (edge_set_predicate): Likewise.
16757         (evaluate_conditions_for_known_args): Likewise.
16758         (evaluate_properties_for_edge): Likewise.
16759         (ipa_fn_summary::reset): Likewise.
16760         (ipa_fn_summary_t::duplicate): Likewise.
16761         (dump_ipa_call_summary): Likewise.
16762         (ipa_dump_fn_summary): Likewise.
16763         (analyze_function_body): Likewise.
16764         (compute_fn_summary): Likewise.
16765         (estimate_edge_devirt_benefit): Likewise.
16766         (estimate_edge_size_and_time): Likewise.
16767         (estimate_calls_size_and_time): Likewise.
16768         (estimate_node_size_and_time): Likewise.
16769         (inline_update_callee_summaries): Likewise.
16770         (remap_edge_change_prob): Likewise.
16771         (remap_edge_summaries): Likewise.
16772         (ipa_merge_fn_summary_after_inlining): Likewise.
16773         (ipa_update_overall_fn_summary): Likewise.
16774         (read_ipa_call_summary): Likewise.
16775         (inline_read_section): Likewise.
16776         (write_ipa_call_summary): Likewise.
16777         (ipa_fn_summary_write): Likewise.
16778         (ipa_free_fn_summary): Likewise.
16779         * ipa-hsa.c (process_hsa_functions): Likewise.
16780         (ipa_hsa_write_summary): Likewise.
16781         (ipa_hsa_read_section): Likewise.
16782         * ipa-icf.c (sem_function::merge): Likewise.
16783         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16784         (do_estimate_edge_time): Likewise.
16785         (estimate_size_after_inlining): Likewise.
16786         (estimate_growth): Likewise.
16787         (growth_likely_positive): Likewise.
16788         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
16789         (inline_call): Likewise.
16790         * ipa-inline.c (caller_growth_limits): Likewise.
16791         (can_inline_edge_p): Likewise.
16792         (can_inline_edge_by_limits_p): Likewise.
16793         (compute_uninlined_call_time): Likewise.
16794         (compute_inlined_call_time): Likewise.
16795         (want_inline_small_function_p): Likewise.
16796         (edge_badness): Likewise.
16797         (update_caller_keys): Likewise.
16798         (update_callee_keys): Likewise.
16799         (recursive_inlining): Likewise.
16800         (inline_small_functions): Likewise.
16801         (inline_to_all_callers_1): Likewise.
16802         (dump_overall_stats): Likewise.
16803         (early_inline_small_functions): Likewise.
16804         (early_inliner): Likewise.
16805         * ipa-inline.h (estimate_edge_growth): Likewise.
16806         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
16807         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16808         * ipa-prop.h (IPA_NODE_REF): Likewise.
16809         (IPA_EDGE_REF): Likewise.
16810         * ipa-pure-const.c (malloc_candidate_p): Likewise.
16811         (propagate_malloc): Likewise.
16812         * ipa-split.c (execute_split_functions): Likewise.
16813         * symbol-summary.h: Rename get to get_create.
16814         (get): Likewise.
16815         (get_create): Likewise.
16816         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
16818 2018-06-08  Martin Liska  <mliska@suse.cz>
16820         * symbol-summary.h (release): Move definition out of class
16821         declaration.
16822         (symtab_removal): Likewise.
16823         (symtab_duplication): Likewise.
16825 2018-06-08  Martin Liska  <mliska@suse.cz>
16827         * symbol-summary.h (function_summary): Move constructor
16828         implementation out of class declaration.
16829         (release): Likewise.
16830         (symtab_insertion): Likewise.
16831         (symtab_removal): Likewise.
16832         (symtab_duplication): Likewise.
16833         (get): Likewise.
16835 2018-06-08  Martin Liska  <mliska@suse.cz>
16837         * Makefile.in: Remove support for MPX (macros, related functions,
16838         fields in cgraph_node, ...).
16839         * builtin-types.def (BT_BND): Likewise.
16840         (BT_FN_BND_CONST_PTR): Likewise.
16841         (BT_FN_CONST_PTR_BND): Likewise.
16842         (BT_FN_VOID_PTR_BND): Likewise.
16843         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
16844         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
16845         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
16846         (expand_builtin_mempcpy_with_bounds): Likewise.
16847         (expand_builtin_memset_with_bounds): Likewise.
16848         (expand_builtin_memset_args): Likewise.
16849         (std_expand_builtin_va_start): Likewise.
16850         (expand_builtin): Likewise.
16851         (expand_builtin_with_bounds): Likewise.
16852         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
16853         (DEF_LIB_BUILTIN_CHKP): Likewise.
16854         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
16855         (DEF_CHKP_BUILTIN): Likewise.
16856         (BUILT_IN_MEMCPY): Likewise.
16857         (BUILT_IN_MEMMOVE): Likewise.
16858         (BUILT_IN_MEMPCPY): Likewise.
16859         (BUILT_IN_MEMSET): Likewise.
16860         (BUILT_IN_STPCPY): Likewise.
16861         (BUILT_IN_STRCAT): Likewise.
16862         (BUILT_IN_STRCHR): Likewise.
16863         (BUILT_IN_STRCPY): Likewise.
16864         (BUILT_IN_STRLEN): Likewise.
16865         (BUILT_IN_MEMCPY_CHK): Likewise.
16866         (BUILT_IN_MEMMOVE_CHK): Likewise.
16867         (BUILT_IN_MEMPCPY_CHK): Likewise.
16868         (BUILT_IN_MEMSET_CHK): Likewise.
16869         (BUILT_IN_STPCPY_CHK): Likewise.
16870         (BUILT_IN_STRCAT_CHK): Likewise.
16871         (BUILT_IN_STRCPY_CHK): Likewise.
16872         * calls.c (store_bounds): Likewise.
16873         (emit_call_1): Likewise.
16874         (special_function_p): Likewise.
16875         (maybe_warn_nonstring_arg): Likewise.
16876         (initialize_argument_information): Likewise.
16877         (finalize_must_preallocate): Likewise.
16878         (compute_argument_addresses): Likewise.
16879         (expand_call): Likewise.
16880         * cfgexpand.c (expand_call_stmt): Likewise.
16881         (expand_return): Likewise.
16882         (expand_gimple_stmt_1): Likewise.
16883         (pass_expand::execute): Likewise.
16884         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
16885         (cgraph_node::remove): Likewise.
16886         (cgraph_node::dump): Likewise.
16887         (cgraph_node::verify_node): Likewise.
16888         * cgraph.h (chkp_function_instrumented_p): Likewise.
16889         (symtab_node::get_alias_target): Likewise.
16890         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
16891         (cgraph_local_p): Likewise.
16892         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
16893         (cgraph_edge::rebuild_references): Likewise.
16894         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
16895         (walk_polymorphic_call_targets): Likewise.
16896         (cgraph_node::expand_thunk): Likewise.
16897         (symbol_table::output_weakrefs): Likewise.
16898         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
16899         (ix86_handle_option): Likewise.
16900         * config/i386/constraints.md: Likewise.
16901         * config/i386/i386-builtin-types.def (BND): Likewise.
16902         (VOID): Likewise.
16903         (PVOID): Likewise.
16904         (ULONG): Likewise.
16905         * config/i386/i386-builtin.def (BDESC_END): Likewise.
16906         (BDESC_FIRST): Likewise.
16907         (BDESC): Likewise.
16908         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
16909         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
16910         * config/i386/i386.c (enum reg_class): Likewise.
16911         (ix86_target_string): Likewise.
16912         (ix86_option_override_internal): Likewise.
16913         (ix86_conditional_register_usage): Likewise.
16914         (ix86_valid_target_attribute_inner_p): Likewise.
16915         (ix86_set_indirect_branch_type): Likewise.
16916         (ix86_set_current_function): Likewise.
16917         (ix86_function_arg_regno_p): Likewise.
16918         (init_cumulative_args): Likewise.
16919         (ix86_function_arg_advance): Likewise.
16920         (ix86_function_arg): Likewise.
16921         (ix86_pass_by_reference): Likewise.
16922         (ix86_function_value_regno_p): Likewise.
16923         (ix86_function_value_1): Likewise.
16924         (ix86_function_value_bounds): Likewise.
16925         (ix86_return_in_memory): Likewise.
16926         (ix86_setup_incoming_vararg_bounds): Likewise.
16927         (ix86_va_start): Likewise.
16928         (indirect_thunk_need_prefix): Likewise.
16929         (print_reg): Likewise.
16930         (ix86_print_operand): Likewise.
16931         (ix86_expand_call): Likewise.
16932         (ix86_output_function_return): Likewise.
16933         (reg_encoded_number): Likewise.
16934         (BDESC_VERIFYS): Likewise.
16935         (ix86_init_mpx_builtins): Likewise.
16936         (ix86_init_builtins): Likewise.
16937         (ix86_emit_cmove): Likewise.
16938         (ix86_emit_move_max): Likewise.
16939         (ix86_expand_builtin): Likewise.
16940         (ix86_builtin_mpx_function): Likewise.
16941         (ix86_get_arg_address_for_bt): Likewise.
16942         (ix86_load_bounds): Likewise.
16943         (ix86_store_bounds): Likewise.
16944         (ix86_load_returned_bounds): Likewise.
16945         (ix86_store_returned_bounds): Likewise.
16946         (ix86_class_likely_spilled_p): Likewise.
16947         (ix86_hard_regno_mode_ok): Likewise.
16948         (x86_order_regs_for_local_alloc): Likewise.
16949         (ix86_mitigate_rop): Likewise.
16950         (ix86_bnd_prefixed_insn_p): Likewise.
16951         (ix86_mpx_bound_mode): Likewise.
16952         (ix86_make_bounds_constant): Likewise.
16953         (ix86_initialize_bounds): Likewise.
16954         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
16955         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
16956         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
16957         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
16958         (TARGET_CHKP_BOUND_MODE): Likewise.
16959         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
16960         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
16961         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
16962         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
16963         * config/i386/i386.h (TARGET_MPX): Likewise.
16964         (TARGET_MPX_P): Likewise.
16965         (VALID_BND_REG_MODE): Likewise.
16966         (FIRST_BND_REG): Likewise.
16967         (LAST_BND_REG): Likewise.
16968         (enum reg_class): Likewise.
16969         (BND_REG_P): Likewise.
16970         (BND_REGNO_P): Likewise.
16971         (BNDmode): Likewise.
16972         (ADJUST_INSN_LENGTH): Likewise.
16973         * config/i386/i386.md: Likewise.
16974         * config/i386/i386.opt: Likewise.
16975         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
16976         (defined): Likewise.
16977         (LINK_MPX): Likewise.
16978         (MPX_SPEC): Likewise.
16979         (LIBMPX_SPEC): Likewise.
16980         (LIBMPXWRAPPERS_SPEC): Likewise.
16981         (CHKP_SPEC): Likewise.
16982         * config/i386/predicates.md: Likewise.
16983         * dbxout.c (dbxout_type): Likewise.
16984         * doc/extend.texi: Likewise.
16985         * doc/invoke.texi: Likewise.
16986         * doc/md.texi: Likewise.
16987         * doc/tm.texi: Likewise.
16988         * doc/tm.texi.in: Likewise.
16989         * dwarf2out.c (is_base_type): Likewise.
16990         (gen_formal_types_die): Likewise.
16991         (gen_subprogram_die): Likewise.
16992         (gen_type_die_with_usage): Likewise.
16993         (gen_decl_die): Likewise.
16994         (dwarf2out_late_global_decl): Likewise.
16995         * expr.c (expand_assignment): Likewise.
16996         (emit_storent_insn): Likewise.
16997         (store_expr_with_bounds): Likewise.
16998         (store_expr): Likewise.
16999         (expand_expr_real_1): Likewise.
17000         * expr.h (store_expr_with_bounds): Likewise.
17001         * function.c (use_register_for_decl): Likewise.
17002         (struct bounds_parm_data): Likewise.
17003         (assign_parms_augmented_arg_list): Likewise.
17004         (assign_parm_find_entry_rtl): Likewise.
17005         (assign_parm_is_stack_parm): Likewise.
17006         (assign_parm_load_bounds): Likewise.
17007         (assign_bounds): Likewise.
17008         (assign_parms): Likewise.
17009         (expand_function_start): Likewise.
17010         * gcc.c (CHKP_SPEC): Likewise.
17011         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
17012         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
17013         (wrestrict_dom_walker::check_call): Likewise.
17014         * gimple.c (gimple_build_call_from_tree): Likewise.
17015         * gimple.h (enum gf_mask): Likewise.
17016         (gimple_call_with_bounds_p): Likewise.
17017         (gimple_call_set_with_bounds): Likewise.
17018         * gimplify.c (gimplify_init_constructor): Likewise.
17019         * ipa-cp.c (initialize_node_lattices): Likewise.
17020         (propagate_constants_across_call): Likewise.
17021         (find_more_scalar_values_for_callers_subset): Likewise.
17022         * ipa-hsa.c (process_hsa_functions): Likewise.
17023         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
17024         * ipa-icf.c (sem_function::merge): Likewise.
17025         * ipa-inline.c (early_inliner): Likewise.
17026         * ipa-pure-const.c (warn_function_noreturn): Likewise.
17027         (warn_function_cold): Likewise.
17028         (propagate_pure_const): Likewise.
17029         * ipa-ref.h (enum GTY): Likewise.
17030         * ipa-split.c (find_retbnd): Likewise.
17031         (consider_split): Likewise.
17032         (split_function): Likewise.
17033         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
17034         * ipa.c (walk_polymorphic_call_targets): Likewise.
17035         (symbol_table::remove_unreachable_nodes): Likewise.
17036         (process_references): Likewise.
17037         (cgraph_build_static_cdtor_1): Likewise.
17038         * lto-cgraph.c (lto_output_node): Likewise.
17039         (output_refs): Likewise.
17040         (compute_ltrans_boundary): Likewise.
17041         (input_overwrite_node): Likewise.
17042         (input_node): Likewise.
17043         (input_cgraph_1): Likewise.
17044         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
17045         * passes.c (pass_manager::execute_early_local_passes): Likewise.
17046         (class pass_chkp_instrumentation_passes): Likewise.
17047         (make_pass_chkp_instrumentation_passes): Likewise.
17048         * passes.def: Likewise.
17049         * rtl.h (struct GTY): Likewise.
17050         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
17051         * stor-layout.c (layout_type): Likewise.
17052         * symtab.c: Likewise.
17053         * target.def: Likewise.
17054         * targhooks.c (default_chkp_bound_type): Likewise.
17055         (default_chkp_bound_mode): Likewise.
17056         (default_builtin_chkp_function): Likewise.
17057         (default_chkp_function_value_bounds): Likewise.
17058         (default_chkp_make_bounds_constant): Likewise.
17059         (default_chkp_initialize_bounds): Likewise.
17060         * targhooks.h (default_chkp_bound_type): Likewise.
17061         (default_chkp_bound_mode): Likewise.
17062         (default_builtin_chkp_function): Likewise.
17063         (default_chkp_function_value_bounds): Likewise.
17064         (default_chkp_make_bounds_constant): Likewise.
17065         (default_chkp_initialize_bounds): Likewise.
17066         * toplev.c (compile_file): Likewise.
17067         (process_options): Likewise.
17068         * tree-core.h (DEF_BUILTIN): Likewise.
17069         (DEF_BUILTIN_CHKP): Likewise.
17070         * tree-inline.c (declare_return_variable): Likewise.
17071         (remap_gimple_stmt): Likewise.
17072         (copy_bb): Likewise.
17073         (initialize_inlined_parameters): Likewise.
17074         (expand_call_inline): Likewise.
17075         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
17076         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
17077         (make_pass_ipa_chkp_produce_thunks): Likewise.
17078         (make_pass_chkp): Likewise.
17079         (make_pass_chkp_opt): Likewise.
17080         (make_pass_chkp_instrumentation_passes): Likewise.
17081         * tree-pretty-print.c (dump_generic_node): Likewise.
17082         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
17083         * tree-ssa-dce.c (propagate_necessity): Likewise.
17084         (eliminate_unnecessary_stmts): Likewise.
17085         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
17086         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
17087         * tree-ssa-sccvn.h: Likewise.
17088         * tree-ssa-strlen.c (get_string_length): Likewise.
17089         (valid_builtin_call): Likewise.
17090         (adjust_last_stmt): Likewise.
17091         (handle_builtin_strchr): Likewise.
17092         (handle_builtin_strcpy): Likewise.
17093         (handle_builtin_stxncpy): Likewise.
17094         (handle_builtin_memcpy): Likewise.
17095         (handle_builtin_strcat): Likewise.
17096         (strlen_check_and_optimize_stmt): Likewise.
17097         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
17098         * tree-streamer-in.c: Likewise.
17099         * tree-streamer.c (record_common_node): Likewise.
17100         * tree.c (tree_code_size): Likewise.
17101         (wide_int_to_tree_1): Likewise.
17102         (type_contains_placeholder_1): Likewise.
17103         (build_common_tree_nodes): Likewise.
17104         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
17105         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
17106         (POINTER_BOUNDS_P): Likewise.
17107         (BOUNDED_TYPE_P): Likewise.
17108         (BOUNDED_P): Likewise.
17109         (CALL_WITH_BOUNDS_P): Likewise.
17110         (pointer_bounds_type_node): Likewise.
17111         * value-prof.c (gimple_ic): Likewise.
17112         * var-tracking.c (vt_add_function_parameters): Likewise.
17113         * varasm.c (make_decl_rtl): Likewise.
17114         (assemble_start_function): Likewise.
17115         (output_constant): Likewise.
17116         (maybe_assemble_visibility): Likewise.
17117         * varpool.c (ctor_for_folding): Likewise.
17118         * chkp-builtins.def: Remove.
17119         * ipa-chkp.c: Remove.
17120         * ipa-chkp.h: Remove.
17121         * rtl-chkp.c: Remove.
17122         * rtl-chkp.h: Remove.
17123         * tree-chkp-opt.c: Remove.
17124         * tree-chkp.c: Remove.
17125         * tree-chkp.h: Remove.
17127 2018-06-07  Carl Love  <cel@us.ibm.com>
17129         * config/rs6000/vsx.md (vextract_fp_from_shorth,
17130         vextract_fp_from_shortl): Add BE support.
17132 2018-06-07  Paul Koning  <ni1d@arrl.net>
17134         * compare-elim.c (try_merge_compare): Don't merge compare if
17135         address contains a side effect.
17136         (try_eliminate_compare): Likewise.
17138 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
17140         * config.gcc: Support "tremont".
17141         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
17142         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17143         PROCESSOR_TREMONT.
17144         * config/i386/i386.c (m_TREMONT): Define.
17145         (processor_target_table): Add "tremont".
17146         (PTA_TREMONT): Define.
17147         (ix86_lea_outperforms): Add TARGET_TREMONT.
17148         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
17149         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
17150         and M_INTEL_GOLDMONT_PLUS.
17151         (fold_builtin_cpu): Add "tremont".
17152         (ix86_add_stmt_cost): Add TARGET_TREMONT.
17153         (ix86_option_override_internal): Add "tremont".
17154         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
17155         (processor_type): Add PROCESSOR_TREMONT.
17156         * config/i386/x86-tune.def: Add m_TREMONT.
17157         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
17159 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17161         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
17162         symbol defined for msp430i* devices to be lower case.
17164 2018-06-07  Richard Biener  <rguenther@suse.de>
17166         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
17167         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
17168         Properly wrap signed arithmetic if overflow wraps.
17170 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
17172         PR tree-optimization/69615
17173         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
17174         of a cast from a same precision integral SSA_NAME in a bb dominated
17175         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
17176         cast to utype if rhs2 has already a compatible type.
17178 2018-06-07  Richard Biener  <rguenther@suse.de>
17180         PR tree-optimization/85935
17181         * graphite-scop-detection.c (find_params_in_bb): Analyze
17182         condition operands with respect to the correct loop.  Assert
17183         the analysis doesn't fail.
17185 2018-06-04  Carl Love  <cel@us.ibm.com>
17187         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
17188         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
17189         as it is slightly cheaper.
17190         (first_match_or_eos_index_<mode>):
17191         Calculate index using natural element order.
17192         (first_match_index_<mode>):
17193         Calculate index using natural element order.
17194         (first_match_or_eos_index_<mode>):
17195         Calculate index using natural order.
17196         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
17197         for BE and LE modes.
17198         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
17199         P9V_BUILTIN_VCLZLSBB_V16QI.
17200         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
17201         specific.
17203 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17205         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
17206         indentation and line wrap for many prototypes.  Add missing
17207         @smallexample directives around block of prototypes for vec_xl and
17208         vec_xst.
17210 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
17212         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
17213         track if we pass or return IEEE 128-bit floating point.
17214         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
17215         C++ mangling that is compatible with GCC 8.1.
17216         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
17217         (init_cumulative_args): Note if we pass or return IEEE 128-bit
17218         floating point types.
17219         (rs6000_function_arg_advance_1): Likewise.
17220         (rs6000_mangle_type): Optionally generate mangled names that match
17221         what GCC 8.1 generated for IEEE 128-bit floating point types.
17222         (rs6000_globalize_decl_name): If we have an external function that
17223         passes or returns IEEE 128-bit types, generate a weak reference
17224         from the mangled name used in GCC 8.1 to the current mangled
17225         name.
17226         (rs6000_init_builtins): Make __ibm128 use the long double type if
17227         long double is IBM extended double.  Make __float128 use the long
17228         double type if long double is IEEE 128-bit.
17230         PR target/85657
17231         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
17232         macro for __ibm128 built-in functions.
17233         (PACK_IF): Add __ibm128 pack/unpack functions.
17234         (UNPACK_IF): Likewise.
17235         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
17236         enable long double built-in functions if long double is IEEE
17237         128-bit floating point.
17238         (rs6000_invalid_builtin): Update long double built-in function
17239         error message.
17240         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
17241         functions, adjust the built-in function to use the long double
17242         built-in function if __ibm128 and long double are the same type.
17243         * doc/extend.texi (PowerPC builtins): Update documention for
17244         __builtin_{,un}pack_longdouble.  Add documentation for
17245         __builtin_{,un}pack_ibm128.
17247 2018-06-06  Jim Wilson  <jimw@sifive.com>
17249         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
17250         (struct machine_function): New field interrupt_mode.
17251         (riscv_handle_type_attribute): New function.  Add forward declaration.
17252         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
17253         (riscv_expand_epilogue): Check interrupt_mode field.
17254         (riscv_set_current_function): Check interrupt attribute args and
17255         set interrupt_mode field.
17256         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
17257         (riscv_sret, riscv_uret): New.
17258         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
17259         new arguments to interrupt attribute.
17261 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
17263         PR target/63177
17264         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
17265         Don't handle -mcpu=power8 if -mpower9-vector is also used.
17267 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17269         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
17270         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
17271         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17272         several redundant entries.
17274 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
17276         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
17277         type from "rtx" to "rtx_insn *".
17278         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
17279         for local "call_insn", removing cast.
17280         (ix86_expand_call): Likewise, introducing a "call_insn" local.
17282 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
17284         PR tree-optimization/86066
17285         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
17286         for BIT_INSERT_EXPR stores.
17288 2018-06-06  Richard Biener  <rguenther@suse.de>
17290         PR tree-optimization/86062
17291         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
17292         component refs ontop
17293         of to be offsetted base.
17295 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17297         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
17298         to be static and remove check on interrupt attribute name.
17300 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17302         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17303         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
17305 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
17307         PR target/79924
17308         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
17309         second argument.
17310         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
17311         Remove second argument, change how error is called.
17312         (aarch64_layout_arg): Remove second argument from
17313         aarch64_err_no_fpadvsimd call.
17314         (aarch64_init_cumulative_args): Ditto.
17315         (aarch64_gimplify_va_arg_expr): Ditto.
17316         * config/aarch64/aarch64.md (mov<mode>): Ditto.
17318 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
17320         * config/i386/i386.md (simple_return_indirect_internal): New expander.
17321         (*simple_return_indirect_internal<mode>): Rename from
17322         simple_return_indirect_internal.  Use W mode iterator.
17323         (rstorssp): New expander.
17324         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
17325         (clrssbsy): New expander.
17326         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
17328 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17330         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
17331         __typeof__.
17332         (cmse_check_pointed_object): Likewise.
17334 2018-06-05  Martin Liska  <mliska@suse.cz>
17336         PR gcov-profile/47618
17337         * doc/invoke.texi: Document how -fprofile-dir format
17338         is extended.
17340 2018-06-05  Richard Biener  <rguenther@suse.de>
17342         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
17343         removal pretend DOM info isn't available so we do not update
17344         it and only remove edges, not dominated blocks.  Actually free
17345         DOM info in case we removed something.  Remove unreachable blocks.
17346         (mfb_keep_latches): Work with either DOM info or marked backedges.
17347         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
17348         first.  Mark backedges if DOM info isn't available.
17349         (Re-)compute DOM info after cleanup_control_flow_pre.
17351 2018-06-05  Richard Biener  <rguenther@suse.de>
17353         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
17354         (locus_discrim_hasher::hash): Adjust.
17355         (locus_discrim_hasher::equal): Likewise.
17356         (next_discriminator_for_locus): Work on line directly.
17357         (same_line_p): Pass in expanded locus1 as well.
17358         (assign_discriminators): Avoid redundant location expansions.
17360 2018-06-05  Richard Biener  <rguenther@suse.de>
17362         PR tree-optimization/86046
17363         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
17364         if required after clearing TREE_ADDRESSABLE.
17366 2018-06-05  Richard Biener  <rguenther@suse.de>
17368         PR tree-optimization/86047
17369         * tree-ssa-loop.c (for_each_index): Glob handling of all
17370         decls and constants and really handle all of them.
17372 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17374         PR target/81497
17375         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
17376         qualifier_void_pointer and qualifier_const_void_pointer.
17377         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
17378         (arm_init_builtins): Handle the above.
17379         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
17380         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
17381         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
17382         void intrinsics.
17384 2018-06-05  Martin Liska  <mliska@suse.cz>
17386         * auto-profile.c (read_autofdo_file): Do not use
17387         gcov_ctr_summary struct.
17388         (afdo_callsite_hot_enough_for_early_inline): Likewise.
17389         * coverage.c (struct counts_entry): Likewise.
17390         (read_counts_file): Read just single summary entry.
17391         (get_coverage_counts): Use gcov_summary struct.
17392         * coverage.h (get_coverage_counts): Likewise.
17393         * gcov-dump.c (dump_working_sets): Likewise.
17394         (tag_summary): Dump just single summary.
17395         * gcov-io.c (gcov_write_summary): Write just histogram
17396         summary.
17397         (gcov_read_summary): Read just single summary.
17398         (compute_working_sets): Use gcov_summary struct.
17399         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
17400         of GCOV_COUNTERS_SUMMABLE.
17401         (GCOV_COUNTERS_SUMMABLE): Remove.
17402         (GCOV_FIRST_VALUE_COUNTER): Replace with
17403         GCOV_COUNTER_V_INTERVAL.
17404         (struct gcov_ctr_summary): Remove.
17405         (struct gcov_summary): Directly use fields of former
17406         gcov_ctr_summary.
17407         (compute_working_sets): Use gcov_summary struct.
17408         * gcov.c (read_count_file): Do not use ctrs fields.
17409         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
17410         struct.
17411         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
17412         struct.
17413         * profile.c: Likewise.
17414         * profile.h: Likewise.
17416 2018-06-05  Martin Liska  <mliska@suse.cz>
17418         PR gcov-profile/84846
17419         * gcov.c (output_lines): Print working directory only
17420         in intermediate format.
17422 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
17424         * config/s390/s390-builtin-types.def: Add void function type.
17425         * config/s390/s390-builtins.def: Use the function type for the
17426         tbeginc builtin.
17428 2018-06-04  Jim Wilson  <jimw@sifive.com>
17430         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
17431         to int.
17432         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
17433         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
17434         handle EH_RETURN_DATA_REGNO registers properly.
17435         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
17436         (riscv_expand_epilogue): Update comment.  Change argument name and
17437         type.  Update code to use new name and type.  Pass new args to
17438         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
17439         EXCEPTION_RETURN.
17440         * config/riscv/riscv.md (NORMAL_RETURN): New.
17441         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
17442         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
17443         (eh_return): Call gen_eh_return_internal and emit barrier.
17444         (eh_return_internal): Call riscv_expand_epilogue.
17446 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
17448         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
17449         bit_insertion field and declare can_be_merged_into method.
17450         (merged_store_group::can_be_merged_into): New method.
17451         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
17452         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
17453         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
17455 2018-06-04  Richard Biener  <rguenther@suse.de>
17457         PR tree-optimization/85955
17458         * builtins.c (fold_builtin_sincos): Convert pointers to
17459         destination to appropriate type before dereferencing.
17461 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
17463         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
17465 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
17467         * expr.c (expand_expr_real_1): Force the operand into memory if
17468         its TYPE_MODE is BLKmode and if there is no integer mode for
17469         the number of bits being extracted.
17471 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17473         PR target/85832
17474         PR target/86036
17475         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17476         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
17478 2018-06-04  Richard Biener  <rguenther@suse.de>
17480         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
17481         (cleanup_tree_cfg_noloop): ... single caller.  Do
17482         start_recording_case_labels later.
17484 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
17486         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
17487         to _IMMINTRIN_H_INCLUDED.
17488         * config/i386/pconfigintrin.h: Ditto.
17489         * config/i386/waitpkgintrin.h: Ditto.
17490         * config/i386/immintrin.h: Add includes for sgxintrin.h,
17491         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
17492         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
17493         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
17494         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
17495         waitpkgintrin.h and cldemoteintrin.h.
17497 2018-06-04  Richard Biener  <rguenther@suse.de>
17499         PR tree-optimization/86038
17500         * tracer.c (find_best_successor): Check probability for
17501         being initialized, bail out if not.
17503 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
17505         PR target/86003
17506         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
17507         of bits to ignore when comparing architectures.
17509 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17511         PR tree-optimization/69615
17512         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
17513         maximum or minimum of the type, try to merge it also as if
17514         range1 is + [-, x - 1] or + [x + 1, -].
17516         PR c++/86025
17517         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
17519 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
17521         PR tree-optimization/86034
17522         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
17523         the unsigned bitfield type in a bit insertion sequence if it does not
17524         have a larger precision than the bitfield size.
17525         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
17527 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
17529         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
17531 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17533         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
17534         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
17535         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
17536         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
17538 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17540         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
17541         Disable -fdelete-null-pointer-checks for ELF toolchain.
17543 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17544             Kito Cheng  <kito.cheng@gmail.com>
17546         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
17547         (nds32le-*-*, nds32be-*-*): Integrate checking process.
17548         (nds32*-*-*): Add glibc and uclibc conditions.
17549         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
17550         (TARGET_EXCEPT_UNWIND_INFO): Define.
17551         * config/nds32/elf.h: New file.
17552         * config/nds32/linux.h: New file.
17553         * config/nds32/nds32-elf.opt: New file.
17554         * config/nds32/nds32-linux.opt: New file.
17555         * config/nds32/nds32-fp-as-gp.c
17556         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
17557         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
17558         TARGET_LINUX_ABI.
17559         (nds32_asm_file_end): Ditto.
17560         (nds32_print_operand): Ditto.
17561         (nds32_insert_attributes): Ditto.
17562         (nds32_init_libfuncs): New function.
17563         (TARGET_HAVE_TLS): Define.
17564         (TARGET_INIT_LIBFUNCS): Define.
17565         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
17566         spec content.
17567         (TARGET_ELF): Apply different mcmodel setting.
17568         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
17569         been migrated into elf.h and linux.h files.
17570         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
17571         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
17572         (mcmodel): The content has been migrated into nds32-elf.opt and
17573         nds32-linux.opt files.
17574         * config/nds32/t-elf: New file.
17575         * config/nds32/t-linux: New file.
17577 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17578             Shiva Chen  <shiva0217@gmail.com>
17580         * config/nds32/constants.md (unspec_volatile_element): Add
17581         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
17582         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
17583         optimization.
17584         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
17585         (make_pass_nds32_fp_as_gp): Declare.
17586         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
17587         optmization pass.
17588         (nds32_asm_function_end_prologue): Remove unused asm output.
17589         (nds32_asm_function_begin_epilogue): Remove unused asm output.
17590         (nds32_asm_file_start): Output necessary fp_as_gp information.
17591         (nds32_option_override): Adjust register usage.
17592         (nds32_expand_prologue): Consider fp_as_gp situation.
17593         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
17594         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
17595         (epilogue): Ditto.
17596         (return): Ditto.
17597         (simple_return): Ditto.
17598         (omit_fp_begin): Output special directive for fp_as_gp.
17599         (omit_fp_end): Output special directive for fp_as_gp.
17600         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
17601         mforbid-fp-as-gp): New options.
17603 2018-06-01  Mark Wielaard  <mark@klomp.org>
17605         * dwarf2out.c (dwarf2out_finish): Remove generation of
17606         DW_AT_loclists_base.
17608 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
17610         * gimple-ssa-store-merging.c: Include gimple-fold.h.
17611         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
17612         (struct merged_store_group): Add bit_insertion field.
17613         (dump_char_array): Use standard hexadecimal format.
17614         (merged_store_group::merged_store_group): Set bit_insertion to false.
17615         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
17616         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
17617         also print the mask in the dump file.
17618         (pass_store_merging::gate): Minor tweak.
17619         (imm_store_chain_info::coalesce_immediate): Fix wrong association
17620         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
17621         stores with INTEGER_CST stores.
17622         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
17623         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
17624         and use it throughout.  Generate bit insertion sequences if need be.
17625         (pass_store_merging::process_store): Remove redundant condition.
17626         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
17628 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
17630         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
17631         the 128-bit floating point types.  Fix function comment.
17633 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17635         * config/aarch64/aarch64-simd.md
17636         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
17637         mnemonics.
17638         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
17639         mnemonics.
17641 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17643         PR tree-optimization/85989
17644         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
17645         variable.
17646         (backprop::intersect_uses): Check it when deciding whether this
17647         is a backedge reference.
17648         (backprop::process_block): Add each phi to m_visited_phis
17649         after visiting it, then clear it at the end.
17651 2018-06-01  Richard Biener  <rguenther@suse.de>
17653         * tree-vectorizer.h (vect_dr_stmt): New function.
17654         (vect_get_load_cost): Adjust.
17655         (vect_get_store_cost): Likewise.
17656         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
17657         Use vect_dr_stmt instead of DR_SMTT.
17658         (vect_record_base_alignments): Likewise.
17659         (vect_calculate_target_alignment): Likewise.
17660         (vect_compute_data_ref_alignment): Likewise and make static.
17661         (vect_update_misalignment_for_peel): Likewise.
17662         (vect_verify_datarefs_alignment): Likewise.
17663         (vector_alignment_reachable_p): Likewise.
17664         (vect_get_data_access_cost): Likewise.  Pass down
17665         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
17666         (vect_get_peeling_costs_all_drs): Likewise.
17667         (vect_peeling_hash_get_lowest_cost): Likewise.
17668         (vect_enhance_data_refs_alignment): Likewise.
17669         (vect_find_same_alignment_drs): Likewise.
17670         (vect_analyze_data_refs_alignment): Likewise.
17671         (vect_analyze_group_access_1): Likewise.
17672         (vect_analyze_group_access): Likewise.
17673         (vect_analyze_data_ref_access): Likewise.
17674         (vect_analyze_data_ref_accesses): Likewise.
17675         (vect_vfa_segment_size): Likewise.
17676         (vect_small_gap_p): Likewise.
17677         (vectorizable_with_step_bound_p): Likewise.
17678         (vect_prune_runtime_alias_test_list): Likewise.
17679         (vect_analyze_data_refs): Likewise.
17680         (vect_supportable_dr_alignment): Likewise.
17681         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
17682         (vect_gen_prolog_loop_niters): Likewise.
17683         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
17684         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
17685         modify DR_STMT.
17686         (vect_recog_mask_conversion_pattern): Likewise.
17687         (vect_try_gather_scatter_pattern): Likewise.
17688         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
17689         to vect_get_store_cost.
17690         (vect_get_store_cost): Get stmt_info instead of DR.
17691         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
17692         (vect_get_load_cost): Get stmt_info instead of DR.
17694 2018-06-01  Richard Biener  <rguenther@suse.de>
17696         PR middle-end/86017
17697         * gimple-fold.c (var_decl_component_p): Also allow offsetted
17698         vars wrapped in MEM_REFs.
17700 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17702         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17703         Fix subreg tests so that we only return a choice between
17704         GENERAL_REGS and FP_REGS if the original classes included both.
17706 2018-06-01  Richard Biener  <rguenther@suse.de>
17708         PR ipa/85960
17709         * tree-ssa-structalias.c (get_function_part_constraint):
17710         Handle NULL fi->decl.
17711         (find_func_aliases_for_call): Properly handle indirect
17712         fi from direct call.
17713         (find_func_clobbers): Likewise.
17714         (ipa_pta_execute): Likewise.
17715         (create_variable_info_for): For functions that are ifunc_resolver
17716         resolve to a varinfo that contains the result of the resolver call.
17717         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
17718         aliases.
17720 2018-05-31  Michael Collison  <michael.collison@arm.com>
17722         * config/aarch64/aarch64.md:
17723         (*fix_to_zero_extenddfdi2): New pattern.
17724         * gcc.target/aarch64/fix_extend1.c: New testcase.
17726 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
17728         PR middle-end/78809
17729         PR middle-end/83026
17730         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
17731         and BUILT_IN_STRNCMP_EQ.
17732         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
17733         BUILT_IN_STRNCMP_EQ.
17734         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
17735         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17736         (gimple_fold_builtin): Likewise.
17737         * tree-ssa-strlen.c (compute_string_length): New function.
17738         (determine_min_obsize): New function.
17739         (handle_builtin_string_cmp): New function to handle calls to
17740         string compare functions.
17741         (strlen_optimize_stmt): Add handling to builtin string compare
17742         calls.
17743         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
17744         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17745         * tree.c (build_common_builtin_nodes): Add new defines of
17746         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
17748 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
17750         PR target/85984
17751         * bb-reorder.c (pass_partition_blocks::gate): Return false for
17752         functions with naked attribute.
17754 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
17756         * config/i386/sse.md (avx_vec_concat<mode>):
17757         Substitute concat_tg_mode mode attribute with xtg_mode.
17758         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
17759         (concat_tg_mode): Remove mode attribute.
17761 2018-05-31  Martin Sebor  <msebor@redhat.com>
17763         PR c/82063
17764         * calls.c (alloc_max_size): Correct a logic error/typo.
17765         Treat excessive arguments as infinite.  Warn for invalid arguments.
17766         * doc/invoke.texi (-Walloc-size-larger-than): Update.
17768 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
17770         PR target/85829
17771         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
17772         and movx for Haswell.
17774 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
17775             Cesar Philippidis  <cesar@codesourcery.com>
17777         PR middle-end/85879
17778         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
17779         when emitting error on private/firstprivate reductions.
17780         * omp-low.c (lower_omp_target): Avoid reference-type processing
17781         on pointers for firstprivate clause.
17783 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17785         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
17786         (st1x2): Likewise.
17787         (st1x3): Likewise.
17788         * config/aarch64/aarch64-simd.md
17789         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
17790         (aarch64_ld1_x3_<mode>): Likewise
17791         (aarch64_st1x2<VALLDIF:mode>): Likewise
17792         (aarch64_st1_x2_<mode>): Likewise
17793         (aarch64_st1x3<VALLDIF:mode>): Likewise
17794         (aarch64_st1_x3_<mode>): Likewise
17795         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
17796         (vld1_s8_x3): Likewise.
17797         (vld1_u16_x3): Likewise.
17798         (vld1_s16_x3): Likewise.
17799         (vld1_u32_x3): Likewise.
17800         (vld1_s32_x3): Likewise.
17801         (vld1_u64_x3): Likewise.
17802         (vld1_s64_x3): Likewise.
17803         (vld1_f16_x3): Likewise.
17804         (vld1_f32_x3): Likewise.
17805         (vld1_f64_x3): Likewise.
17806         (vld1_p8_x3): Likewise.
17807         (vld1_p16_x3): Likewise.
17808         (vld1_p64_x3): Likewise.
17809         (vld1q_u8_x3): Likewise.
17810         (vld1q_s8_x3): Likewise.
17811         (vld1q_u16_x3): Likewise.
17812         (vld1q_s16_x3): Likewise.
17813         (vld1q_u32_x3): Likewise.
17814         (vld1q_s32_x3): Likewise.
17815         (vld1q_u64_x3): Likewise.
17816         (vld1q_s64_x3): Likewise.
17817         (vld1q_f16_x3): Likewise.
17818         (vld1q_f32_x3): Likewise.
17819         (vld1q_f64_x3): Likewise.
17820         (vld1q_p8_x3): Likewise.
17821         (vld1q_p16_x3): Likewise.
17822         (vld1q_p64_x3): Likewise.
17823         (vst1_s64_x2): Likewise.
17824         (vst1_u64_x2): Likewise.
17825         (vst1_f64_x2): Likewise.
17826         (vst1_s8_x2): Likewise.
17827         (vst1_p8_x2): Likewise.
17828         (vst1_s16_x2): Likewise.
17829         (vst1_p16_x2): Likewise.
17830         (vst1_s32_x2): Likewise.
17831         (vst1_u8_x2): Likewise.
17832         (vst1_u16_x2): Likewise.
17833         (vst1_u32_x2): Likewise.
17834         (vst1_f16_x2): Likewise.
17835         (vst1_f32_x2): Likewise.
17836         (vst1_p64_x2): Likewise.
17837         (vst1q_s8_x2): Likewise.
17838         (vst1q_p8_x2): Likewise.
17839         (vst1q_s16_x2): Likewise.
17840         (vst1q_p16_x2): Likewise.
17841         (vst1q_s32_x2): Likewise.
17842         (vst1q_s64_x2): Likewise.
17843         (vst1q_u8_x2): Likewise.
17844         (vst1q_u16_x2): Likewise.
17845         (vst1q_u32_x2): Likewise.
17846         (vst1q_u64_x2): Likewise.
17847         (vst1q_f16_x2): Likewise.
17848         (vst1q_f32_x2): Likewise.
17849         (vst1q_f64_x2): Likewise.
17850         (vst1q_p64_x2): Likewise.
17851         (vst1_s64_x3): Likewise.
17852         (vst1_u64_x3): Likewise.
17853         (vst1_f64_x3): Likewise.
17854         (vst1_s8_x3): Likewise.
17855         (vst1_p8_x3): Likewise.
17856         (vst1_s16_x3): Likewise.
17857         (vst1_p16_x3): Likewise.
17858         (vst1_s32_x3): Likewise.
17859         (vst1_u8_x3): Likewise.
17860         (vst1_u16_x3): Likewise.
17861         (vst1_u32_x3): Likewise.
17862         (vst1_f16_x3): Likewise.
17863         (vst1_f32_x3): Likewise.
17864         (vst1_p64_x3): Likewise.
17865         (vst1q_s8_x3): Likewise.
17866         (vst1q_p8_x3): Likewise.
17867         (vst1q_s16_x3): Likewise.
17868         (vst1q_p16_x3): Likewise.
17869         (vst1q_s32_x3): Likewise.
17870         (vst1q_s64_x3): Likewise.
17871         (vst1q_u8_x3): Likewise.
17872         (vst1q_u16_x3): Likewise.
17873         (vst1q_u32_x3): Likewise.
17874         (vst1q_u64_x3): Likewise.
17875         (vst1q_f16_x3): Likewise.
17876         (vst1q_f32_x3): Likewise.
17877         (vst1q_f64_x3): Likewise.
17878         (vst1q_p64_x3): Likewise.
17880 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17882         * config/msp430/msp430.c (msp430_output_labelref): Prepend
17883         user_label_prefix to name.
17885         * tree-core.h: Update comment about the format of NAME string
17886         passed to handler in attribute_spec.
17888         * config/msp430/msp430.md: Remove erroneous subreg expression from
17889         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
17890         zero_extend{q,h}isi2.
17892 2018-05-30  Borislav Petkov  <bp@suse.de>
17894         * doc/extend.texi: Document some architecture specific
17895         constraints and sort entries.
17897 2018-05-30  Martin Sebor  <msebor@redhat.com>
17899         PR middle-end/85369
17900         * builtins.c (expand_builtin_stpcpy_1): New function.
17901         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
17902         only if the former succeeds.
17904 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17906         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
17907         in saphira.
17909 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17911         * doc/invoke.texi (-flinker-output): Document
17913 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17915         * passes.c (ipa_write_summaries): Only modify statements if body
17916         is in memory.
17917         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
17918         incrementally linking.
17919         (ipa_passes): Likewise.
17920         * lto-cgraph.c (lto_output_node): When incrementally linking do not
17921         pass down resolution info.
17922         * common.opt (flag_incremental_link): Update info.
17923         * gcc.c (plugin specs): Turn flinker-output=* to
17924         -plugin-opt=-linker-output-known
17925         * toplev.c (compile_file): Also cut compilation when doing incremental
17926         link.
17927         * flag-types. (enum lto_partition_model): Add
17928         LTO_LINKER_OUTPUT_NOLTOREL.
17929         (invoke.texi): Add -flinker-output docs.
17930         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
17931         link same way as WPA; do not stream in dead initializers.
17933         * dwarf2out.c (dwarf2out_die_ref_for_decl,
17934         darf2out_register_external_decl): Support incremental link.
17936 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17938         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
17940 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17942         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
17943         it down to simple_object_copy_lto_debug_sections.
17944         (run_gcc): Determine incremental LTO link time and configure
17945         lto1 into non-wpa mode, disable renaming of debug sections.
17947 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17949         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17950         descriptions of various incorrectly documented functions.
17952 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17954         Revert:
17955         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17956         address check not strict.
17958 2018-05-30  Richard Biener  <rguenther@suse.de>
17960         PR tree-optimization/85964
17961         * tracer.c (better_p): Drop initialized count check, we only
17962         call the function with initialized counts now.
17963         (find_best_successor): Do find a best edge if one
17964         has uninitialized count.
17965         (find_best_predecessor): Likewise.  Do BB frequency check only
17966         if count is initialized.
17968 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
17970         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
17971         (aarch64_ldrstr_offset_compare): New.
17972         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
17973         load/store orderings.
17974         (aarch64_gen_adjusted_ldpstp): Likewise.
17976 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
17978         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17979         Check for subset of GENERAL_REGS and FP_REGS.
17980         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
17981         r=w alternative.
17983 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
17985         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
17986         and wi::to_poly_offset.  Add the current offset and then check
17987         whether the sum fits, rather than using an unchecked addition of
17988         a checked term.  Check for a shwi rather than a uhwi.
17989         * expr.c (get_bit_range): Use tree_to_poly_uint64.
17990         (store_constructor): Use poly_int_tree_p.
17991         (expand_expr_real_1): Likewise.
17992         * function.c (assign_temp): Likewise.
17993         * fold-const.c (const_binop): Use poly_int_tree_p and
17994         wi::to_poly_offset.
17995         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
17996         division.
17997         * ipa-icf-gimple.c (func_checker::compare_operand): Use
17998         to_poly_offset for MEM offsets.
17999         * ipa-icf.c (sem_variable::equals): Likewise.
18000         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
18001         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
18002         wi::to_poly_offset for BIT_FIELD_REF offsets.
18003         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
18004         wi::to_poly_offset.
18005         * var-tracking.c (emit_note_insn_var_location): Use
18006         tree_to_poly_uint64.
18008 2018-05-29  Jim Wilson  <jimw@sifive.com>
18010         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
18012 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
18014         PR target/85950
18015         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
18016         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
18017         sequence.
18018         (sse4_1_round<mode>2): Use nonimmediate_operand
18019         for operand 1 predicate.
18021 2018-05-29  Martin Sebor  <msebor@redhat.com>
18022             Richard Biener  <rguenther@suse.de>
18024         PR testsuite/85888
18025         * calls.c (get_size_range): Call determine_value_range instead
18026         of get_value_range..
18027         * tree-vrp.h (determine_value_range): Declared new function.
18028         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
18030 2018-05-29  Richard Biener  <rguenther@suse.de>
18032         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
18033         sure to use non-pattern stmts for get_earlier_stmt arguments.
18034         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
18035         called on pattern stmts.
18036         (get_later_stmt): Likewise.
18038 2018-05-29  Martin Liska  <mliska@suse.cz>
18040         PR gcov-profile/85759
18041         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
18042         env variables.
18044 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
18046         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
18047         VEC_UNPACK_*_EXPR.
18048         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
18049         VEC_PACK_*_EXPR.
18051         PR target/85918
18052         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
18053         VEC_PACK_FLOAT_EXPR): New tree codes.
18054         * tree-pretty-print.c (op_code_prio): Handle
18055         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
18056         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
18057         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
18058         * tree-inline.c (estimate_operator_cost): Likewise.
18059         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
18060         * fold-const.c (const_binop): Likewise.
18061         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
18062         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
18063         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
18064         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
18065         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
18066         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
18067         * expr.c (expand_expr_real_2): Likewise.
18068         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
18069         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
18070         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
18071         optabs.
18072         * optabs.c (expand_widen_pattern_expr): For
18073         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
18074         sign from result type rather than operand's type.
18075         (expand_binop_directly): For vec_packu_float_optab and
18076         vec_packs_float_optab allow result type to be different from operand's
18077         type.
18078         * optabs-tree.c (optab_for_tree_code): Handle
18079         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
18080         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
18081         * tree-vect-generic.c (expand_vector_operations_1):  Handle
18082         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
18083         VEC_PACK_FLOAT_EXPR.
18084         * tree-vect-stmts.c (supportable_widening_operation): Handle
18085         FIX_TRUNC_EXPR.
18086         (supportable_narrowing_operation): Handle FLOAT_EXPR.
18087         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
18088         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
18089         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
18090         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
18091         mode attributes.
18092         (vec_pack<floatprefix>_float_<mode>): New expander.
18093         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
18094         attributes.
18095         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
18096         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
18097         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
18098         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
18099         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
18100         Document.
18101         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
18102         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
18103         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
18104         VEC_PACK_FLOAT_EXPR): Document.
18106 2018-05-29  Richard Biener  <rguenther@suse.de>
18108         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
18109         member.
18110         (stmt_vec_info_vec): Make pointer.
18111         (init_stmt_vec_info_vec): Remove.
18112         (free_stmt_vec_info_vec): Likewise.
18113         (set_stmt_vec_info_vec): New function.
18114         (free_stmt_vec_infos): Likewise.
18115         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
18116         (set_vinfo_for_stmt): Likewise.
18117         (get_earlier_stmt): Likewise.
18118         (get_later_stmt): Likewise.
18119         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
18120         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
18121         (vec_info::~vec_info): Free stmt_vec_infos.
18122         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
18123         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
18124         (pass_slp_vectorize::execute): Likewise.
18125         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
18126         (free_stmt_vec_info_vec): Likewise.
18127         (set_stmt_vec_info_vec): New function.
18128         (free_stmt_vec_infos): Likewise.
18129         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
18130         the global stmt_vec_info_vec.
18131         * tree-parloops.c (gather_scalar_reductions): Use
18132         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
18133         vector.
18135 2018-05-29  Richard Biener  <rguenther@suse.de>
18137         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
18139 2018-05-29  Martin Liska  <mliska@suse.cz>
18140             David Malcolm  <dmalcolm@redhat.com>
18142         * vec.c (test_reverse): New.
18143         (vec_c_tests): Add new test.
18144         * vec.h (vl_ptr>::reverse): New function.
18146 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
18148         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
18150         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
18151         and later.
18153 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
18155         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
18157 2018-05-28  Richard Biener  <rguenther@suse.de>
18159         PR tree-optimization/85933
18160         * tree-vect-data-refs.c (vect_record_base_alignments): Only
18161         look at stmts marked as vectorizable.
18163 2018-05-28  Richard Biener  <rguenther@suse.de>
18165         PR tree-optimization/85934
18166         * tree-vect-generic.c (expand_vector_operations_1): Hoist
18167         vector boolean check before scalar optimization.
18169 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
18171         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
18172         for armv5te.
18174 2018-05-28  Mark Wielaard  <mark@klomp.org>
18176         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
18177         if it is an expression containing a minus sign.
18179 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
18181         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
18183 2018-05-27  Paul Koning  <ni1d@arrl.net>
18185         * config/pdp11/pdp11.md (truncsihi2): Remove.
18187 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
18188             Chung-Ju Wu  <jasonwucj@gmail.com>
18190         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
18191         implementation.
18192         (unaligned_store_dw): Ditto.
18193         * config/nds32/nds32-memory-manipulation.c
18194         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
18195         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
18196         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
18197         (emit_setmem_word_loop): Rename to ...
18198         (emit_setmem_doubleword_loop): ... this.
18199         (nds32_gen_dup_4_byte_to_word_value): New function.
18200         (nds32_gen_dup_8_byte_to_double_word_value): New function.
18201         (nds32_expand_setmem_loop): Refine implementation.
18202         (nds32_expand_setmem_loop_v3m): Ditto.
18203         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
18204         pattern.
18206 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18208         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
18210 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18212         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
18213         (nds32_init_machine_status): Initialize machine->attr_naked_p and
18214         machine->attr_no_prologue_p.
18215         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
18216         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
18217         (nds32_expand_epilogue): Consider attr_naked_p.
18218         (nds32_expand_epilogue_v3pop): Likewise.
18219         (nds32_can_use_return_insn): Likewise.
18220         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
18221         attr_no_prologue_p fields.
18222         * config/nds32/nds32.opt (mret-in-naked-func): New option.
18224 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
18226         PR target/85918
18227         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
18228         attributes.
18229         * config/i386/sse.md
18230         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
18231         Rename to ...
18232         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
18233         ... this.
18234         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
18235         Rename to ...
18236         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
18237         ... this.
18238         (*<floatsuffix>floatv2div2sf2): Rename to ...
18239         (*float<floatunssuffix>v2div2sf2): ... this.
18240         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
18241         (float<floatunssuffix>v2div2sf2_mask): ... this.
18242         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
18243         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
18244         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
18245         to ...
18246         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
18247         ... this.
18248         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18249         Rename to ...
18250         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18251         ... this.
18252         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18253         Rename to ...
18254         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18255         ... this.
18256         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
18257         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
18258         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
18259         gen_ufix_truncv8dfv8si2.
18260         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
18261         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
18262         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
18263         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
18264         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
18265         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
18266         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
18267         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
18269 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18271         PR target/85900
18272         PR target/85345
18273         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
18275 2018-05-25  Jim Wilson  <jimw@sifive.com>
18277         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
18278         * config/riscv/riscv.c (struct machine_function): Add
18279         interrupt_handler_p and attribute_checked_p fields.
18280         (riscv_attribute_table): Add interrupt.
18281         (riscv_interrupt_type_p): New.
18282         (riscv_save_reg_p): Save extra regs for interrupt handler.
18283         (riscv_use_save_libcall): Return false  for interrupt handler.
18284         (riscv_first_stack_step): Add forward declaration.
18285         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
18286         for interrupt handler with large frame.  Use it for saved reg list.
18287         (riscv_expand_prologue): Move flag_stack_usage_info support to
18288         eliminate duplication.
18289         (riscv_expand_epilogue): Generate mret for interrupt handler.
18290         (riscv_epilogue_uses): New.
18291         (riscv_can_use_return_insn): Return false for interrupt handler.
18292         (riscv_function_ok_for_sibcall): Likewise.
18293         (riscv_set_current_function): Add interrupt handler support.
18294         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
18295         * config/riscv/riscv.md (UNSPECV_MRET): New.
18296         (GP_REGNUM): New.
18297         (riscv_frflags, riscv_fsflags): Use tab after opcode.
18298         (riscv_mret): New.
18299         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
18301 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
18303         PR tree-optimization/85712
18304         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
18305         this candidate has already been replaced in-situ by a copy.
18307 2018-05-25  Jason Merrill  <jason@redhat.com>
18309         PR c++/80485 - inline function non-zero address.
18310         * symtab.c (nonzero_address): Check DECL_COMDAT.
18312 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
18314         PR target/83628
18315         * config/alpha/alpha.md (ashlsi3): New insn pattern.
18316         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
18317         extension of SImode operation.  Use const123_operand predicate.
18318         (*saddsi_1): Remove.
18319         (*saddl_se_1): Ditto.
18320         (*ssubsi_1): Ditto.
18321         (*ssubl_se_1): Ditto.
18322         * config/alpha/predicates.md (const123_operand): New predicate.
18323         * config/alpha/constraints.md (P): Use IN_RANGE.
18325 2018-05-25  Richard Biener  <rguenther@suse.de>
18327         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
18328         defaulted to true.
18329         (ref_maybe_used_by_stmt_p): Likewise.
18330         (stmt_may_clobber_ref_p): Likewise.
18331         (stmt_may_clobber_ref_p_1): Likewise.
18332         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
18333         and pass it along.
18334         (ref_maybe_used_by_stmt_p): Likewise.
18335         (stmt_may_clobber_ref_p): Likewise.
18336         (stmt_may_clobber_ref_p_1): Likewise.
18337         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
18338         the alias oracle to disambiguate DRs with stmts DR analysis
18339         couldn't handle.
18340         (vect_analyze_data_refs): Do not give up on not analyzable
18341         DRs for BB vectorization.  Remove code truncating the dataref
18342         vector.
18344 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
18346         PR target/85832
18347         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
18348         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
18349         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
18351 2018-05-25  Richard Biener  <rguenther@suse.de>
18353         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
18354         function, combining stmt data ref gathering and fatal analysis
18355         parts.
18356         (vect_analyze_data_refs): Remove now redudnant code and simplify.
18357         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
18358         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
18359         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
18360         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
18362 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18364         PR tree-optimization/85720
18365         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
18366         SCC if all partitions are builtins.
18367         (version_loop_by_alias_check): New parameter.  Generate cancelable
18368         runtime alias check if all partitions are builtins.
18369         (distribute_loop): Update call to above function.
18371 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18373         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
18374         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
18375         (parm_default_def_partition_arg): Ditto.
18376         (set_parm_default_def_partition): Ditto.
18377         (get_parm_default_def_partitions): Ditto and make it static.
18378         (get_undefined_value_partitions): Ditto and make it static.
18379         (remove_ssa_form): Refactor call to init_var_map here.
18380         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
18381         computation for loop region.
18382         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
18383         (register_default_def): Delete.
18384         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
18385         (parm_default_def_partition_arg): Ditto.
18386         (set_parm_default_def_partition): Ditto.
18387         (get_parm_default_def_partitions): Ditto and make it static.
18388         (get_undefined_value_partitions): Ditto and make it static.
18389         (coalesce_with_default, coalesce_with_default): Update comment.
18390         (create_coalesce_list_for_region): New func factored out from
18391         create_outofssa_var_map.
18392         (populate_coalesce_list_for_outofssa): New func factored out from
18393         create_outofssa_var_map and coalesce_ssa_name.
18394         (create_outofssa_var_map): Delete.
18395         (coalesce_ssa_name): Refactor to support live range computation.
18396         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
18397         (get_parm_default_def_partitions): Delete.
18398         (get_undefined_value_partitions): Ditto.
18399         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
18400         computation for loop region.
18401         (new_tree_live_info, loe_visit_block): Ditto.
18402         (live_worklist, set_var_live_on_entry): Ditto.
18403         (calculate_live_on_exit, verify_live_on_entry): Ditto.
18404         * tree-ssa-live.h (struct _var_map): New fields.
18405         (init_var_map): Change decl.
18406         (region_contains_p): New.
18408 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18410         * tree-ssa-live.h (live_merge_and_clear): Delete.
18412 2018-05-25  Richard Biener  <rguenther@suse.de>
18414         PR c++/85912
18415         * tree-dump.c (dequeue_and_dump): Remove access to removed
18416         operand 2 of a SWITCH_EXPR.
18418 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18420         * doc/sourcebuild.texi (vect_double_cond_arith): Include
18421         multiplication and division.
18422         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
18423         (cond_udiv@var{m}, cond_umod@var{m}): Document.
18424         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
18425         (cond_udiv_optab, cond_umod_optab): New optabs.
18426         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
18427         (IFN_COND_RDIV): New internal functions.
18428         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
18429         TRUNC_MOD_EXPR and RDIV_EXPR.
18430         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
18431         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
18432         New unspecs.
18433         (SVE_INT_BINARY): Include mult.
18434         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
18435         (optab, sve_int_op): Handle mult.
18436         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
18437         UNSPEC_COND_DIV.
18438         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
18439         for SVE_INT_BINARY_SD.
18441 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18443         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
18444         (optab, sve_int_op): Handle div and udiv.
18445         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
18446         for SVE_INT_BINARY_SD.
18447         (*<optab><mode>3): New insn for the same.
18449 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18451         * tree-vect-patterns.c: Include predict.h.
18452         (vect_recog_divmod_pattern): Restrict check for division support
18453         to when optimizing for size.
18455 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18457         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
18458         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
18459         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
18460         (gimple_match_op::set_op): Likewise.
18461         (gimple_resimplify4): Declare.
18462         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
18463         (expr::gen_transform): Likewise.
18464         (decision_tree::gen): Generate a simplification routine for 4 operands.
18465         * gimple-match-head.c (gimple_simplify): Add an overload for
18466         4 operands.  In the top-level function, handle up to 4 call
18467         arguments and call gimple_resimplify4.
18468         (gimple_resimplify4): New function.
18469         (build_call_internal): Pass a fourth operand.
18470         (maybe_push_to_seq): Likewise.
18471         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
18472         Fold VEC_COND_EXPRs of an operation and a default value into
18473         an IFN_COND_* function if possible.
18474         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
18475         New unspecs.
18476         (SVE_COND_FP_BINARY): Include them.
18477         (optab, sve_fp_op): Handle them.
18478         (SVE_INT_BINARY_REV): New code iterator.
18479         (SVE_COND_FP_BINARY_REV): New int iterator.
18480         (commutative): New int attribute.
18481         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
18482         Declare.
18483         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
18484         function.
18485         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
18486         (*cond_<optab><mode>): New patterns for reversed operands.
18488 2018-05-25  Richard Biener  <rguenther@suse.de>
18490         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
18491         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
18492         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
18493         (STMT_VINFO_GROUPED_ACCESS): Adjust.
18494         * tree-vect-data-refs.c (everywhere): Adjust users.
18495         * tree-vect-loop.c (everywhere): Likewise.
18496         * tree-vect-slp.c (everywhere): Likewise.
18497         * tree-vect-stmts.c (everywhere): Likewise.
18498         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
18500 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18502         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
18503         Rename to...
18504         (gcc_cv_as_section_exclude): ... this.
18505         Try Solaris as #exclude syntax.
18506         * configure: Regenerate.
18507         * config.in: Regenerate.
18508         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
18509         SECTION_EXCLUDE.
18510         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
18511         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
18513         * varasm.c (default_elf_asm_named_section): Don't check if
18514         HAVE_GAS_SECTION_EXCLUDE is defined.
18516 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18518         * doc/md.texi: Update the documentation of the cond_* optabs
18519         to mention the new final operand.  Fix GET_MODE_NUNITS call.
18520         Describe the scalar case too.
18521         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
18522         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
18523         instead of 2.
18524         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
18525         (get_conditional_internal_fn): Update comment.
18526         * tree-vect-loop.c (vectorizable_reduction): Pass the original
18527         accumulator value as a final argument to conditional functions.
18528         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
18529         a define_expand and add an "else" operand.  Assert for now that
18530         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
18531         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
18532         (*cond_<optab><mode>): New patterns.
18533         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
18534         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
18535         (UNSPEC_COND_EOR): Delete.
18536         (optab): Remove associated mappings.
18537         (SVE_INT_BINARY): New code iterator.
18538         (sve_int_op): Remove int attribute and add "minus" to the code
18539         attribute.
18540         (SVE_COND_INT_OP): Delete.
18541         (SVE_COND_FP_OP): Rename to...
18542         (SVE_COND_FP_BINARY): ...this.
18544 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18546         * optabs.c (can_reuse_operands_p): New function.
18547         (maybe_legitimize_operands): Try to reuse the results for
18548         earlier operands.
18550 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18552         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
18553         Add {q} suffix to insn mnemonic.
18555 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18557         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
18558         (msp430_warn_func_return): New.
18560 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
18562         * fold-const.c (tree_nonzero_bits): New function.
18563         * fold-const.h (tree_nonzero_bits): Likewise.
18564         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
18565         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
18567 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18569         PR target/85900
18570         PR target/85345
18571         * varasm.c (assemble_alias): Check ifunc_resolver only on
18572         FUNCTION_DECL.
18574 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18576         PR target/85903
18577         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
18578         when memory input operand is handled.
18580 2018-05-24  Luis Machado  <luis.machado@linaro.org>
18582         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
18583         global.
18584         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
18586 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18588         * match.pd: Delay FMA folds until after vectorization.
18590 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18592         PR target/83009
18593         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
18594         address check not strict.
18596 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18598         * gimple-match.h (gimple_match_op): New class.
18599         (mprts_hook): Replace parameters with a gimple_match_op *.
18600         (maybe_build_generic_op): Likewise.
18601         (gimple_simplified_result_is_gimple_val): Replace parameters with
18602         a const gimple_match_op *.
18603         (gimple_simplify): Replace code_helper * and tree * parameters with
18604         a gimple_match_op * parameter.
18605         (gimple_resimplify1): Replace code_helper *, tree and tree *
18606         parameters with a gimple_match_op * parameter.
18607         (gimple_resimplify2): Likewise.
18608         (gimple_resimplify3): Likewise.
18609         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
18610         parameters with a gimple_match_op * parameter.
18611         * gimple-match-head.c (gimple_simplify): Change prototypes of
18612         auto-generated functions to take a gimple_match_op * instead of
18613         separate code_helper * and tree * parameters.  Make the same
18614         change in the top-level overload and update calls to the
18615         gimple_resimplify routines.  Update calls to the auto-generated
18616         functions and to maybe_push_res_to_seq in the publicly-facing
18617         operation-specific gimple_simplify overloads.
18618         (gimple_match_op::MAX_NUM_OPS): Define.
18619         (gimple_resimplify1): Replace rcode and ops with a single res_op
18620         parameter.  Update call to gimple_simplify.
18621         (gimple_resimplify2): Likewise.
18622         (gimple_resimplify3): Likewise.
18623         (mprts_hook): Replace parameters with a gimple_match_op *.
18624         (maybe_build_generic_op): Likewise.
18625         (build_call_internal): Replace type, nargs and ops with
18626         a gimple_match_op *.
18627         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
18628         with a single gimple_match_op *.  Update calls to mprts_hook,
18629         build_call_internal and gimple_simplified_result_is_gimple_val.
18630         Factor out code that is common to the tree_code and combined_fn cases.
18631         * genmatch.c (expr::gen_transform): Replace tem_code and
18632         tem_ops with a gimple_match_op called tem_op.  Update calls
18633         to the gimple_resimplify functions and maybe_push_res_to_seq.
18634         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
18635         res_ops.  Update call to the gimple_resimplify functions.
18636         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
18637         (decision_tree::gen): Make the functions take a gimple_match_op *
18638         called res_op instead of separate res_code and res_ops parameters.
18639         Update call accordingly.
18640         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
18641         and ops with a single res_op parameter.  Update calls to
18642         maybe_build_generic_op and maybe_push_res_to_seq.
18643         (fold_stmt_1): Update calls to gimple_simplify and
18644         replace_stmt_with_simplification.
18645         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
18646         and gimple_simplified_result_is_gimple_val.
18647         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
18648         gimple_simplify.
18649         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
18650         with a gimple_match_op *.
18651         (vn_nary_build_or_lookup): Likewise.  Update call to
18652         vn_nary_build_or_lookup_1.
18653         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
18654         gimple_match_op *.  Update calls to the gimple_resimplify routines
18655         and to gimple_simplified_result_is_gimple_val.
18656         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
18657         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
18658         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
18659         (visit_nary_op): Likewise.
18660         (visit_reference_op_load): Likewise.
18662 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18664         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
18665         modifier for printing the step amount.
18667 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
18669         PR target/78849
18670         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
18671         types.
18673 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
18675         * doc/sourcebuild.texi (Endianness): New subsubsection.
18677 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18679         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18680         <prefetch_dynamic_strides>: New const bool field.
18681         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18682         prefetch_dynamic_strides.
18683         (exynosm1_prefetch_tune): Likewise.
18684         (thunderxt88_prefetch_tune): Likewise.
18685         (thunderx_prefetch_tune): Likewise.
18686         (thunderx2t99_prefetch_tune): Likewise.
18687         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
18688         false.
18689         (aarch64_override_options_internal): Update to set
18690         PARAM_PREFETCH_DYNAMIC_STRIDES.
18691         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18692         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18693         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18694         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18695         prefetch-dynamic-strides setting.
18697 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18699         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18700         <minimum_stride>: New const int field.
18701         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18702         minimum_stride field defaulting to -1.
18703         (exynosm1_prefetch_tune): Likewise.
18704         (thunderxt88_prefetch_tune): Likewise.
18705         (thunderx_prefetch_tune): Likewise.
18706         (thunderx2t99_prefetch_tune): Likewise.
18707         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
18708         <default_opt_level>: Set to 3.
18709         (aarch64_override_options_internal): Update to set
18710         PARAM_PREFETCH_MINIMUM_STRIDE.
18711         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18712         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18713         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18714         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18715         stride is constant and is below the minimum stride threshold.
18717 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18719         * config/arm/arm-cpus.in (mode26): Delete.
18720         (armv4): Delete mode26 reference.
18721         * config/arm/arm.c (arm_configure_build_target): Delete use of
18722         isa_bit_mode26.
18724 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18726         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
18727         New insn pattern.
18728         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
18729         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
18730         for non-SSE modes.
18731         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
18732         (floatunsdidf2): Ditto.
18734 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18736         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
18737         (fixuns_trunc<mode>si2_avx512f): Ditto.
18738         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
18739         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
18740         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
18742 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18744         PR rtl-optimization/79985
18745         * df-scan.c (df_insn_refs_collect): Remove special case for
18746         global registers and asm statements.
18748 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18750         * extend.texi (Global Register Variables): Rewrite the bullet list.
18751         Note that the register is available for allocation. Note that access
18752         via inline asm must use constraints. Add note about async-signal
18753         handlers. Remove paragraph about automagic register selection.
18755 2018-05-23  Richard Biener  <rguenther@suse.de>
18757         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
18758         of fixed offset from memset VN.
18760 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
18762         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
18763         first_interp field.
18764         (alloc_cand_and_find_basis): Initialize first_interp field.
18765         (slsr_process_mul): Modify first_interp field.
18766         (slsr_process_add): Likewise.
18767         (slsr_process_cast): Modify first_interp field for each new
18768         interpretation.
18769         (slsr_process_copy): Likewise.
18770         (dump_candidate): Dump first_interp field.
18771         (replace_mult_candidate): Process all interpretations, not just
18772         subsequent ones.
18773         (replace_rhs_if_not_dup): Likewise.
18774         (replace_one_candidate): Likewise.
18776 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
18778         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
18779         Add new boolean.
18780         (aarch64_needs_frame_chain): New function.
18781         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
18783 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
18785         PR target/84882
18786         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
18787         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
18788         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
18789         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
18790         as true for strict-align.
18791         (aarch64_can_inline_p): Perform checks even when callee has no
18792         attributes to check for strict alignment.
18793         * doc/extend.texi (AArch64 Function Attributes): Document
18794         no-strict-align.
18795         * doc/invoke.texi: (AArch64 Options): Likewise.
18797 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
18799         PR tree-optimization/85853
18800         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
18801         the handling of the root of the node to...
18802         (vect_slp_analyze_node_operations_1): ...this new function,
18803         and run the whole thing with the child nodes' def types
18804         set according to their SLP node's def type.
18806 2018-05-23  Richard Biener  <rguenther@suse.de>
18808         PR middle-end/85874
18809         * tree-data-ref.c (create_runtime_alias_checks): Defer
18810         and ignore overflow warnings.
18812 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
18814         PR tree-optimization/85822
18815         * tree-vrp.c (is_masked_range_test): Fix handling of negative
18816         constants.
18818 2018-05-23  Richard Biener  <rguenther@suse.de>
18820         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
18821         memset constants via native_interpret_expr.
18823 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
18825         PR target/85345
18826         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
18827         attribute.
18828         (cgraph_node::create_alias): Likewise.
18829         (cgraph_node::get_availability): Check ifunc_resolver instead
18830         of looking up ifunc attribute.
18831         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
18832         * varasm.c (do_assemble_alias): Likewise.
18833         (assemble_alias): Likewise.
18834         (default_binds_local_p_3): Likewise.
18835         * cgraph.h (cgraph_node): Add ifunc_resolver.
18836         (cgraph_node::only_called_directly_or_aliased_p): Return false
18837         for IFUNC resolver.
18838         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
18839         attribute.
18840         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
18841         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
18842         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
18843         instead of looking up ifunc attribute.
18845 2018-05-22  Luis Machado  <luis.machado@linaro.org>
18847         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
18849 2018-05-22  Martin Sebor  <msebor@redhat.com>
18851         PR middle-end/85359
18852         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
18853         only when expasion succeeds.
18854         (expand_builtin_strcmp): Same.
18855         (expand_builtin_strncmp): Same.
18857 2018-05-22  Martin Sebor  <msebor@redhat.com>
18859         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
18861 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18862             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18864         * config/aarch64/aarch64-ldpstp.md: Replace uses of
18865         aarch64_mem_pair_operand with memory_operand and delete operand swapping
18866         code.
18867         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18868         Add check for legitimate_address.
18869         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
18870         (aarch64_swap_ldrstr_operands): New.
18871         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
18872         Define prototype.
18874 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18875             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18877         * config/aarch64/aarch64.md: New patterns to generate stp
18878         and ldp.
18879         (store_pair_sw, store_pair_dw): New patterns to generate stp for
18880         single words and double words.
18881         (load_pair_sw, load_pair_dw): Likewise.
18882         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
18883         Delete.
18884         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
18885         Delete.
18886         * config/aarch64/aarch64-ldpstp.md: Modify peephole
18887         for different mode ldpstp and add peephole for merged zero stores.
18888         Likewise for loads.
18889         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18890         Add size check.
18891         (aarch64_gen_store_pair): Rename calls to match new patterns.
18892         (aarch64_gen_load_pair): Rename calls to match new patterns.
18893         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
18894         (load_pair<DREG:mode><DREG2:mode>): ... This.
18895         (store_pair<mode>): Rename to...
18896         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
18897         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
18898         New mode iterators.
18899         (V_INT_EQUIV): Handle SImode.
18900         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
18901         New predicate.
18903 2018-05-22  Martin Sebor  <msebor@redhat.com>
18905         PR c/85623
18906         * calls.c (maybe_warn_nonstring_arg): Use string length to set
18907         or ajust the presumed bound on an operation to avoid unnecessary
18908         warnings.
18910 2018-05-22  Martin Sebor  <msebor@redhat.com>
18912         PR tree-optimization/85826
18913         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
18914         assuming that a DECL necesarily has a constant size.
18916 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
18918         PR middle-end/85862
18919         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
18921 2018-05-22  Richard Biener  <rguenther@suse.de>
18923         PR tree-optimization/85834
18924         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
18925         non-constant and non-zero memset arguments.
18927 2018-05-22  Martin Liska  <mliska@suse.cz>
18929         PR ipa/85607
18930         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
18932 2018-05-22  Richard Biener  <rguenther@suse.de>
18934         PR tree-optimization/85863
18935         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
18936         comparisons when vectype is specified.
18937         (vectorizable_condition): Do not specify vectype for
18938         vect_is_simple_cond when SLP vectorizing.
18940 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
18942         PR target/85657
18943         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
18944         define __ibm128 as long double.
18945         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
18946         as a distinct type when IEEE 128-bit support is enabled.
18947         (init_float128_ieee): Fix up conversions between IFmode and IEEE
18948         128-bit types to use the correct functions.
18949         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
18950         convert between 128-bit floating point types that have different
18951         modes but the same representation, instead of using gen_lowpart to
18952         makean alias.
18953         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
18954         KFmode.
18955         (IFKF_reg): New attributes to give the register constraints for
18956         IFmode and KFmode.
18957         (extend<mode>tf2_internal): New insns to mark an explicit
18958         conversion between 128-bit floating point types that have a
18959         different mode but share the same representation.
18961 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
18963         PR tree-optimization/85814
18964         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
18965         a null return from get_strinfo when unsharing the next
18966         strinfo in the chain.
18968 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
18970         PR gcc/84923
18971         * varasm.c (weak_finish): Clean up weak_decls.
18973 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18975         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
18976         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
18977         UNSPEC_UADALP values.
18978         * config/aarch64/iterators.md (ABAL): New int iterator.
18979         (ABDL2): Likewise.
18980         (ADALP): Likewise.
18981         (sur): Add mappings for the above.
18982         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
18983         New define_insn.
18984         (aarch64_<sur>abal<mode>_4): Likewise.
18985         (aarch64_<sur>adalp<mode>_3): Likewise.
18986         (<sur>sadv16qi): New define_expand.
18988 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
18990         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
18991         (*movdf_internal): Ditto.
18992         (*rcpsf2_sse): Ditto.
18993         (*rsqrtsf2_sse): Ditto.
18994         (*sqrt<mode>2_sse): Ditto.
18996 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
18998         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
18999         eor3q<mode>4.
19000         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
19001         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
19002         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
19003         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
19004         vbcaxq_s64): New.
19005         * config/aarch64/arm_neon.h: Likewise.
19006         * config/aarch64/iterators.md (VQ_I): New.
19008 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
19010         * config.gcc: Add arc/t-multilib-linux to tmake_file for
19011         arc*-*-linux*.
19012         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
19013         MULTILIB_DIRNAMES
19015 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
19017         * config/nds32/constraints.md (S): New constraint.
19018         * config/nds32/nds32.md (call_internal): Use constraint S.
19019         (call_value_internal): Likewise.
19020         (sibcall_internal): Likewise.
19021         (sibcall_value_internal): Likewise.
19023 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
19024             Chung-Ju Wu  <jasonwucj@gmail.com>
19026         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
19027         into consideration.
19029 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
19030             Chung-Ju Wu  <jasonwucj@gmail.com>
19032         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
19033         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
19034         (nds32_rtx_costs_impl): Simplify.
19035         (nds32_address_cost_impl): Simplify.
19036         (nds32_init_rtx_costs): New function.
19037         (nds32_rtx_costs_speed_prefer): Likewise.
19038         (nds32_rtx_costs_size_prefer): Likewise.
19039         (nds32_address_cost_speed_prefer): Likewise.
19040         (nds32_address_cost_speed_fwprop): Likewise.
19041         (nds32_address_cost_size_prefer): Likewise.
19042         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
19043         * config/nds32/nds32.c (nds32_option_override): Use
19044         nds32_init_rtx_costs function.
19046 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
19048         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
19049         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
19050         (TARGET_PIPELINE_N8): Likewise.
19051         (TARGET_PIPELINE_N10): Likewise.
19052         (TARGET_PIPELINE_N13): Likewise.
19053         (TARGET_PIPELINE_GRAYWOLF): Likewise.
19055 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
19057         * config/nds32/nds32-fpu.md: Update copyright year.
19059 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19061         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
19063 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19065         * config/nds32/nds32.c
19066         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
19067         * config/nds32/nds32.opt (minline-asm-r15): New option.
19069 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19071         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
19072         MASK_HW_ABS.
19073         * config/nds32/nds32.md (abssi2): New pattern.
19075 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
19077         * config/i386/i386.md (rex64namesuffix): New mode attribute.
19078         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
19079         Merge insn pattern from sse_cvtsi2ss<round_name> and
19080         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
19081         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
19082         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
19083         using SWI48 mode iterator.
19084         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
19085         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
19086         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
19087         pattern from sse_cvttss2si<round_saeonly_name>
19088         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
19089         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
19090         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
19091         using SWI48 mode iterator.
19092         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
19093         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
19094         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
19095         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
19096         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
19097         using SWI48 mode iterator.
19098         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
19099         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
19100         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
19101         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
19102         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
19103         SWI48 mode iterator.
19104         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
19105         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
19106         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
19107         pattern from sse_cvttsd2si<round_saeonly_name>
19108         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
19110 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19112         * config/nds32/nds32-md-auxiliary.c
19113         (nds32_valid_smw_lwm_base_p): Refine.
19114         (nds32_output_smw_single_word): Refine.
19115         (nds32_output_smw_double_word): New.
19116         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
19118 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19120         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
19121         (nds32_output_stack_pop): Refine.
19122         (nds32_expand_unaligned_load): Refine.
19123         (nds32_expand_unaligned_store): Refine.
19125 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
19126             Chung-Ju Wu  <jasonwucj@gmail.com>
19128         * config/nds32/constants.md: Add TP_REGNUM constant.
19129         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
19130         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
19131         UNSPEC_ADD32.
19132         * config/nds32/nds32-doubleword.md: Consider flag_pic.
19133         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
19134         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
19135         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
19136         and PIC code generation.
19137         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
19138         code generation.
19139         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
19140         optimization.
19141         * config/nds32/nds32.md: Support TLS and PIC.
19142         * config/nds32/nds32.c: Support TLS and PIC.
19143         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
19144         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
19145         predicate.
19147 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19149         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
19150         mode with E_ prefix.
19152 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
19153             Chung-Ju Wu  <jasonwucj@gmail.com>
19155         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
19156         * config/nds32/nds32-md-auxiliary.c
19157         (symbolic_reference_mentioned_p): New.
19158         (nds32_legitimize_ict_address): New.
19159         (nds32_expand_ict_move): New.
19160         (nds32_indirect_call_referenced_p): New.
19161         (nds32_symbol_binds_local_p): Delete.
19162         (nds32_long_call_p): Modify.
19163         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
19164         * config/nds32/nds32-protos.h
19165         (symbolic_reference_mentioned_p): Declare.
19166         (nds32_legitimize_ict_address): Declare.
19167         (nds32_expand_ict_move): Declare.
19168         (nds32_indirect_call_referenced_p): Declare.
19169         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
19170         (nds32_relax_group): Use nds32_ict_const_p as condition.
19171         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
19172         (nds32_asm_file_start): Output ict_model directive in asm code.
19173         (nds32_legitimate_address_p): Consider indirect call.
19174         (nds32_print_operand): Consider indirect call.
19175         (nds32_print_operand_address): Consider indirect call.
19176         (nds32_insert_attributes): Handle "indirect_call" attribute.
19177         (TARGET_LEGITIMATE_ADDRESS_P): Define.
19178         (TARGET_LEGITIMATE_CONSTANT_P): Define.
19179         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
19180         (TARGET_DELEGITIMIZE_ADDRESS): Define.
19181         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
19182         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
19183         (TARGET_ICT_MODEL_SMALL): Define.
19184         (TARGET_ICT_MODEL_LARGE): Define.
19185         * config/nds32/nds32.md (movsi): Consider ict model.
19186         (call, call_value): Consider ict model.
19187         (sibcall, sibcall_value): Consider ict model.
19188         * config/nds32/nds32.opt (mict-model): New option.
19189         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
19190         model.
19192 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
19193             Monk Chiang  <sh.chiang04@gmail.com>
19194             Jim Wilson <jimw@sifive.com>
19196         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
19197         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
19198         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
19199         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
19200         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
19201         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
19202         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
19203         compute save_libcall_adjustment properly.
19204         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
19205         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
19206         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
19207         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
19208         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
19209         (ABI_SPEC): Handle mabi=ilp32e.
19210         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
19211         (RVE): Add RVE mask.
19212         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
19213         <-march>: Add rv32e as an example.
19215 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
19217         PR c++/82899
19218         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
19219         (intra_create_variable_infos): Handle C++ constructors.
19221 2018-05-18  Martin Liska  <mliska@suse.cz>
19223         * passes.def: Remove a redundant pass.
19225 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
19227         PR bootstrap/85838
19228         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
19230 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19232         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
19233         (ARMv4): Update.
19234         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
19235         (ARMv6m): Update.
19236         (armv2, armv2a, armv3, armv3m): Delete architectures.
19237         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
19238         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
19239         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
19240         Delete cpus.
19241         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
19242         (*mulsidi3adddi): Likewise.
19243         (mulsidi3): Likewise.
19244         (*mulsidi3_nov6): Likewise.
19245         (umulsidi3): Likewise.
19246         (umulsidi3_nov6): Likewise.
19247         (umaddsidi4): Likewise.
19248         (*umulsidi3adddi): Likewise.
19249         (smulsi3_highpart): Likewise.
19250         (*smulsi3_highpart_nov6): Likewise.
19251         (umulsi3_highpart): Likewise.
19252         (*umulsi3_highpart_nov6): Likewise.
19253         * config/arm/arm.h (arm_arch3m): Delete.
19254         * config/arm/arm.c (arm_arch3m): Delete.
19255         (arm_option_override_internal): Update armv3-related comment.
19256         (arm_configure_build_target): Delete use of isa_bit_mode32.
19257         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
19258         (arm_rtx_costs_internal): Delete check of arm_arch3m.
19259         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
19260         (mulsa3): Likewise.
19261         (mulusa3): Likewise.
19262         * config/arm/arm-protos.h (arm_arch3m): Delete.
19263         * config/arm/arm-tables.opt: Regenerate.
19264         * config/arm/arm-tune.md: Likewise.
19265         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
19266         deleted architectures.
19268 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19270         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
19271         (armv5t, armv5te): New features.
19272         (ARMv5, ARMv5e): Delete fgroups.
19273         (ARMv5t, ARMv5te): Adjust for above changes.
19274         (ARMv6m): Likewise.
19275         (armv5, armv5e): Delete arches.
19276         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
19277         arm_arch5.
19278         (*call_reg_arm): Likewise.
19279         (*call_value_reg_armv5): Likewise.
19280         (*call_value_reg_arm): Likewise.
19281         (*call_symbol): Likewise.
19282         (*call_value_symbol): Likewise.
19283         (*sibcall_insn): Likewise.
19284         (*sibcall_value_insn): Likewise.
19285         (clzsi2): Likewise.
19286         (prefetch): Likewise.
19287         (define_split and define_peephole2 dependent on arm_arch5):
19288         Likewise.
19289         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
19290         arm_arch5e.
19291         (TARGET_ARM_QBIT): Likewise.
19292         (TARGET_DSP_MULTIPLY): Likewise.
19293         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
19294         (arm_arch5, arm_arch5e): Delete.
19295         (arm_arch5t, arm_arch5te): Declare.
19296         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
19297         (arm_arch5t): Declare.
19298         (arm_option_reconfigure_globals): Update for the above.
19299         (arm_options_perform_arch_sanity_checks): Update comment, replace
19300         use of arm_arch5 with arm_arch5t.
19301         (use_return_insn): Likewise.
19302         (arm_emit_call_insn): Likewise.
19303         (output_return_instruction): Likewise.
19304         (arm_final_prescan_insn): Likewise.
19305         (arm_coproc_builtin_available): Likewise.
19306         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
19307         arm_arch5e with arm_arch5t and arm_arch5te.
19308         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
19309         (arm_arch5t, arm_arch5te): Declare.
19310         * config/arm/arm-tables.opt: Regenerate.
19311         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
19312         * config/arm/t-multilib: Likewise.
19313         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
19314         instead of arm_arch5.
19315         (*call_reg_thumb1): Likewise.
19316         (*call_value_reg_thumb1_v5): Likewise.
19317         (*call_value_reg_thumb1): Likewise.
19318         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
19319         unreachable path.
19320         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
19322 2018-05-18  Martin Liska  <mliska@suse.cz>
19324         PR gcov-profile/84846
19325         * doc/gcov.texi: Document -t option of gcov tool.
19327 2018-05-18  Martin Liska  <mliska@suse.cz>
19329         PR gcov-profile/84846
19330         * gcov.c (print_usage): Add new -t option.
19331         (process_args): Handle the option.
19332         (generate_results): Use stdout as output when requested by
19333         the option.
19335 2018-05-18  Martin Liska  <mliska@suse.cz>
19337         PR gcov-profile/84846
19338         * coverage.c (coverage_init): Write PWD to .gcno file.
19339         * doc/gcov.texi: Document how working directory is printed.
19340         * gcov-dump.c (dump_gcov_file): Print PWD.
19341         * gcov.c (output_intermediate_file): Likewise.
19342         (read_graph_file): Read PWD string.
19343         (output_lines): Print PWD.
19345 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19347         PR middle-end/85817
19348         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
19349         for retval and return false if all args to phi are zero.
19351 2018-05-18  Richard Biener  <rguenther@suse.de>
19353         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
19354         method.
19355         (evrp_dom_walker::before_dom_children): Call it.
19357 2018-05-18  Richard Biener  <rguenther@suse.de>
19359         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
19360         results when processing array refs with variable index.
19362 2018-05-18  Toon Moene  <toon@moene.org>
19364         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
19365         directly after that of -floop-interchange. Indicate that both
19366         options are enabled by default when specifying -O3.
19368 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19370         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
19371         iterator.  Delete separate integer-mode vec_set<mode> expander.
19372         (aarch64_simd_vec_setv2di): Delete.
19373         (vec_setv2di): Delete.
19374         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
19375         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
19376         the "w, r" alternative.
19378 2018-05-18  Martin Liska  <mliska@suse.cz>
19380         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
19381         * tree-pass.h (make_pass_lower_switch_O0): New function.
19382         * tree-switch-conversion.c (node_has_low_bound): Remove.
19383         (node_has_high_bound): Likewise.
19384         (node_is_bounded): Likewise.
19385         (class pass_lower_switch): Make it a template type and create
19386         two instances.
19387         (pass_lower_switch::execute): Add template argument.
19388         (make_pass_lower_switch): New function.
19389         (make_pass_lower_switch_O0): New function.
19390         (do_jump_if_equal): Remove.
19391         (emit_case_nodes): Simplify to just handle all 3 cases and leave
19392         all the hard work to tree optimization passes.
19394 2018-05-18  Martin Liska  <mliska@suse.cz>
19396         * dbgcnt.c (limit_low): Renamed from limit.
19397         (limit_high): New variable.
19398         (dbg_cnt_is_enabled): Check for upper limit.
19399         (dbg_cnt): Adjust dumping.
19400         (dbg_cnt_set_limit_by_index): Add new argument for high
19401         value.
19402         (dbg_cnt_set_limit_by_name): Likewise.
19403         (dbg_cnt_process_single_pair): Parse new format.
19404         (dbg_cnt_process_opt): Use strtok.
19405         (dbg_cnt_list_all_counters): Remove 'value' and add
19406         'limit_high'.
19407         * doc/invoke.texi: Document changes.
19409 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
19411         * doc/sourcebuild.texi (scalar_all_fma): Document.
19412         * tree.def (FMA_EXPR): Delete.
19413         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
19414         * internal-fn.c (ternary_direct): New macro.
19415         (expand_ternary_optab_fn): Likewise.
19416         (direct_ternary_optab_supported_p): Likewise.
19417         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
19418         * builtins.c (fold_builtin_fma): Delete.
19419         (fold_builtin_3): Don't call it.
19420         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
19421         * expr.c (expand_expr_real_2): Likewise.
19422         * fold-const.c (operand_equal_p): Likewise.
19423         (fold_ternary_loc): Likewise.
19424         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
19425         * gimple.c (DEFTREECODE): Likewise.
19426         * gimplify.c (gimplify_expr): Likewise.
19427         * optabs-tree.c (optab_for_tree_code): Likewise.
19428         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19429         * tree-eh.c (operation_could_trap_p): Likewise.
19430         (stmt_could_throw_1_p): Likewise.
19431         * tree-inline.c (estimate_operator_cost): Likewise.
19432         * tree-pretty-print.c (dump_generic_node): Likewise.
19433         (op_code_prio): Likewise.
19434         * tree-ssa-loop-im.c (stmt_cost): Likewise.
19435         * tree-ssa-operands.c (get_expr_operands): Likewise.
19436         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
19437         * fold-const-call.h (fold_fma): Delete.
19438         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
19439         CFN_FNMA and CFN_FNMS.
19440         (fold_fma): Delete.
19441         * genmatch.c (combined_fn): New enum.
19442         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
19443         (commutative_op): New function.
19444         (commutate): Use it.  Handle more than 2 operands.
19445         (dt_operand::gen_gimple_expr): Use commutative_op.
19446         (parser::parse_expr): Allow :c to be used with non-binary
19447         operators if the commutative operand is known.
19448         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
19449         CFN_FMS, CFN_FNMA and CFN_FNMS.
19450         (backprop::process_assign_use): Remove FMA_EXPR handling.
19451         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
19452         (gen_hsa_fma): New function.
19453         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
19454         IFN_FNMA and IFN_FNMS.
19455         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
19456         * gimple-fold.h (follow_all_ssa_edges): Declare.
19457         * gimple-fold.c (follow_all_ssa_edges): New function.
19458         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
19459         gimple_build interface and use follow_all_ssa_edges to fold the result.
19460         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
19461         instead of checking for optabs directly.
19462         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
19463         rather than FMA_EXPRs.
19464         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
19465         call to IFN_FMA instead of an FMA_EXPR.
19467 2018-05-17  Jim Wilson  <jimw@sifive.com>
19469         * expr.c (do_tablejump): When converting index to Pmode, if we have a
19470         sign extended promoted subreg, and the range does not have the sign bit
19471         set, then do a sign extend.
19473         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
19474         test, check for sign extended subreg and/or constant operands, and
19475         do a sign extend in that case.
19477 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
19479         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
19480         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
19481         Add untyped.
19482         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
19483         Change logics_shift_reg to logics_shift_imm.
19484         (thunderx2t99_fp_loadpair_basic): Delete.
19485         (thunderx2t99_fp_storepair_basic): Delete.
19486         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
19487         (thunderx2t99_asimd_polynomial): Delete.
19488         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
19489         and neon_fp_mul_d_scalar_q.
19490         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
19491         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
19492         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
19493         (thunderx2t99_asimd_lut): Add missing tbl types.
19494         (thunderx2t99_asimd_ext): Delete.
19495         (thunderx2t99_asimd_load1_1_mult): Delete.
19496         (thunderx2t99_asimd_load1_2_mult): Delete.
19497         (thunderx2t99_asimd_load1_ldp): New.
19498         (thunderx2t99_asimd_load1): New.
19499         (thunderx2t99_asimd_load2): Add missing *load2* types.
19500         (thunderx2t99_asimd_load3): New.
19501         (thunderx2t99_asimd_load4): New.
19502         (thunderx2t99_asimd_store1_1_mult): Delete.
19503         (thunderx2t99_asimd_store1_2_mult): Delete.
19504         (thunderx2t99_asimd_store2_mult): Delete.
19505         (thunderx2t99_asimd_store2_onelane): Delete.
19506         (thunderx2t99_asimd_store_stp): New.
19507         (thunderx2t99_asimd_store1): New.
19508         (thunderx2t99_asimd_store2): New.
19509         (thunderx2t99_asimd_store3): New.
19510         (thunderx2t99_asimd_store4): New.
19512 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
19514         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
19515         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
19517 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
19518             Segher Boessenkool  <segher@kernel.crashing.org>
19520         PR target/85698
19521         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
19522         operand.
19524 2018-05-17  Richard Biener  <rguenther@suse.de>
19526         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
19527         for pruning loop and prune defs feeding only already visited PHIs.
19529 2018-05-17  Richard Biener  <rguenther@suse.de>
19531         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
19533 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
19534             Richard Biener  <rguenther@suse.de>
19536         PR tree-optimization/85793
19537         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
19538         for VMAT_ELEMENTWISE.
19540 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19542         * internal-fn.h (lookup_internal_fn): Declare
19543         * internal-fn.c (lookup_internal_fn): New function.
19544         * gimple.c (gimple_build_call_from_tree): Handle calls to
19545         internal functions.
19546         * gimple-pretty-print.c (dump_gimple_call): Print "." before
19547         internal function names.
19548         * tree-pretty-print.c (dump_generic_node): Likewise.
19549         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
19551 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19553         * gimple-fold.h (gimple_build): Make the function forms take
19554         combined_fn rather than built_in_function.
19555         (gimple_simplify): Likewise.
19556         * gimple-match-head.c (gimple_simplify): Likewise.
19557         * gimple-fold.c (gimple_build): Likewise.
19558         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
19559         rather than gimple_build_call_internal.
19560         (get_initial_defs_for_reduction): Likewise.
19561         (vect_create_epilog_for_reduction): Likewise.
19562         (vectorizable_live_operation): Likewise.
19564 2018-05-17  Martin Liska  <mliska@suse.cz>
19566         * gimple-ssa-sprintf.c (format_directive): Do not use
19567         space in between 'G_' and '('.
19569 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
19571         PR target/85323
19572         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
19573         even if the mask is not all ones.
19575         PR target/85323
19576         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
19577         vector.
19578         (ix86_gimple_fold_builtin): Likewise.
19580         PR target/85323
19581         * config/i386/i386.c: Include tree-vector-builder.h.
19582         (ix86_vector_shift_count): New function.
19583         (ix86_fold_builtin): Fold shift builtins by scalar count.
19584         (ix86_gimple_fold_builtin): Likewise.
19586         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
19587         _mm512_setzero): New intrinsics.
19589 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
19590             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19592         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
19593         code generation for cases where splatting a value is not useful.
19594         * simplify-rtx.c (simplify_ternary_operation): Simplify
19595         vec_merge across a vec_duplicate and a paradoxical subreg forming
19596         a vector mode to a vec_concat.
19598 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
19600         * config.gcc: Support "goldmont-plus".
19601         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19602         "goldmont-plus".
19603         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19604         PROCESSOR_GOLDMONT_PLUS.
19605         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
19606         (processor_target_table): Add "goldmont-plus".
19607         (PTA_GOLDMONT_PLUS): Define.
19608         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
19609         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
19610         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
19611         (fold_builtin_cpu): Add "goldmont-plus".
19612         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
19613         (ix86_option_override_internal): Add "goldmont-plus".
19614         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
19615         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
19616         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
19617         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
19619 2018-05-17  Richard Biener  <rguenther@suse.de>
19621         PR tree-optimization/85757
19622         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
19623         remove defs that only feed that PHI from further processing.
19625 2018-05-16  Jim Wilson  <jimw@sifive.com>
19627         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
19628         asterisk to name.
19629         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
19631 2018-05-16  Mark Wielaard  <mark@klomp.org>
19633         * dwarf2out.c (count_index_strings): New function.
19634         (output_indirect_strings): Call count_index_strings and generate
19635         header for dwarf_version >= 5.
19637 2018-05-16  Mark Wielaard  <mark@klomp.org>
19639         * dwarf2out.c (dwarf_FORM): New function.
19640         (set_indirect_string): Use dwarf_FORM.
19641         (reset_indirect_string): Likewise.
19642         (size_of_die): Likewise.
19643         (value_format): Likewise.
19644         (output_die): Likewise.
19645         (add_skeleton_AT_string): Likewise.
19646         (output_macinfo_op): Likewise.
19647         (index_string): Likewise.
19648         (output_index_string_offset): Likewise.
19649         (output_index_string): Likewise.
19650         (count_index_strings): Likewise.
19652 2018-05-16  Carl Love  <cel@us.ibm.com>
19654         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19655         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
19657 2018-05-16  Martin Jambor  <mjambor@suse.cz>
19659         * ipa-prop.c (ipa_free_all_edge_args): Remove.
19660         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
19662 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
19664         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
19665         (fnma<mode>4): Likewise.
19666         (fms<mode>4): Likewise.
19667         (fnms<mode>4): Likewise.
19668         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
19669         (aarch64_fnma<mode>4): Likewise.
19670         (aarch64_fms<mode>4): Likewise.
19671         (aarch64_fnms<mode>4): Likewise.
19672         (aarch64_fnmadd<mode>4): Likewise.
19674 2018-05-16  Jason Merrill  <jason@redhat.com>
19676         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
19678 2018-05-16  Richard Biener  <rguenther@suse.de>
19680         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
19681         (dump_stmt_cost): Declare.
19682         (add_stmt_cost): Dump cost we add.
19683         (add_stmt_costs): New function.
19684         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
19685         No longer exported.
19686         (vect_analyze_stmt): Adjust prototype.
19687         (vectorizable_condition): Likewise.
19688         (vectorizable_live_operation): Likewise.
19689         (vectorizable_reduction): Likewise.
19690         (vectorizable_induction): Likewise.
19691         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
19692         cost vector to pass to vectorizable_ and record afterwards.
19693         (vect_model_reduction_cost): Take cost vector argument and adjust.
19694         (vect_model_induction_cost): Likewise.
19695         (vectorizable_reduction): Likewise.
19696         (vectorizable_induction): Likewise.
19697         (vectorizable_live_operation): Likewise.
19698         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
19699         SLP_TREE_NUMBER_OF_VEC_STMTS.
19700         (vect_analyze_slp_cost_1): Remove.
19701         (vect_analyze_slp_cost): Likewise.
19702         (vect_slp_analyze_node_operations): Take visited args and
19703         a target cost vector.  Avoid processing already visited stmt sets.
19704         (vect_slp_analyze_operations): Use a local cost vector to gather
19705         costs and register those of non-discarded instances.
19706         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
19707         (vect_schedule_slp_instance): Remove copying of
19708         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
19709         zero.
19710         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
19711         adding cost.  Record cost entry location.
19712         (vect_prologue_cost_for_slp_op): Function to compute cost of
19713         a constant or invariant generated for SLP vect in the prologue,
19714         split out from vect_analyze_slp_cost_1.
19715         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
19716         (vect_model_promotion_demotion_cost): Likewise.
19717         (vect_model_store_cost): Likewise, make static.
19718         (vect_model_load_cost): Likewise.
19719         (vectorizable_bswap): Add cost vector arg and adjust.
19720         (vectorizable_call): Likewise.
19721         (vectorizable_simd_clone_call): Likewise.
19722         (vectorizable_conversion): Likewise.
19723         (vectorizable_assignment): Likewise.
19724         (vectorizable_shift): Likewise.
19725         (vectorizable_operation): Likewise.
19726         (vectorizable_store): Likewise.
19727         (vectorizable_load): Likewise.
19728         (vectorizable_condition): Likewise.
19729         (vectorizable_comparison): Likewise.
19730         (can_vectorize_live_stmts): Likewise.
19731         (vect_analyze_stmt): Likewise.
19732         (vect_transform_stmt): Adjust calls to vectorizable_*.
19733         * tree-vectorizer.c: Include gimple-pretty-print.h.
19734         (dump_stmt_cost): New function.
19736 2018-05-16  Richard Biener  <rguenther@suse.de>
19738         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
19739         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
19740         * tree-ssa-dse.c: Include tree-ssa-loop.h.
19741         (check_name): New callback.
19742         (dse_classify_store): Track cycles via a visited bitmap of PHI
19743         defs and simplify handling of in-loop and across loop dead stores
19744         and properly fail for loop-variant refs.  Handle byte-tracking with
19745         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
19746         limiting the walk.
19748 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
19750         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
19751         (vect_get_mask_type_for_stmt): Likewise.
19752         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
19753         split out from...
19754         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
19755         to determine the statement's vector type and the vector type that
19756         should be used for calculating nunits.  Deal with cases in which
19757         the type has to be deferred.
19758         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
19759         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
19760         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
19761         (vect_determine_vf_for_stmt): New functions, split out from...
19762         (vect_determine_vectorization_factor): ...here.
19763         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
19764         (vect_get_mask_type_for_stmt): New functions, split out from
19765         vect_determine_vectorization_factor.
19767 2018-05-16  Richard Biener  <rguenther@suse.de>
19769         * tree-cfg.c (verify_gimple_assign_ternary): Properly
19770         verify the [VEC_]COND_EXPR embedded comparison.
19772 2018-05-15  Martin Sebor  <msebor@redhat.com>
19774         PR tree-optimization/85753
19775         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
19776         RECORD_TYPE in addition to ARRAY_TYPE.
19778 2018-05-15  Martin Sebor  <msebor@redhat.com>
19780         PR middle-end/85643
19781         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
19783 2018-05-15  Richard Biener  <rguenther@suse.de>
19785         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
19786         add by_clobber_p one.  Change algorithm to collect all defs
19787         representing uses we need to walk and try reducing them to
19788         a single one before failing.
19789         (dse_dom_walker::dse_optimize_stmt): Adjust.
19791 2018-05-13  Mark Wielaard  <mark@klomp.org>
19793         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
19794         (size_of_loc_descr): Likewise.
19795         (output_loc_operands): Likewise.
19796         (output_loc_operands_raw): Likewise.
19797         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
19798         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
19799         (hash_loc_operands): Likewise.
19800         (compare_loc_operands): Likewise.
19802 2018-05-14  Mark Wielaard  <mark@klomp.org>
19804         * dwarf2out.c (count_index_addrs): New function.
19805         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
19807 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19809         PR tree-optimization/83648
19810         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
19811         return value as malloc candidate.
19813 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19815         PR ipa/85734
19816         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
19817         param as true in call to suggest_attribute.
19819 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
19821         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
19822         -mreadonly-in-sdata.
19824 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19826         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
19827         New pattern.
19828         (aarch64_crypto_aesd_fused): Likewise.
19830 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
19832         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
19833         (movsi_aarch64): Likewise.
19834         (load_pairsi): Likewise.
19835         (load_pairdi): Likewise.
19836         (store_pairsi): Likewise.
19837         (store_pairdi): Likewise.
19838         (load_pairsf): Likewise.
19839         (load_pairdf): Likewise.
19840         (store_pairsf): Likewise.
19841         (store_pairdf): Likewise.
19842         (zero_extend): Likewise.
19843         (trunc): Swap alternatives.
19844         (fcvt_target): Add '?' to prefer w over r.
19846 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
19848         PR target/85756
19849         * config/i386/i386.md: Disallow non-commutative arithmetics in
19850         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
19851         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
19852         in the peephole2 before it.
19854 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
19856         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
19857         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
19858         (ix86_handle_option): Handle -mcldemote.
19859         * config.gcc: New header.
19860         * config/i386/cldemoteintrin.h: New file.
19861         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
19862         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19863         -mcldemote.
19864         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19865         OPTION_MASK_ISA_CLDEMOTE.
19866         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
19867         (ix86_valid_target_attribute_inner_p): Ditto.
19868         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
19869         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
19870         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
19871         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
19872         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
19873         (cldemote): New.
19874         * config/i386/i386.opt: Add -mcldemote.
19875         * config/i386/x86intrin.h: New header.
19876         * doc/invoke.texi: Add -mcldemote.
19878 2018-05-14  Richard Biener  <rguenther@suse.de>
19880         * doc/match-and-simplify.texi: Adjust :s documentation.
19882 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
19884         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
19885         intended memcpy size.
19886         (REORDER_45): Likewise.
19888 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
19890         * sort.cc: New file.
19891         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
19892         * vec.c (qsort_chk): Use gcc_qsort.
19893         * Makefile.in (OBJS-libcommon): Add sort.o.
19894         (build/sort.o): New target.  Use it...
19895         (BUILD_RTL): ... here, and...
19896         (build/gencfn-macros): ... here, and...
19897         (build/genmatch): ... here.
19899 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19900             Chung-Ju Wu  <jasonwucj@gmail.com>
19902         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
19903         * config/nds32/nds32-graywolf.md: New file.
19904         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
19905         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
19906         pipeline.
19907         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
19908         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
19909         * config/nds32/nds32.md (pipeline_model): Add graywolf.
19910         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
19911         * config/nds32/pipelines.md: Include n15 settings.
19913 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19914             Chung-Ju Wu  <jasonwucj@gmail.com>
19916         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
19917         * config/nds32/nds32-n13.md: New file.
19918         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
19919         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
19920         pipeline.
19921         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
19922         * config/nds32/nds32.md (pipeline_model): Add n13.
19923         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
19924         * config/nds32/pipelines.md: Include n13 settings.
19926 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19927             Chung-Ju Wu  <jasonwucj@gmail.com>
19929         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
19930         * config/nds32/nds32-n10.md: New file.
19931         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
19932         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
19933         pipeline.
19934         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
19935         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
19936         * config/nds32/nds32.md (pipeline_model): Add n10.
19937         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
19938         * config/nds32/pipelines.md: Include n10 settings.
19940 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
19941             Kito Cheng  <kito.cheng@gmail.com>
19942             Chung-Ju Wu  <jasonwucj@gmail.com>
19944         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
19945         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19946         Add enum values for DSP extension instructions.
19947         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
19948         New constraints.
19949         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
19950         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
19951         New code iterators.
19952         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
19953         * config/nds32/nds32-dspext.md: New file for DSP implementation.
19954         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
19955         * config/nds32/nds32-intrinsic.md: Likewise.
19956         * config/nds32/nds32_intrinsic.h: Likewise.
19957         * config/nds32/nds32-md-auxiliary.c: Likewise.
19958         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
19959         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
19960         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
19961         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
19962         * config/nds32/nds32-protos.h: New declarations for DSP extension.
19963         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
19964         TYPE_DMAC in switch statement.
19965         * config/nds32/nds32.c: New checking and implementation for DSP
19966         extension instructions.
19967         * config/nds32/nds32.h: Likewise.
19968         * config/nds32/nds32.md: Likewise.
19969         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
19970         * config/nds32/predicates.md: Implement new predicates for DSP
19971         extension.
19973 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
19975         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
19976         Reformat alternatives and attributes so it is easier to identify
19977         which constraints/attributes go with which instruction.
19978         (mov<mode>_hardfloat32, FMOVE64): Likewise.
19979         (mov<mode>_softfloat32, FMOVE64): Likewise.
19980         (mov<mode>_hardfloat64, FMOVE64): Likewise.
19981         (mov<mode>_softfloat64, FMOVE64): Likewise.
19983 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19985         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19986         subsection.
19987         (Basic PowerPC Built-in Functions): The new name of the
19988         subsection previously known as "PowerPC Built-in Functions".
19989         (Basic PowerPC Built-in Functions Available on all Configurations):
19990         New subsubsection.
19991         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
19992         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
19993         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
19994         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
19996 2018-05-11  Martin Jambor  <mjambor@suse.cz>
19998         PR ipa/85655
19999         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
20000         single const.
20002 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
20004         PR target/85733
20005         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
20007 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
20009         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
20010         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
20011         (ix86_handle_option): Handle -mwaitpkg.
20012         * config.gcc: New header.
20013         * config/i386/cpuid.h (bit_WAITPKG): New bit.
20014         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
20015         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
20016         function type.
20017         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20018         OPTION_MASK_ISA_WAITPKG.
20019         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
20020         (ix86_option_override_internal): Add PTA_WAITPKG.
20021         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
20022         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
20023         IX86_BUILTIN_TPAUSE.
20024         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
20025         __builtin_ia32_umwait and __builtin_ia32_tpause.
20026         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
20027         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
20028         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
20029         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
20030         UNSPECV_TPAUSE): New.
20031         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
20032         * config/i386/i386.opt: Add -mwaitpkg.
20033         * config/i386/waitpkgintrin.h: New file.
20034         * config/i386/x86intrin.h: New header.
20035         * doc/invoke.texi: Add -mwaitpkg.
20037 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
20039         PR target/85606
20040         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
20041         equivalent.
20042         (cortex-m0): Use armv6s-m isa.
20043         (cortex-m0plus): Likewise.
20044         (cortex-m1): Likewise.
20045         (cortex-m0.small-multiply): Likewise.
20046         (cortex-m0plus.small-multiply): Likewise.
20047         (cortex-m1.small-multiply): Likewise.
20049 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
20050             Jakub Jelinek  <jakub@redhat.com>
20052         PR tree-optimization/85692
20053         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
20054         source permute as well.
20056 2018-05-11  Martin Liska  <mliska@suse.cz>
20058         PR sanitizer/85556
20059         * doc/extend.texi: Document LLVM style format for no_sanitize
20060         attribute.
20062 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
20064         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
20065         mode_supports_vsx_dform_quad to mode_supports_dq_form.
20066         (mode_supports_vsx_dform_quad): Likewise.
20067         (mode_supports_vmx_dform): Move these functions to be next to the
20068         other mode_supports functions.
20069         (mode_supports_dq_form): Likewise.
20070         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
20071         mode_supports_dq_form.
20072         (reg_offset_addressing_ok_p): Likewise.
20073         (offsettable_ok_by_alignment): Likewise.
20074         (rs6000_legitimate_offset_address_p): Likewise.
20075         (legitimate_lo_sum_address_p): Likewise.
20076         (rs6000_legitimize_address): Likewise.
20077         (rs6000_legitimize_reload_address): Likewise.
20078         (rs6000_secondary_reload_inner): Likewise.
20079         (rs6000_preferred_reload_class): Likewise.
20080         (rs6000_output_move_128bit): Likewise.
20082 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
20084         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
20085         Generate SImode target register for null target.
20086         <case IX86_BUILTIN_XGETBV>: Ditto.
20087         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
20088         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
20090 2018-05-10  Carl Love  <cel@us.ibm.com>
20092         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
20093         dcbtt and dcbtstt if operands[2] is 0.
20095 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
20097         PR target/85693
20098         * config/i386/sse.md (usadv64qi): New expander.
20100 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
20102         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
20103         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
20104         -maltivec=be support.
20105         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
20106         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
20107         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
20108         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
20109         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
20110         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
20111         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
20112         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
20113         altivec_vsumsws): Adjust.
20114         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
20115         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
20116         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
20117         support.
20118         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
20119         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
20120         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
20121         (altivec_lve<VI_char>x): Delete expand.
20122         (*altivec_lve<VI_char>x_internal): Rename to...
20123         (altivec_lve<VI_char>x): ... this.
20124         (altivec_lvxl_<mode>): Delete expand.
20125         (*altivec_lvxl_<mode>_internal): Rename to ...
20126         (altivec_lvxl_<mode>): ... this.
20127         (altivec_stvxl_<mode>): Delete expand.
20128         (*altivec_stvxl_<mode>_internal): Rename to ...
20129         (altivec_stvxl_<mode>): ... this.
20130         (altivec_stve<VI_char>x): Delete expand.
20131         (*altivec_stve<VI_char>x_internal): Rename to ...
20132         (altivec_stve<VI_char>x): ... this.
20133         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
20134         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
20135         reduc_plus_scal_<mode>): Adjust.
20136         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
20137         comment.
20138         (rs6000_cpu_cpp_builtins): Adjust.
20139         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
20140         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
20141         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
20142         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
20143         -maltivec=be support.
20144         (rs6000_split_vec_extract_var): Adjust.
20145         (rs6000_split_v4si_init): Adjust.
20146         (swap_selector_for_mode): Delete.
20147         (altivec_expand_lvx_be, altivec_expand_stvx_be,
20148         altivec_expand_stvex_be): Delete.
20149         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
20150         -maltivec=be support.
20151         (rs6000_gimple_fold_builtin): Ditto.
20152         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
20153         Adjust.
20154         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
20155         (TARGET_DIRECT_MOVE_64BIT): Adjust.
20156         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
20157         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
20158         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
20159         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
20160         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
20161         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
20162         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
20163         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
20164         anonymous split): Adjust.
20165         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
20166         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
20168 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
20170         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
20171         when --with-gxx-include-dir is also specified.
20172         * configure: Regenerate.
20174 2018-05-09  Jim Wilson  <jimw@sifive.com>
20176         PR target/84797
20177         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
20178         * config/riscv/t-withmultilib: New.
20179         * config/riscv/withmultilib.h: New.
20180         * doc/install.texi: Document RISC-V --with-multilib-list support.
20182 2018-05-09  Richard Biener  <rguenther@suse.de>
20184         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
20185         vector.
20186         (vect_bb_vectorization_profitable_p): Adjust.  Compute
20187         actual scalar cost using the cost vector and the add_stmt_cost
20188         machinery.
20190 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20192         PR rtl-optimization/85645
20193         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
20194         in the REG_CFA_REGISTER note for LR, don't leave it empty.
20196 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20198         PR rtl-optimization/85645
20199         * shrink-wrap.c (spread_components): Return a boolean saying if
20200         anything was changed.
20201         (try_shrink_wrapping_separate): Iterate spread_components until
20202         nothing changes anymore.
20204 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20206         PR rtl-optimization/85645
20207         * regrename.c (build_def_use): Also kill the chains that include the
20208         destination of a REG_CFA_REGISTER note.
20210 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20212         PR rtl-optimization/85645
20213         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
20214         insn that has a REG_CFA_REGISTER note.
20216 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
20218         * cfgexpand.c (expand_clobber): New function.
20219         (expand_gimple_stmt_1): Use it.
20220         * tree-vect-stmts.c (vect_clobber_variable): New function,
20221         split out from...
20222         (vectorizable_simd_clone_call): ...here.
20223         (vectorizable_store): Emit a clobber either side of an
20224         IFN_STORE_LANES sequence.
20225         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
20227 2018-05-09  Tom de Vries  <tom@codesourcery.com>
20229         PR target/85626
20230         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
20231         (define_insn "trap_if_false"): Add exit after trap.
20233 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
20235         PR rtl-optimization/85638
20236         * bb-reorder.c: Include common/common-target.h.
20237         (create_forwarder_block): New function extracted from...
20238         (fix_up_crossing_landing_pad): ...here.  Rename into...
20239         (dw2_fix_up_crossing_landing_pad): ...this.
20240         (sjlj_fix_up_crossing_landing_pad): New function.
20241         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
20242         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
20243         from both partitions and exit the loop after one iteration.
20245 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20247         Revert:
20248         * doc/extend.texi (PowerPC Built-in Functions): Rename this
20249         subsection.
20250         (Basic PowerPC Built-in Functions): The new name of the
20251         subsection previously known as "PowerPC Built-in Functions".
20252         (Basic PowerPC Built-in Functions Available on all Configurations):
20253         New subsubsection.
20254         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20255         subsubsection.
20256         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20257         subsubsection.
20258         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20259         subsubsection.
20260         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20261         subsubsection.
20263 2018-05-08  Jim Wilson  <jimw@sifive.com>
20265         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
20266         (LD_EMUL_SUFFIX): New.
20267         (LINK_SPEC): Use it.
20269 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20271         * doc/extend.texi (PowerPC Built-in Functions): Rename this
20272         subsection.
20273         (Basic PowerPC Built-in Functions): The new name of the
20274         subsection previously known as "PowerPC Built-in Functions".
20275         (Basic PowerPC Built-in Functions Available on all Configurations):
20276         New subsubsection.
20277         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20278         subsubsection.
20279         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20280         subsubsection.
20281         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20282         subsubsection.
20283         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20284         subsubsection.
20286 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
20288         PR target/85683
20289         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
20290         after cmpelim optimization.
20292 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
20294         * config.gcc: Support "goldmont".
20295         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
20296         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20297         PROCESSOR_GOLDMONT.
20298         * config/i386/i386.c (m_GOLDMONT): Define.
20299         (processor_target_table): Add "goldmont".
20300         (PTA_GOLDMONT): Define.
20301         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
20302         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
20303         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
20304         (fold_builtin_cpu): Add "goldmont".
20305         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
20306         (ix86_option_override_internal): Add "goldmont".
20307         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
20308         (processor_type): Add PROCESSOR_GOLDMONT.
20309         * config/i386/i386.md: Add CPU "glm".
20310         * config/i386/glm.md: New file.
20311         * config/i386/x86-tune.def: Add m_GOLDMONT.
20312         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
20314 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
20316         PR target/85572
20317         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
20318         E_V4DImode.
20319         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
20320         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
20321         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
20323         PR target/85317
20324         * config/i386/i386.c (ix86_fold_builtin): Handle
20325         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
20327         PR target/85480
20328         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
20329         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
20331 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
20333         PR target/85658
20334         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
20335         (check_arch): Likewise.
20336         (check_fpu): Return the result rather than printing it.
20337         (end arch): Fix operator precedence.
20338         (end cpu): Likewise.
20339         (END): Print the result from check_fpu.
20341 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20342             Alan Hayward  <alan.hayward@arm.com>
20343             David Sherwood  <david.sherwood@arm.com>
20345         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
20346         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
20347         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
20348         (*fcmuo<mode>_and): New patterns.
20350 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20352         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
20353         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
20354         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
20355         (cmp_op, sve_imm_con): New code attributes.
20356         (SVE_COND_INT_CMP, imm_con): Delete.
20357         (cmp_op): Remove above unspecs from int attribute.
20358         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
20359         to...
20360         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
20361         comparison-specific unspecs.
20362         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
20363         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
20364         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
20365         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
20366         (*vec_fcm<cmp_op><mode>): Rename to...
20367         (*fcm<cmp_op><mode>): ...this and adjust likewise.
20368         (*vec_fcmuo<mode>): Rename to...
20369         (*fcmuo<mode>): ...this and adjust likewise.
20370         (*pred_fcm<cmp_op><mode>): New pattern.
20371         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
20372         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
20373         functions.
20374         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
20375         and UNORDERED.
20376         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
20377         (aarch64_emit_sve_predicated_cond): New function.
20378         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
20379         (aarch64_emit_unspec_cond_or): Replace with...
20380         (aarch64_emit_sve_or_conds): ...this new function.  Use
20381         aarch64_emit_sve_ptrue_op for the individual comparisons and
20382         aarch64_emit_binop to OR them together.
20383         (aarch64_emit_inverted_unspec_cond): Replace with...
20384         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
20385         aarch64_emit_sve_ptrue_op for the comparison and
20386         aarch64_emit_unop to invert the result.
20387         (aarch64_expand_sve_vec_cmp_float): Update after the above
20388         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
20390 2018-05-07  Nathan Sidwell  <nathan@acm.org>
20392         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
20393         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
20394         (Backwards Compatibility): Likewise.
20396 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20398         PR bootstrap/85681
20399         Revert:
20400         2018-05-07  Luis Machado  <luis.machado@linaro.org>
20402         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20403         <prefetch_dynamic_strides>: New const bool field.
20404         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20405         prefetch_dynamic_strides.
20406         (exynosm1_prefetch_tune): Likewise.
20407         (thunderxt88_prefetch_tune): Likewise.
20408         (thunderx_prefetch_tune): Likewise.
20409         (thunderx2t99_prefetch_tune): Likewise.
20410         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20411         to false.
20412         (aarch64_override_options_internal): Update to set
20413         PARAM_PREFETCH_DYNAMIC_STRIDES.
20414         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20415         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20416         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20417         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20418         prefetch-dynamic-strides setting.
20420         2018-05-07  Luis Machado  <luis.machado@linaro.org>
20422         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20423         <minimum_stride>: New const int field.
20424         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20425         minimum_stride field.
20426         (exynosm1_prefetch_tune): Likewise.
20427         (thunderxt88_prefetch_tune): Likewise.
20428         (thunderx_prefetch_tune): Likewise.
20429         (thunderx2t99_prefetch_tune): Likewise.
20430         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20431         (aarch64_override_options_internal): Update to set
20432         PARAM_PREFETCH_MINIMUM_STRIDE.
20433         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20434         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20435         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20436         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20437         stride is constant and is below the minimum stride threshold.
20439 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20441         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
20442         to 512.
20444 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20446         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20447         <prefetch_dynamic_strides>: New const bool field.
20448         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20449         prefetch_dynamic_strides.
20450         (exynosm1_prefetch_tune): Likewise.
20451         (thunderxt88_prefetch_tune): Likewise.
20452         (thunderx_prefetch_tune): Likewise.
20453         (thunderx2t99_prefetch_tune): Likewise.
20454         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20455         to false.
20456         (aarch64_override_options_internal): Update to set
20457         PARAM_PREFETCH_DYNAMIC_STRIDES.
20458         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20459         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20460         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20461         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20462         prefetch-dynamic-strides setting.
20464 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20466         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20467         <minimum_stride>: New const int field.
20468         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20469         minimum_stride field.
20470         (exynosm1_prefetch_tune): Likewise.
20471         (thunderxt88_prefetch_tune): Likewise.
20472         (thunderx_prefetch_tune): Likewise.
20473         (thunderx2t99_prefetch_tune): Likewise.
20474         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20475         (aarch64_override_options_internal): Update to set
20476         PARAM_PREFETCH_MINIMUM_STRIDE.
20477         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20478         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20479         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20480         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20481         stride is constant and is below the minimum stride threshold.
20483 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
20485         PR c++/85659
20486         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
20487         the type is addressable.  Don't force op into register if it has
20488         BLKmode.
20490 2018-05-05  Roland McGrath  <mcgrathr@google.com>
20492         PR other/77609
20493         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
20494         any section for which we don't know a specific type it should have,
20495         regardless of name.  Previously this was done only for the exact
20496         names ".init_array", ".fini_array", and ".preinit_array".
20497         (default_elf_asm_named_section): Add comment about
20498         relationship with default_section_type_flags and SECTION_NOTYPE.
20499         (get_section): Don't consider it a type conflict if one side has
20500         SECTION_NOTYPE and the other doesn't, as long as neither has the
20501         SECTION_BSS et al used in the default_section_type_flags logic.
20503 2018-05-05  Tom de Vries  <tom@codesourcery.com>
20505         PR target/85653
20506         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
20507         (workaround_barsyncs): New function.
20508         (nvptx_reorg): Use workaround_barsyncs.
20509         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
20510         (define_expand "nvptx_membar_cta"): New define_expand.
20511         (define_insn "*nvptx_membar_cta"): New insn.
20513 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
20515         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
20516         To improve optimization opportunities.
20517         * builtin-types.def: The new needed builtin types for the above.
20519 2018-05-04  Richard Biener  <rguenther@suse.de>
20521         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
20522         * gimple-ssa-store-merging.c
20523         (imm_store_chain_info::output_merged_store): Remove redundant create,
20524         release split_store vector contents on failure.
20525         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
20526         scalar stmt vector on cache hit.
20528 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
20530         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
20531         Xilinx FP support.
20532         * config.gcc (powerpc-xilinx-eabi*): Remove.
20533         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
20534         support.
20535         (fusion_addis_mem_combo_load): Ditto.
20536         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
20537         FP support.
20538         (rs6000_cpu_cpp_builtins): Ditto.
20539         * config/rs6000/rs6000-linux.c
20540         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
20541         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
20542         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
20543         support.
20544         (rs6000_setup_reg_addr_masks): Ditto.
20545         (rs6000_init_hard_regno_mode_ok): Ditto.
20546         (rs6000_option_override_internal): Ditto.
20547         (legitimate_lo_sum_address_p): Ditto.
20548         (rs6000_legitimize_address): Ditto.
20549         (rs6000_legitimize_reload_address): Ditto.
20550         (rs6000_legitimate_address_p): Ditto.
20551         (abi_v4_pass_in_fpr): Ditto.
20552         (setup_incoming_varargs): Ditto.
20553         (rs6000_gimplify_va_arg): Ditto.
20554         (rs6000_split_multireg_move): Ditto.
20555         (rs6000_savres_strategy): Ditto.
20556         (rs6000_emit_prologue_components): Ditto.
20557         (rs6000_emit_epilogue_components): Ditto.
20558         (rs6000_emit_prologue): Ditto.
20559         (rs6000_emit_epilogue): Ditto.
20560         (rs6000_elf_file_end): Ditto.
20561         (rs6000_function_value): Ditto.
20562         (rs6000_libcall_value): Ditto.
20563         * config/rs6000/rs6000.h: Ditto.
20564         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
20565         (TARGET_MINMAX): ... this.  New.
20566         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
20567         * config/rs6000/rs6000.md: Remove Xilinx FP support.
20568         (*movsi_internal1_single): Delete.
20569         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
20570         mfpu=, mxilinx-fpu): Delete.
20571         * config/rs6000/singlefp.h: Delete.
20572         * config/rs6000/sysv4.h: Remove Xilinx FP support.
20573         * config/rs6000/t-rs6000: Ditto.
20574         * config/rs6000/t-xilinx: Delete.
20575         * config/rs6000/titan.md: Adjust for fp_type removal.
20576         * config/rs6000/vsx.md: Remove Xilinx FP support.
20577         (VStype_simple): Delete.
20578         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
20579         * config/rs6000/xfpu.h: Delete.
20580         * config/rs6000/xfpu.md: Delete.
20581         * config/rs6000/xilinx.h: Delete.
20582         * config/rs6000/xilinx.opt: Delete.
20583         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
20584         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
20586 2018-05-04  Tom de Vries  <tom@codesourcery.com>
20588         PR libgomp/85639
20589         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
20590         if ignore == 0.
20592 2018-05-04  Richard Biener  <rguenther@suse.de>
20594         PR middle-end/85627
20595         * tree-complex.c (update_complex_assignment): We are always in SSA form.
20596         (expand_complex_div_wide): Likewise.
20597         (expand_complex_operations_1): Likewise.
20598         (expand_complex_libcall): Preserve EH info of the original stmt.
20599         (tree_lower_complex): Handle removed blocks.
20600         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
20601         on complex multiplication and division libcall builtins.
20603 2018-05-04  Richard Biener  <rguenther@suse.de>
20605         PR middle-end/85574
20606         * fold-const.c (negate_expr_p): Restrict negation of operand
20607         zero of a division to when we know that can happen without
20608         overflow.
20609         (fold_negate_expr_1): Likewise.
20611 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
20613         PR libstdc++/85466
20614         * real.h (real_nextafter): Declare.
20615         * real.c (real_nextafter): New function.
20616         * fold-const-call.c (fold_const_nextafter): New function.
20617         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
20618         CASE_CFN_NEXTTOWARD.
20619         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
20620         even when arg1_mode is different from arg0_mode.
20622 2018-05-03  Nathan Sidwell  <nathan@acm.org>
20624         * doc/extend.texi (Deprecated Features): Remove
20625         -ffriend-injection.
20626         (Backwards Compatibility): Likewise.
20627         * doc/invoke.texi (C++ Language Options): Likewise.
20628         (C++ Dialect Options): Likewise.
20630 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
20632         PR target/85530
20633         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
20634         _mm512_mask_mullox_epi64): New intrinsics.
20636 2018-05-03  Tom de Vries  <tom@codesourcery.com>
20638         PR testsuite/85106
20639         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20640         dump files): Add offload-tree.
20642 2018-05-03  Richard Biener  <rguenther@suse.de>
20644         PR tree-optimization/85615
20645         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
20646         to loops not nested in BBs loop father to avoid creating multi-entry
20647         loops.
20649 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20651         PR tree-optimization/70291
20652         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
20653         arguments.  Change return type to tree.  Emit libcall as a new
20654         statement rather than replacing existing one when inplace_p is true.
20655         (expand_complex_multiplication_components): New function.
20656         (expand_complex_multiplication): Expand floating-point complex
20657         multiplication using the above.
20658         (expand_complex_division): Rename inner_type parameter to type.
20659         Update expand_complex_libcall call-site.
20660         (expand_complex_operations_1): Update expand_complex_multiplication
20661         and expand_complex_division call-sites.
20663 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20665         PR target/85582
20666         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20667         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20668         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
20669         the highest significant bit of the shift count mask is clear.  In
20670         check whether and[sq]i3 is needed verify that all significant bits
20671         of the shift count other than the highest are set.
20673 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20675         PR libgomp/82428
20676         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
20677         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
20678         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
20679         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
20680         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
20681         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
20682         __builtin_goacc_parlevel_size.
20684 2018-05-02  Richard Biener  <rguenther@suse.de>
20686         PR tree-optimization/85597
20687         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
20688         do not use split vect_get_vec_defs call but call vect_get_slp_defs
20689         directly.
20691 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20693         PR testsuite/85106
20694         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20695         dump files): Add ltrans-tree.
20697 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20699         PR testsuite/85106
20700         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20701         dump files): Add wpa-ipa.
20703 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
20705         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
20706         powerpc*-*-linux*paired* target.
20707         * config/rs6000/750cl.h: Delete.
20708         * config/rs6000/paired.h: Delete.
20709         * config/rs6000/paired.md: Delete.
20710         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
20711         float support.
20712         * config/rs6000/rs6000-builtin.def: Remove paired float support.
20713         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20714         comment.  Remove paired float support.
20715         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
20716         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
20717         VECTOR_PAIRED.
20718         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
20719         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
20720         declarations.
20721         * config/rs6000/rs6000.c: Remove paired float support.
20722         (paired_expand_vector_init, paired_expand_vector_move,
20723         paired_emit_vector_compare, paired_emit_vector_cond_expr,
20724         (paired_expand_lv_builtin, paired_expand_stv_builtin,
20725         paired_expand_builtin, paired_expand_predicate_builtin,
20726         paired_init_builtins): Delete.
20727         * config/rs6000/rs6000.h: Remove paired float support.
20728         * config/rs6000/rs6000.md: Remove paired float support.
20729         (move_from_CR_ov_bit): Delete.
20730         * config/rs6000/rs6000.opt (mpaired): Delete.
20731         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
20732         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
20734 2018-05-02  Richard Biener  <rguenther@suse.de>
20736         PR middle-end/85567
20737         * gimplify.c (gimplify_save_expr): When in SSA form allow
20738         SAVE_EXPRs to compute to SSA vars.
20740 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20742         PR target/85582
20743         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20744         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20745         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
20746         clobber operands[2], instead use a new pseudo.  Formatting fixes.
20748 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
20750         PR tree-optimization/85586
20751         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
20752         exit early for statements in the same group if the accesses are
20753         not strided.
20755 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20757         PR lto/85451
20758         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
20759         error message.
20761 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
20763         PR tree-optimization/85143
20764         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
20766 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20768         PR lto/85451
20769         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
20770         not found" error message.
20772 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20774         PR other/83786
20775         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
20776         * vec.c (test_ordered_remove_if): New function.
20777         (vec_c_tests): Call test_ordered_remove_if.
20778         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
20779         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
20780         * tree-vect-patterns.c (vect_pattern_recog_1): Use
20781         VEC_ORDERED_REMOVE_IF.
20783 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20785         PR tree-optimization/82665
20786         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
20787         pointer subtraction where arguments come from a memchr call.
20789 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
20791         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
20792         --push-state --as-needed and --pop-state instead of --as-needed and
20793         --no-as-needed if ld supports it.
20794         * configure: Regenerated.
20796         PR web/85578
20797         * doc/install.texi2html: Replace _002d with - and _002a with * in
20798         generated html files using sed.
20800 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20802         PR c++/85523
20803         * gcc-rich-location.c (blank_line_before_p): New function.
20804         (use_new_line): New function.
20805         (gcc_rich_location::add_fixit_insert_formatted): New function.
20806         * gcc-rich-location.h
20807         (gcc_rich_location::add_fixit_insert_formatted): New function.
20809 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20811         * selftest.c (assert_streq): Rename "expected" and "actual" to
20812         "val1" and "val2".  Extend NULL-handling to cover both inputs
20813         symmetrically, while still requiring both to be non-NULL for a pass.
20814         * selftest.h (assert_streq): Rename "expected" and "actual" to
20815         "val1" and "val2".
20816         (ASSERT_EQ): Likewise.
20817         (ASSERT_EQ_AT): Likewise.
20818         (ASSERT_KNOWN_EQ): Likewise.
20819         (ASSERT_KNOWN_EQ_AT): Likewise.
20820         (ASSERT_NE): Likewise.
20821         (ASSERT_MAYBE_NE): Likewise.
20822         (ASSERT_MAYBE_NE_AT): Likewise.
20823         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
20824         the assertion to pass.
20825         (ASSERT_STREQ_AT): Likewise.
20827 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20829         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
20830         interaction with -pie.
20832 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20834         * selftest.h: Fix alphabetization of per-source-file selftest
20835         declarations.
20837 2018-04-30  Jason Merrill  <jason@redhat.com>
20839         PR c++/61982 - dead stores to destroyed objects.
20840         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
20841         of clobber.
20843 2018-04-30  Jason Merrill  <jason@redhat.com>
20845         * tree.c (build_clobber): New.
20846         * tree.h: Declare it.
20847         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
20849 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20851         * diagnostic-show-locus.c (layout::layout): Update for
20852         location_get_source_line returning a char_span.
20853         (struct char_span): Move to input.h.
20854         (struct correction): Update for fields in char_span becoming
20855         private.
20856         (struct source_line): Update for location_get_source_line
20857         returning a char_span.
20858         (layout::print_line): Likewise.
20859         * edit-context.c (edited_file::print_content): Likewise.
20860         (edited_file::print_diff_hunk): Likewise.
20861         (edited_file::print_run_of_changed_lines): Likewise.
20862         (edited_file::get_num_lines): Likewise.
20863         (edited_line::edited_line): Likewise.
20864         * final.c (asm_show_source): Likewise.
20865         * input.c (location_get_source_line): Convert return type
20866         from const char * to char_span, losing the final "line_len"
20867         param.
20868         (dump_location_info): Update for the above.
20869         (get_substring_ranges_for_loc): Likewise.  Use a char_span
20870         when handling the literal within the line.
20871         (test_reading_source_line): Update for location_get_source_line
20872         returning a char_span.
20873         * input.h (class char_span): Move here from
20874         diagnostic-show-locus.c, converting from a struct to a class.
20875         Make data members private.
20876         (char_span::operator bool): New.
20877         (char_span::length): New.
20878         (char_span::get_buffer): New.
20879         (char_span::operator[]): New.
20880         (char_span::subspan): Make const.
20881         (char_span::xstrdup): New.
20882         (location_get_source_line): Convert return type from const char *
20883         to char_span, losing the final "line_size" param.
20885 2018-04-30  Jan Hubicka  <jh@suse.cz>
20887         * lto-wrapper.c (ltrans_priorities): New static var.
20888         (cmp_priority): New.
20889         (run_gcc): Read priorities and if doing parallel build order
20890         the Makefile by them.
20892 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20894         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
20896 2018-04-30  Richard Biener  <rguenther@suse.de>
20898         * tree-cfg.c (verify_address): Remove base argument, add
20899         flag whether to check TREE_ADDRESSABLE and do that.
20900         (verify_expr): Remove.
20901         (verify_types_in_gimple_reference): Add pieces from verify_expr.
20902         (verify_gimple_assign_single): Likewise.
20903         (verify_gimple_switch): Likewise.
20904         (verify_expr_location_1): Dereference tp once.  Add (disabled)
20905         piece from verify_expr.
20906         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
20908 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20910         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
20912 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20914         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
20915         (small_data_pattern): Likewise.
20916         (arc_rewrite_small_data): Likewise.
20917         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
20918         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
20919         (get_symbol_alignment): New function.
20920         (legitimate_small_data_address_p): Likewise.
20921         (legitimate_scaled_address): Update, call
20922         legitimate_small_data_address_p.
20923         (output_sdata): New static variable.
20924         (arc_print_operand): Update how we handle small data operands.
20925         (arc_print_operand_address): Likewise.
20926         (arc_legitimate_address_p): Update, use
20927         legitimate_small_data_address_p.
20928         (arc_rewrite_small_data_p): Remove.
20929         (arc_rewrite_small_data_1): Likewise.
20930         (arc_rewrite_small_data): Likewise.
20931         (small_data_pattern): Likewise.
20932         (compact_sda_memory_operand): Update to use
20933         legitimate_small_data_address_p and get_symbol_alignment.
20934         (prepare_move_operands): Don't rewite sdata pattern.
20935         (prepare_extend_operands): Remove.
20936         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
20937         pattern.
20938         (zero_extendqisi2): Likewise.
20939         (zero_extendhisi2): Likewise.
20940         (extendqihi2): Likewise.
20941         (extendqisi2): Likewise.
20942         (extendhisi2): Likewise.
20943         (addsi3): Likewise.
20944         (subsi3): Likewise.
20945         (andsi3): Likewise.
20946         * config/arc/constraints.md (Usd): Change it to memory constraint.
20948 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20950         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
20951         as source of std instructions.
20952         * config/arc/arc.md (movsi_insn): Update pattern predicate to
20953         allow 6-bit constants as source for store instructions.
20954         (movdi_insn): Update instruction pattern to allow 6-bit constants
20955         as source for store instructions.
20957 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20959         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
20961 2018-04-30  Nathan Sidwell  <nathan@acm.org>
20962             Sandra Loosemore <sandra@codesourcery.com>
20964         * dumpfile.c (dump_open): Allow '-' for stdout.
20965         * doc/invoke.texi (Developer Options): Document dump filename
20966         determination early.  Document stdin/stdout selection.
20968 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
20970         Microblaze Target: PIC data text relative
20972         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
20973         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
20974         Add declaration.
20975         * config/microblaze/microblaze.h (microblaze_constant_address_p):
20976         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
20977         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
20978         New addressing mode for data-text relative position indepenedent code.
20979         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
20980         'ADDRESS_SYMBOLIC_TXT_REL'.
20981         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
20982         (microblaze_legitimate_pic_operand): Exclude function calls from
20983         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
20984         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
20985         addresses cases.
20986         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20987         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20988         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
20989         for 'address + offset'.
20990         (microblaze_expand_prologue): Add new function prologue call for
20991         'r20' assignation.
20992         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
20993         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
20994         table in case of TARGET_PIC_DATA_TEXT_REL.
20995         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
20996         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
20997         Add new macros 'UNSPEC_TEXT',
20998         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
20999         + exclude function calls from 'UNSPEC_PLT' in case of data text
21000         relative mode.
21001         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
21002         new target hook for generating address diff vector tables in case of
21003         flag_pic.
21004         * doc/tm.texi : Regenerate.
21005         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
21006         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
21007         of addr diff vector generation.
21008         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
21009         target hook definition.
21010         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
21011         Add default function for generate_pic_addr_diff_vec -> flag_pic.
21012         * doc/invoke.texi (Add new pic option): Add new microblaze pic
21013         option for data text relative.
21015 2018-04-30  Richard Biener  <rguenther@suse.de>
21017         * tree-chrec.h (evolution_function_is_constant_p): Remove
21018         redundant check.
21019         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
21021 2018-04-30  Richard Biener  <rguenther@suse.de>
21023         PR bootstrap/85571
21024         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
21026 2018-04-30  Richard Biener  <rguenther@suse.de>
21028         PR tree-optimization/28364
21029         PR tree-optimization/85275
21030         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
21031         copying first exit test.
21033 2018-04-28  Mark Wielaard  <mark@klomp.org>
21035         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
21036         dwarf_version >= 5.
21037         (dwarf_AT): Handle DW_AT_addr_base.
21038         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
21040 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
21042         PR target/84431
21043         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
21044         (*ashl<dwi>3_doubleword_mask_1): Ditto.
21045         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
21046         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
21048 2018-04-28  Richard Biener  <rguenther@suse.de>
21050         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
21051         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
21052         to reflect use.  Only add interesting stmts.
21054 2018-04-27  Martin Jambor  <mjambor@suse.cz>
21056         PR ipa/85549
21057         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
21058         the jump function allows for passing through aggregate values.
21060 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
21062         * input.h (in_system_header_at): Convert from macro to inline
21063         function.
21064         (from_macro_expansion_at): Likewise.
21065         (from_macro_definition_at): Likewise.
21067 2018-04-27  Jeff Law  <law@redhat.com>
21069         * config.gcc: Mark tile* targets as deprecated/obsolete.
21071 2018-04-27  Richard Biener  <rguenther@suse.de>
21073         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
21074         fix for ILP32.
21076 2018-04-27  Richard Biener  <rguenther@suse.de>
21078         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
21080 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
21082         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
21083         with Yd constraint. Set "preferred_for_speed" attribute from
21084         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
21085         with Yd constraint.
21086         (*movdi_internal): Ditto.
21087         (movti_interunit splitters): Remove
21088         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
21089         (movdi_interunit splitters): Ditto.
21090         * config/i386/constraints.md (Ye): Remove.
21091         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
21093 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21095         PR target/85512
21096         * config/aarch64/constraints.md (Usg): Limit to 31.
21097         (Usj): Limit to 63.
21099 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
21101         PR tree-optimization/85529
21102         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
21103         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
21104         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
21105         zero extension or masking of the MSB bit.
21106         (optimize_range_tests): Add FIRST_BB argument, pass it through
21107         to optimize_range_tests_var_bound.
21108         (maybe_optimize_range_tests, reassociate_bb): Adjust
21109         optimize_range_tests callers.
21111 2018-04-26  Richard Biener  <rguenther@suse.de>
21112             Jakub Jelinek  <jakub@redhat.com>
21114         * cgraph.h (symbol_table): Just declare debug method here.
21115         * symtab.c (symbol_table::debug): Define.
21117 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
21119         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
21121 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
21123         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
21124         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
21125         (*movdi_internal): Substitute Yi and Yj constraint with x
21126         and Ym and Yn constraint with y constraint.  Update "isa"
21127         attribute and set "preferred_for_speed" attribute from
21128         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
21129         (*movsi_internal): Ditto.
21130         (*movdf_internal): Ditto.
21131         (*movsf_internal): Ditto.
21132         (*zero_extendsidi2): Ditto.
21133         * config/i386/sse.md (vec_set<mode>_0): Ditto.
21134         (sse2_loadld): Ditto.
21135         (*vec_extract<ssevecmodelower>_0): Ditto.
21136         (*vec_extractv4si_0_zext_sse4): Ditto.
21137         (vec_concatv2di): Ditto.
21138         (*vec_dup<mode>): Ditto.
21139         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
21140         * config/i386/constraints.md (Yi): Remove.
21141         (Yj): Remove.
21142         (Ym): Remove.
21143         (Yn): Remove.
21145 2018-04-26  Nathan Sidwell  <nathan@acm.org>
21147         * dumpfile.c (dump_open): New.
21148         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
21149         (dump_finish): Detect stdio/stderr by value not name.
21151 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
21153         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
21155 2018-04-26  Tom de Vries  <tom@codesourcery.com>
21157         PR target/84952
21158         * config/nvptx/nvptx.c (verify_neutering_jumps)
21159         (verify_neutering_labels): New function
21160         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
21162 2018-04-26  Tom de Vries  <tom@codesourcery.com>
21164         PR target/84025
21165         * config/nvptx/nvptx.c (needs_neutering_p): New function.
21166         (nvptx_single): Use needs_neutering_p to skip over insns that do not
21167         need neutering.
21169 2018-04-26  Richard Biener <rguenther@suse.de>
21170             Tom de Vries  <tom@codesourcery.com>
21172         PR lto/85422
21173         * lto-streamer-out.c (output_function): Fixup loops if required to match
21174         discovery done in the reader.
21176 2018-04-26  Richard Biener  <rguenther@suse.de>
21178         PR tree-optimization/85116
21179         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
21180         have a loop exit from the single latch predecessor.  Remove
21181         case of header with just condition.
21182         (ch_base::copy_headers): Exclude infinite loops from any
21183         processing.
21184         (pass_ch::execute): Record exits.
21186 2018-04-26  Richard Biener  <rguenther@suse.de>
21188         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
21189         prologue cost vector and pass it to vect_get_load_cost.
21190         (vect_get_peeling_costs_all_drs): Likewise.
21191         (vect_peeling_hash_get_lowest_cost): Likewise.
21192         (vect_enhance_data_refs_alignment): Likewise.
21194 2018-04-26  Richard Biener  <rguenther@suse.de>
21196         PR middle-end/85450
21197         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
21198         checking of integer<->pointer conversions.
21199         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
21200         sign-/zero-extending pointer types.
21201         (expand_omp_for_static_chunk): Likewise.
21203 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
21204             Jean Lee  <xiaoyur347@gmail.com>
21206         * config/mips/mips.c (mips_asan_shadow_offset): New function.
21207         (TARGET_ASAN_SHADOW_OFFSET): Define.
21208         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
21209         true for -fsanitize=address.
21211 2018-04-25  Mark Wielaard  <mark@klomp.org>
21213         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
21214         shorter ones.
21216 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
21218         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
21219         than "alu", remove explicit "memory" and "imm_disp" attributes.
21220         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
21222         PR middle-end/85414
21223         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
21224         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
21225         gen_lowpart_no_emit.
21227 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
21229         PR target/85473
21230         * config/i386/i386.c (ix86_expand_builtin): Change memory
21231         operand to XI, extend p0 to Pmode.
21232         * config/i386/i386.md: Change unspec volatile and operand
21233         1 mode to XI, change operand 0 mode to P.
21235 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21237         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
21238         GET_MODE_MASK before any checking.
21239         (nds32_can_use_bset_p): Likewise.
21240         (nds32_can_use_btgl_p): Likewise.
21242 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21244         * config/nds32/nds32-doubleword.md: New define_split pattern for
21245         illegal register number.
21247 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21249         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
21251 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21253         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
21255 2018-04-25  Richard Biener  <rguenther@suse.de>
21257         * lto-streamer.h (LTO_major_version): Bump to 8.
21259 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
21261         * BASE-VER: Set to 9.0.0.
21263 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
21265         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
21266         in __abskf2 and __powikf2.
21268 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21270         PR target/85512
21271         * config/aarch64/constraints.md (Usg, Usj): New constraints.
21272         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
21273         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
21274         Use the above on operand 2.  Reindent.
21275         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
21277 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
21279         PR target/85485
21280         * common/config/i386/i386-common.c (ix86_handle_option): Don't
21281         handle OPT_mcet.
21282         * config/i386/i386.opt (mcet): Removed.
21283         * doc/install.texi: Remove -mcet documentation.
21284         * doc/invoke.texi: Likewise.
21286 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
21288         PR target/85485
21289         * doc/install.texi: Remove -mcet from bootstrap-cet.
21291 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21293         PR target/85511
21294         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
21295         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
21296         if TARGET_64BIT.
21298         PR target/85503
21299         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
21300         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
21301         containing a CONST_VECTOR.
21303 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
21305         * doc/install.texi: Update newlib dependency for nvptx.
21307 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21309         PR target/85508
21310         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
21311         instead of INTVAL when shifting x left.
21313 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
21315         PR tree-optimization/85478
21316         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
21317         vect_grouped_store_supported for single element vectors.
21319 2018-04-24  Richard Biener  <rguenther@suse.de>
21321         PR target/85491
21322         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
21323         load cost increase to the case of non-constant step.
21325 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21327         PR target/84828
21328         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
21329         destination if any_malformed_asm.
21331 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
21333         PR middle-end/85496
21334         * expr.c (store_field): In the bitfield case, if the value comes from
21335         a function call and is returned in registers by means of a PARALLEL,
21336         do not change the mode of the temporary unless BLKmode and VOIDmode.
21338 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
21340         PR rtl-optimization/85423
21341         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
21342         dependencies to debug insns when the previous insn is non-debug.
21344 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21346         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
21347         enums into a single definition.
21348         (fls): Fix predicates and printing.
21349         (seti): Likewise.
21351 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21353         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
21354         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
21355         and short u6 immediate.
21356         (check_if_valid_sleep_operand): Remove.
21357         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
21359 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21361         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
21362         flag_always_save_lp condition.
21363         * config/nds32/nds32.opt (malways-save-lp): New option.
21365 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21367         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
21368         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
21369         * config/nds32/nds32.h
21370         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
21371         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
21373 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21375         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
21376         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
21378 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21379             Chung-Ju Wu  <jasonwucj@gmail.com>
21381         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
21382         Declare.
21383         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
21384         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
21386 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21388         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
21390 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21392         * config/nds32/nds32-protos.h (nds32_data_alignment,
21393         nds32_local_alignment): Declare.
21394         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
21395         nds32_local_alignment): New functions.
21396         (TARGET_CONSTANT_ALIGNMENT): Define.
21397         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
21399 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21401         * config/nds32/nds32.c
21402         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
21403         (TARGET_MODES_TIEABLE_P): Likewise.
21405 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21407         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
21408         level Ofast and Og.
21410 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
21411             Chung-Ju Wu  <jasonwucj@gmail.com>
21413         * config/nds32/constants.md (unspec_volatile_element): Add enum values
21414         for unaligned access.
21415         * config/nds32/nds32-intrinsic.c: Implementation of expanding
21416         unaligned access.
21417         * config/nds32/nds32-intrinsic.md: Likewise.
21418         * config/nds32/nds32_intrinsic.h: Likewise.
21419         * config/nds32/nds32.h (nds32_builtins): Likewise.
21420         * config/nds32/nds32.opt (munaligned-access): New option.
21421         * config/nds32/nds32.c (nds32_asm_file_start): Display
21422         flag_unaligned_access status.
21424 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
21426         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
21427         -mno-relax is present.
21428         * config/riscv/linux.h (LINK_SPEC): Ditto.
21430 2018-04-20  Martin Sebor  <msebor@redhat.com>
21432         PR c/85365
21433         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
21434         for null pointers.
21435         (gimple_fold_builtin_stxcpy_chk): Same.
21436         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
21438 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
21440         PR target/85456
21441         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
21442         __powikf2 when long double is IEEE 128-bit.
21444 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
21446         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
21447         step to make sure stack always aligned.
21449 2018-04-20  Carl Love  <cel@us.ibm.com>
21451         PR target/83402
21452         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
21453         size check for arg0.
21455 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
21456             Tom de Vries  <tom@codesourcery.com>
21458         PR target/85445
21459         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
21460         Emit insns for calls too.
21461         (nvptx_find_par): Always look for worker-level predecessor insn.
21462         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
21463         calls.
21464         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
21465         (nvptx_process_pars): Propagate frames for calls.
21467 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
21469         PR target/85469
21470         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
21471         Removed.
21472         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
21473         (ix86_handle_option): Don't handle OPT_mibt.
21474         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
21475         __SHSTK__.
21476         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
21477         has_ibt and ibt.
21478         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
21479         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
21480         (ix86_target_macros): Define __CET__ with flag_cf_protection
21481         for -fcf-protection.
21482         * config/i386/i386.c (isa2_opts): Remove -mibt.
21483         * config/i386/i386.h (TARGET_IBT): Removed.
21484         (TARGET_IBT_P): Likewise.
21485         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
21486         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
21487         * config/i386/i386.opt (mcet): Update help message.
21488         (mshstk): Likewise.
21489         (mibt): Removed.
21490         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
21491         -mcet as an alias for -mshstk.
21493 2018-04-20  Richard Biener <rguenther@suse.de>
21495         PR middle-end/85475
21496         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
21497         complexity by forcing a single use of the multiply operand.
21499 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21501         ipa/85449
21502         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
21503         recursion dependency to only apply to non-clones.
21505 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21507         ipa/85447
21508         * ipa-cp.c (create_specialized_node): Check that clones of
21509         self-recursive edges exist during IPA-CP.
21511 2018-04-19  Toon Moene  <toon@moene.org>
21513         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
21514         by -O3.
21516 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21518         PR tree-optimization/85467
21519         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
21520         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
21521         VECTOR_CST element to type.
21523 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21525         PR target/85397
21526         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
21527         * config/i386/i386.md (builtin_setjmp_setup): Removed.
21528         (builtin_longjmp): Likewise.
21529         (save_stack_nonlocal): New pattern.
21530         (restore_stack_nonlocal): Likewise.
21532 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21534         PR target/85404
21535         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21536         Replace ASM_OUTPUT_LABEL with fprintf.
21538 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21540         PR target/85417
21541         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21542         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
21543         * config/i386/i386-c.c (ix86_target_macros_internal): Also
21544         define __IBT__ and __SHSTK__ for -fcf-protection.
21545         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
21546         TARGET_IBT.
21547         (ix86_trampoline_init): Likewise.
21548         (x86_output_mi_thunk): Likewise.
21549         (ix86_notrack_prefixed_insn_p): Likewise.
21550         (ix86_option_override_internal): Don't disallow -fcf-protection.
21551         * config/i386/i386.md (rdssp<mode>): Also enable for
21552         -fcf-protection.
21553         (incssp<mode>): Likewise.
21554         (nop_endbr): Likewise.
21555         * config/i386/i386.opt (mcet): Change help message to built-in
21556         functions only.
21557         (mibt): Likewise.
21558         (mshstk): Likewise.
21559         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
21560         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
21561         enable CET built-in functions.
21563 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
21565         * common/config/i386/i386-common.c
21566         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
21567         OPTION_MASK_ISA_MOVDIRI_UNSET,
21568         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
21569         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
21570         * config.gcc (movdirintrin.h): New header.
21571         * config/i386/cpuid.h (bit_MOVDIRI,
21572         bit_MOVDIR64B): New bits.
21573         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
21574         and -mmvodir64b.
21575         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
21576         (VOID, PVOID, PCVOID)): New function types.
21577         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
21578         __builtin_ia32_directstoreu_u64,
21579         __builtin_ia32_movdir64b): New builtins.
21580         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
21581         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
21582         and -mmovdiri.
21583         (ix86_valid_target_attribute_inner_p): Ditto.
21584         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
21585         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
21586         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
21587         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
21588         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
21589         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
21590         (movdiri<mode>, movdir64b_<mode>): New.
21591         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
21592         * config/i386/immintrin.h: Include movdirintrin.h.
21593         * config/i386/movdirintrin.h: New file.
21594         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
21596 2018-04-19  Richard Biener  <rguenther@suse.de>
21598         PR middle-end/85455
21599         * cfg.c (clear_bb_flags): When loop state says we have
21600         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
21602 2018-04-19  Richard Biener  <rguenther@suse.de>
21604         PR tree-optimization/84737
21605         * tree-vect-data-refs.c (vect_copy_ref_info): New function
21606         copying restrict info.
21607         (vect_setup_realignment): Use it.
21608         * tree-vectorizer.h (vect_copy_ref_info): Declare.
21609         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
21610         the first DR to all generated stores.
21611         (vectorizable_load): Likewise for loads.
21613 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21615         PR tree-optimization/85446
21616         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
21617         the integral and pointer types to have the same precision.
21619         * doc/install.texi: Document --disable-cet being the default and
21620         --enable-cet=auto.
21622 2018-04-18  Martin Liska  <mliska@suse.cz>
21624         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
21625         style.
21627 2018-04-18  Martin Liska  <mliska@suse.cz>
21629         Revert
21630         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
21632         PR ipa/83983
21633         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21634         arguments if they are comparable.
21636 2018-04-18  Martin Liska  <mliska@suse.cz>
21638         Revert
21639         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21641         PR lto/84805
21642         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21643         incomplete types.
21645 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
21647         PR target/85388
21648         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
21649         ENDBR after calling __morestack.
21651 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
21653         PR jit/85384
21654         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
21655         by using gcc_base_ver to generate a gcc_driver_version, and use
21656         it when generating GCC_DRIVER_NAME.
21657         * configure: Regenerate.
21659 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
21661         PR target/81084
21662         * config.gcc: Obsolete powerpc*-*-*spe*.
21664 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21666         PR debug/84637
21667         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
21668         (stabstr_D): Change type of unum from unsigned int to
21669         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
21670         type.
21672 2018-04-17  Jim Wilson  <jimw@sifive.com>
21674         PR 84856
21675         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
21676         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
21677         Set arg_pointer_offset after using pretend_args_size.
21679 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21681         PR rtl-optimization/85431
21682         * dse.c (record_store): Ignore zero width stores.
21684         PR sanitizer/85230
21685         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
21686         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
21687         __builtin_stack_restore rather than after it.
21688         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
21689         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
21690         argument instead of virtual_dynamic_stack_rtx.
21692 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21694         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
21695         New prototype.
21696         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21697         Add note to error message to explain internal mapping of overloaded
21698         built-in function name to non-overloaded built-in function name.
21699         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
21700         function.
21702 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
21704         PR target/85424
21705         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
21706         where the inputs overlap with the output.
21708 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21710         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
21711         (=v, v) alternative and explicit "memory" attribute.
21712         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
21713         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21714         attributes.
21715         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
21716         "sselog1" type instead of "sselog".
21717         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
21718         "sselog".  Remove explicit "memory" attribute.
21719         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
21720         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21721         attributes.
21722         (vec_extract_hi_v32hi): Merge all alternatives into one, use
21723         "sselog1" type instead of "sselog".  Remove explicit "memory"
21724         attribute.
21725         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
21726         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21727         attribute.
21728         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
21729         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21730         attributes.
21731         (vec_extract_hi_v64qi): Merge all alternatives into one, use
21732         "sselog1" type instead of "sselog".  Remove explicit "memory"
21733         attribute.
21734         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
21735         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21736         attribute.
21738         PR target/85430
21739         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
21741         PR middle-end/85414
21742         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
21743         on a SUBREG.
21745 2018-04-17  Martin Jambor  <mjambor@suse.cz>
21747         PR ipa/85421
21748         * ipa-cp.c (create_specialized_node): Call
21749         expand_all_artificial_thunks if necessary.
21751 2018-04-17  Martin Liska  <mliska@suse.cz>
21753         PR lto/85405
21754         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
21755         in message, remote space in between '_G' and '('.
21757 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21759         PR target/85281
21760         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
21761         avx512f_vmcmp<mode>3<round_saeonly_name>,
21762         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
21763         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
21764         avx512f_rndscale<mode><round_saeonly_name>,
21765         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
21766         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
21767         Use %<iptr>2 instead of %2 for -masm=intel.
21768         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
21769         avx512f_vcvttss2usi<round_saeonly_name>,
21770         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
21771         -masm=intel.
21772         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
21773         avx512f_vcvttsd2usi<round_saeonly_name>,
21774         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
21775         Use %q1 instead of %1 for -masm=intel.
21776         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
21777         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
21778         of %3 for -masm=intel.
21779         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
21780         -masm=intel.
21781         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
21782         -masm=intel.
21783         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
21784         -masm=intel.
21785         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
21786         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
21787         %g1.
21788         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
21789         -masm=intel.
21790         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
21791         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
21792         %g1 and one with %0 and %1.
21793         (avx512er_vmrcp28<mode><round_saeonly_name>,
21794         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
21795         %1 for -masm=intel.
21796         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
21797         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
21798         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
21799         of %0 and %{%4%} for -masm=intel.
21800         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
21801         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
21802         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
21803         order of %0 and %{%5%}%{z%} for -masm=intel.
21805 2018-04-17  Jan Hubicka  <jh@suse.cz>
21807         PR lto/85405
21808         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
21810 2018-04-17  Martin Liska  <mliska@suse.cz>
21812         PR ipa/85329
21813         * multiple_target.c (create_dispatcher_calls): Set apostrophes
21814         for target_clone error message.  Make default implementation
21815         clone to be a local declaration.
21816         (separate_attrs): Add new argument and check for an empty
21817         string.
21818         (expand_target_clones): Handle it.
21819         (ipa_target_clone): Make redirection just for target_clones
21820         functions.
21822 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
21823             Tom de Vries  <tom@codesourcery.com>
21825         PR middle-end/84955
21826         * omp-expand.c (expand_oacc_for): Add dummy false branch for
21827         tiled basic blocks without omp continue statements.
21829 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
21831         PR target/83660
21832         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
21833         vec_extract expression as having side effects to make sure it gets
21834         a cleanup point.
21836 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
21838         PR target/85403
21839         * config/i386/i386.c (get_builtin_code_for_version): Check
21840         error_mark_node.
21842 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
21844         PR target/84331
21845         * config.gcc: Support "skylake".
21846         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21847         PROCESSOR_SKYLAKE.
21848         * config/i386/i386.c (m_SKYLAKE): Define.
21849         (processor_target_table): Add "skylake".
21850         (ix86_option_override_internal): Add "skylake".
21851         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
21852         PROCESSOR_CANNONLAKE.
21853         (get_builtin_code_for_version): Fix priority for
21854         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
21855         PROCESSOR_SKYLAKE-AVX512.
21856         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
21857         (processor_type): Add PROCESSOR_SKYLAKE.
21859 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
21860             Jason Merrill  <jason@redhat.com>
21862         PR c++/85112
21863         * convert.c (convert_to_integer_1): Use direct recursion for
21864         enumeral types and types with a precision less than the number
21865         of bits in their mode.
21867 2018-04-16  Julia Koval  <julia.koval@intel.com>
21869         PR target/84413
21870         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
21871         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
21873 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
21875         PR target/85293
21876         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
21877         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
21878         and -mno-direct-move.
21880 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
21882         PR target/83402
21883         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
21884         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
21885         Ensure negative shifts result in {0}.
21887 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
21889         PR rtl-optimization/79916
21890         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
21891         regs (if any) to define how to gnerate SD moves when LRA is in
21892         progress.
21894 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
21896         PR rtl-optimization/85393
21897         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
21898         * except.c (expand_dw2_landing_pad_for_region): Make static.
21899         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
21900         a label and unconditional jump to old_bb, rather than
21901         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
21902         basic block.
21904         PR rtl-optimization/85376
21905         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
21906         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
21907         instead of a specific value.
21909 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21910             Bin Cheng  <bin.cheng@arm.com>
21912         PR tree-optimization/82965
21913         PR tree-optimization/83991
21914         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
21915         by_profile_only parameter.
21916         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
21917         information if the loop was predicted to iterate too many times.
21918         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
21920 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21922         PR lto/71991
21923         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
21924         always inline.
21926 2018-04-13  Martin Liska  <mliska@suse.cz>
21927             Jakub Jelinek  <jakub@redhat.com>
21929         PR middle-end/81657
21930         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
21931         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
21932         * builtins.c (expand_builtin_memory_copy_args): Use
21933         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
21934         handle dest_addr == pc_rtx.
21936 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
21938         PR target/85291
21939         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
21940         asked to not generate direct moves.
21941         (fix_trunc<mode>si2_stfiwx): Similar.
21942         (fix_trunc<mode>si2_internal): Similar.
21944 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21946         PR debug/83157
21947         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
21948         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
21949         lookup if dest in some wider mode is known to be const0_rtx and
21950         if so, record permanent equivalence for it to be ZERO_EXTEND of
21951         the narrower mode destination.
21953 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21955         * lto-streamer-out.c (output_function): Revert 259346.
21956         * omp-expand.c (expand_oacc_for): Likewise.
21958 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
21960         PR rtl-optimization/85354
21961         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
21962         * sel-sched.c (sel_global_init): ... here.
21964 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
21966         PR target/85238
21967         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
21968         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
21969         mode for PE-COFF targets.
21970         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
21971         (i386_pe_asm_lto_end): Likewise.
21972         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
21973         (TARGET_ASM_LTO_END): Likewise.
21974         * config/i386/winnt.c (saved_debug_info_level): New static variable.
21975         (i386_pe_asm_lto_start): New function.
21976         (i386_pe_asm_lto_end): Likewise.
21978 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21979             Richard Biener  <rguenther@suse.de>
21981         PR middle-end/84955
21982         * lto-streamer-out.c (output_function): Fix CFG loop state before
21983         streaming out.
21984         * omp-expand.c (expand_oacc_for): Handle calls to internal
21985         functions like regular functions.
21987 2018-04-12  Richard Biener  <rguenther@suse.de>
21989         PR lto/85371
21990         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
21991         for the early LTO debug to properly generate references to it
21992         during DIE emission.  Do not re-use that for the skeleton for
21993         split-dwarf.
21994         (dwarf2out_early_finish): Likewise.
21996 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21998         PR target/85328
21999         * config/i386/sse.md
22000         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
22001         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
22002         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
22003         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
22004         and output is a reg, avoid creating invalid lowpart subreg, but
22005         instead split into a 512-bit move.  Don't split if not AVX512VL,
22006         input is xmm16+ reg and output is a mem.
22007         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
22008         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
22009         xmm16+ reg and output is a mem.
22011 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22013         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
22014         also for flag_dwarf2_cfi_asm.
22016 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
22018         PR rtl-optimization/85342
22019         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
22020         a bool scalar var inside of the loop instead.  Don't try to update
22021         recog_data.operand after failed apply_change_group.
22023 2018-04-12  Tom de Vries  <tom@codesourcery.com>
22025         PR target/85296
22026         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
22027         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
22028         array with flexible array member as array without given dimension.
22029         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
22030         argument for undefined param to true.
22032 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
22034         PR target/85321
22035         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
22036         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
22037         from PowerPC section.
22038         * config/rs6000/sysv4.opt (mcall-): Improve help text.
22039         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
22040         help text that is too long.
22041         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
22042         help text that is too long.
22043         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
22044         help text that is too long.
22046 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
22048         * config/alpha/alpha.md (stack_probe_internal): Rename
22049         from "probe_stack".  Update all callers.
22051 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
22053         PR rtl-optimization/84566
22054         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
22055         sched_macro_fuse_insns.
22057 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
22059         PR target/84301
22060         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
22061         (compute_block_dependences): ... from here.
22063 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22065         PR tree-optimization/85331
22066         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
22067         from int to HOST_WIDE_INT.
22069 2018-04-11  Martin Jambor  <mjambor@suse.cz>
22071         PR ipa/84149
22072         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
22073         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
22074         not the same as the source val.
22075         (cgraph_edge_brings_value_p): New parameter.
22076         (gather_edges_for_value): Pass destination value to
22077         cgraph_edge_brings_value_p.
22078         (perhaps_add_new_callers): Likewise.
22079         (get_info_about_necessary_edges): Likewise and exclude values brought
22080         only by self-recursive edges.
22081         (create_specialized_node): Redirect only clones of self-calling edges.
22082         (+self_recursive_pass_through_p): New function.
22083         (find_more_scalar_values_for_callers_subset): Use it.
22084         (find_aggregate_values_for_callers_subset): Likewise.
22085         (known_aggs_to_agg_replacement_list): Removed.
22086         (decide_whether_version_node): Re-calculate known constants for all
22087         remaining context clones.
22089 2018-04-11  Richard Biener  <rguenther@suse.de>
22091         PR lto/85339
22092         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
22093         from early DWARF output.
22094         (dwarf2out_early_finish): Output line info unconditionally into
22095         early DWARF and add reference to it.
22097 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22099         PR target/85281
22100         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
22101         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
22102         other than V2DFmode using iptr mode attribute.
22103         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
22105 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
22107         PR rtl-optimization/84659
22108         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
22110 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22112         PR debug/85302
22113         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
22114         SIZEP is NULL.
22115         (output_loc_list): Pass address of a dummy size variable even in the
22116         locview handling loop.
22117         (index_location_lists): Add comment on why skip_loc_list_entry can't
22118         call size_of_locs.
22120 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22122         PR target/85261
22123         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
22124         into register.
22126 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
22128         PR target/85321
22129         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
22130         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
22131         and -mstring-compare-inline-limit.
22133 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22135         PR target/85287
22136         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
22137         for stack clash protection in a register whenever we need it to be in
22138         a register.
22140 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22142         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
22143         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
22145 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22147         PR target/85321
22148         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
22149         the help text.
22150         (mlong-double-): Ditto.
22151         * config/rs6000/sysv4.opt (msdata=): Ditto.
22152         (mtls-size=): Ditto.
22154 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22156         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
22157         erroneous entries for
22158         "vector int vec_ldl (int, long int *)", and
22159         "vector unsigned int vec_ldl (int, unsigned long int *)".
22160         Add comments and entries for
22161         "vector bool char vec_ldl (int, bool char *)",
22162         "vector bool short vec_ldl (int, bool short *)",
22163         "vector bool int vec_ldl (int, bool int *)",
22164         "vector bool long long vec_ldl (int, bool long long *)",
22165         "vector pixel vec_ldl (int, pixel *)",
22166         "vector long long vec_ldl (int, long long *)",
22167         "vector unsigned long long vec_ldl (int, unsigned long long *)".
22168         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
22169         type tree bool_long_long_type_node and correct definition of
22170         bool_V2DI_type_node to make reference to this new type tree.
22171         (rs6000_mangle_type): Replace erroneous reference to
22172         bool_long_type_node with bool_long_long_type_node.
22173         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
22174         comments to emphasize sign distinctions for char and int types and
22175         replace RS6000_BTI_bool_long constant with
22176         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
22177         use of RS6000_BTI_pixel.
22178         (bool_long_type_node): Remove this macro definition.
22179         (bool_long_long_type_node): New macro definition
22181 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
22183         PR rtl-optimization/85300
22184         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
22185         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
22186         simplify_unary_operation fails.
22188 2018-04-10  Martin Liska  <mliska@suse.cz>
22190         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
22191         cgraph_edge and ipa_ref.
22193 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
22195         PR target/85177
22196         PR target/85255
22197         * config/i386/sse.md
22198         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
22199         computation of the VEC_MERGE selector from mask.
22200         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
22201         Fix decoding of the VEC_MERGE selector into mask.
22203 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
22205         PR tree-optimization/85286
22206         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
22208 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
22210         * final.c (final_1): Set insn_last_address as well as
22211         insn_current_address.
22213 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22215         PR target/85173
22216         * explow.c (emit_stack_probe): Call validize_mem on memory location
22217         before passing it to gen_probe_stack.  Create address operand and
22218         legitimize it for the probe_stack_address case.
22220 2018-04-09  Jan Hubicka  <jh@suse.cz>
22222         PR lto/85078
22223         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
22224         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
22225         * tree.c (free_lang_data_in_type): Fix handling of binfos;
22226         walk basetypes.
22227         (free_lang_data): Rebuild type inheritance graph.
22229 2018-04-09  Martin Sebor  <msebor@redhat.com>
22231         * invoke.texi (-finline-small-functions): Mention other optimization
22232         options.
22233         (-findirect-inlining, -fpartial-inlining): Same.
22234         (-finline-functions-called-once): Same.
22235         (-freorder-blocks-and-partition): Same.
22237 2018-04-09  Jan Hubicka  <jh@suse.cz>
22239         PR rtl/84058
22240         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
22241         jumps; choose last target that matches the criteria (i.e.
22242         no partition changes for non-crossing jumps).
22243         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
22244         support for redirecting crossing jumps to non-crossing.
22246 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
22248         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
22249         also for naked functions.
22251 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
22253         * config/arc/arc.md (add_shift): New pattern.
22254         (add_shift2): Likewise.
22255         (sub_shift): Likewise.
22256         (sub_shift_cmp0_noout): Likewise.
22257         (compare_si_ashiftsi): Likewise.
22258         (xbfu_cmp0_noout): New combine pattern.
22259         (xbfu_cmp0"): Likewise.
22260         (movsi_set_cc_insn): Place the predicable variant first.
22261         (commutative_binary_cmp0_noout): Remove clobber.
22262         (commutative_binary_cmp0): New pattern.
22263         (noncommutative_binary_cmp0): Likewise.
22264         (noncommutative_binary_cmp0_noout): Likewise.
22265         (noncommutative_binary_comparison_result_used): Removed.
22266         (rsub_cmp0): New pattern.
22267         (rsub_cmp0_noout): Likewise.
22268         (extzvsi): Changed, keep only meaningful variants.
22269         (SQH, SEZ): New iterators.
22270         (SQH_postfix): New mode attribute.
22271         (SEZ_prefix): New code attribute.
22272         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
22273         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
22274         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
22275         of numerical value.
22276         (noncommutative_operator): Check the availability of barrel
22277         shifter option.
22279 2018-04-09  Richard Biener  <rguenther@suse.de>
22281         PR tree-optimization/85284
22282         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22283         Only use the niter constraining form of simple_iv when the exit
22284         is always executed.
22286 2018-04-09  Tom de Vries  <tom@codesourcery.com>
22288         PR target/84041
22289         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
22290         (define_expand "*memory_barrier"): New define_expand.
22291         (define_insn "memory_barrier"): New insn.
22293 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22295         PR rtl-optimization/80463
22296         PR rtl-optimization/83972
22297         PR rtl-optimization/83480
22299         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
22300         correct producer for the insn.
22301         (tidy_control_flow): Fixup seqnos in case of debug insns.
22303 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22305         PR rtl-optimization/83913
22307         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
22308         different sched-times when merging exprs.
22310 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22312         PR rtl-optimization/83962
22314         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
22315         tidy_fallthru_edge and tidy_control_flow.
22317 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22319         PR rtl-optimization/83530
22321         * sel-sched.c (force_next_insn): New global variable.
22322         (remove_insn_for_debug): When force_next_insn is true, also leave only
22323         next insn in the ready list.
22324         (sel_sched_region): When the region wasn't scheduled, make another pass
22325         over it with force_next_insn set to 1.
22327 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
22329         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
22330         into tm_file.
22331         * config/nds32/constants.md (unspec_volatile_element): Add enum values
22332         for interrupt control.
22333         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22334         functions for interrupt control.
22335         * config/nds32/nds32-intrinsic.md: Likewise.
22336         * config/nds32/nds32_intrinsic.h: Likewise.
22337         * config/nds32/nds32.h (nds32_builtins): Likewise.
22339 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
22341         * config/nds32/nds32.c (nds32_init_machine_status,
22342         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
22343         strict_aligned_p field.
22344         (nds32_expand_to_rtl_hook): New function.
22345         (TARGET_EXPAND_TO_RTL_HOOK): Define.
22346         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
22348 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22349             Chung-Ju Wu  <jasonwucj@gmail.com>
22351         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
22352         * config/nds32/nds32-n7.md: New file.
22353         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
22354         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
22355         pipeline.
22356         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
22357         * config/nds32/nds32.md (pipeline_model): Add n7.
22358         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
22359         * config/nds32/pipelines.md: Include n7 settings.
22361 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22362             Chung-Ju Wu  <jasonwucj@gmail.com>
22364         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
22365         * config/nds32/nds32-e8.md: New file.
22366         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
22367         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
22368         pipeline.
22369         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
22370         * config/nds32/nds32.md (pipeline_model): Add e8.
22371         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
22372         * config/nds32/pipelines.md: Include e8 settings.
22374 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22375             Chung-Ju Wu  <jasonwucj@gmail.com>
22377         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
22378         * config/nds32/nds32-n8.md: New file.
22379         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
22380         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
22381         pipeline.
22382         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
22383         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
22384         * config/nds32/nds32.md (pipeline_model): Add n8.
22385         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
22386         * config/nds32/pipelines.md: Include n8 settings.
22388 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22389             Chung-Ju Wu  <jasonwucj@gmail.com>
22391         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
22392         * config/nds32/nds32-n9-2r1w.md: New file.
22393         * config/nds32/nds32-n9-3r2w.md: New file.
22394         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
22395         nds32_register_ports): New or modify for cpu n9.
22396         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
22397         pipeline.
22398         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
22399         * config/nds32/nds32-utils.c: New file.
22400         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
22401         TARGET_MUL_SLOW): Define.
22402         * config/nds32/nds32.md (pipeline_model): New attribute.
22403         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
22404         New options that support cpu n9.
22405         * config/nds32/pipelines.md: Include n9 settings.
22406         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
22408 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
22410         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
22411         information if necessary.
22412         (output_cond_branch_compare_zero): Likewise.
22413         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
22414         (nds32_target_alignment): Refine for alignment.
22415         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
22416         (FUNCTION_BOUNDARY): Modify.
22417         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
22418         align case.
22419         * config/nds32/nds32.opt (malways-align, malign-functions): New.
22421 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
22423         * config/nds32/constants.md (unspec_volatile_element): Add values for
22424         TLB operation and data prefetch.
22425         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22426         functions for TLB operation and data prefetch.
22427         * config/nds32/nds32-intrinsic.md: Likewise.
22428         * config/nds32/nds32_intrinsic.h: Likewise.
22429         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
22430         (nds32_print_operand): Likewise.
22431         * config/nds32/nds32.h (nds32_builtins): Likewise.
22433 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22434         Andrew Pinski <pinsika@gcc.gnu.org>
22436         PR middle-end/82976
22437         * match.pd: Use constant_boolean_node of correct type instead of
22438         boolean_true_node or boolean_false_node for simplifying
22439         pointer comparisons to zero.
22441 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
22443         PR tree-optimization/80021
22444         * tree.c (verify_type_variant): Make error call in verify_variant_match
22445         translatable and remove final full stop.
22447 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22449         * config/nds32/constants.md (unspec_volatile_element): Add
22450         UNSPEC_VOLATILE_EH_RETURN.
22451         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
22452         nds32_output_stack_pop): Support dwarf exception handling process.
22453         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
22454         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
22455         exception handling process.
22456         (nds32_compute_stack_frame): Likewise.
22457         (nds32_return_addr_rtx): Likewise.
22458         (nds32_initial_elimination_offset): Likewise.
22459         (nds32_expand_prologue): Likewise.
22460         (nds32_expand_epilogue): Likewise.
22461         (nds32_dynamic_chain_address): New function.
22462         * config/nds32/nds32.h (machine_function): Add fields for dwarf
22463         exception handling.
22464         (DYNAMIC_CHAIN_ADDRESS): Define.
22465         (EH_RETURN_DATA_REGNO): Define.
22466         (EH_RETURN_STACKADJ_RTX): Define.
22467         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
22468         patterns for dwarf exception handling.
22470 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22472         * config/nds32/nds32.h: Clean up obsolete macros.
22474 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22476         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22477         Add enum values for particular instructions.
22478         * config/nds32/nds32-intrinsic.c: Implementation of expanding
22479         particular intrinsic functions.
22480         * config/nds32/nds32-intrinsic.md: Likewise.
22481         * config/nds32/nds32_intrinsic.h: Likewise.
22482         * config/nds32/nds32.h (nds32_builtins): Likewise.
22483         * config/nds32/nds32.md (type): Add pbsad and pbsada.
22484         (btst, ave): New patterns for particular instructions.
22486 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22488         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22489         Add enum values for atomic load/store and memory sync.
22490         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
22491         and memory sync.
22492         * config/nds32/nds32-intrinsic.md: Likewise.
22493         * config/nds32/nds32_intrinsic.h: Likewise.
22494         * config/nds32/nds32.h (nds32_builtins): Likewise.
22496 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
22498         PR tree-optimization/85257
22499         * fold-const.c (native_encode_vector): If not all elts could fit
22500         and off is -1, return 0 rather than offset.
22501         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
22502         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
22503         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
22504         adjust buffer in native_interpret_expr call.
22506 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22508         * config/nds32/constants.md (unspec_volatile_element): Add cache
22509         control enum values.
22510         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
22511         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
22512         * config/nds32/nds32.c (nds32_cctl_names): New.
22513         (nds32_print_operand): Handle cache control register names.
22514         * config/nds32/nds32.h (nds32_builtins): New enum values.
22515         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
22516         macros.
22517         * config/nds32/nds32.md (type): Add mmu.
22518         * config/nds32/pipelines.md (simple_insn): Add mmu.
22520 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22522         * config/nds32/nds32.md (type): Remove call.
22523         * config/nds32/pipelines.md (simple_insn): Likewise.
22525 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22527         * config/nds32/constants.md (unspec_volatile_element): Add
22528         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
22529         UNSPEC_VOLATILE_FMFCFG.
22530         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
22531         description for fmfcfg and fmfcsr.
22532         (bdesc_1arg): Add fmtcsr.
22533         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
22534         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
22535         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
22536         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
22537         unspec_fmfcfg): New patterns.
22538         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
22539         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
22540         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
22541         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
22542         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
22543         __nds32__fmfcfg): Define.
22545 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22547         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
22548         intrinsic register names.
22549         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
22550         intrinsic register enum values and macros.
22552 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22554         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
22555         for load/store addressing form.
22556         (nds32_print_operand_address): Likewise.
22558 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
22560         PR target/85196
22561         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
22562         based on LABEL_REF.  Remove useless assertion.
22563         (pic_address_needs_scratch): Fix formatting.
22564         (sparc_legitimize_pic_address): Minor tweaks.
22565         (sparc_delegitimize_address): Adjust assertion accordingly.
22566         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
22567         into symbolic_operand.
22568         (movsi_high_pic_label_ref): Likewise.
22569         (movsi_lo_sum_pic_label_ref): Likewise.
22570         (movdi_pic_label_ref): Likewise.
22571         (movdi_high_pic_label_ref): Likewise.
22572         (movdi_lo_sum_pic_label_ref): Likewise.
22574 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
22576         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
22577         custom LIB_SPEC setup.
22579 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
22580             Kito Cheng  <kito.cheng@gmail.com>
22582         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
22583         * config/riscv/freebsd.h: New.
22585 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22587         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
22588         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
22589         file.
22591 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22592             Kito Cheng  <kito.cheng@gmail.com>
22594         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
22595         nds32_output_call, nds32_symbol_binds_local_p): New functions.
22596         * config/nds32/nds32-protos.h (nds32_output_call,
22597         nds32_output_return): Declare.
22598         * config/nds32/nds32.md: Refine all the call and return patterns.
22600 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22602         PR debug/85252
22603         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
22604         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
22606         PR rtl-optimization/84872
22607         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
22608         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
22609         EDGE_CROSSING edge.
22611 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
22613         * expr.c (copy_blkmode_to_reg): Revert 254862.
22614         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
22616 2018-04-06  Richard Biener  <rguenther@suse.de>
22618         PR middle-end/85244
22619         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
22620         after seeing a component reference with an adjacent field.  Treat
22621         refs to arrays at struct end of external decls similar to
22622         refs to unconstrained commons.
22624 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22626         PR sanitizer/85213
22627         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
22628         look through SAVE_EXPRs with non-side-effects argument.  Adjust
22629         recursive calls.
22630         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
22631         save_p here.
22633 2018-04-06  Richard Biener  <rguenther@suse.de>
22635         PR middle-end/85180
22636         * alias.c (find_base_term): New wrapper around find_base_term
22637         unwinding CSELIB_VAL_PTR changes.
22638         (find_base_term): Do not restore CSELIB_VAL_PTR during the
22639         recursion.
22641 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22643         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
22644         instructions.
22645         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
22646         constant definitions.
22647         ("nop"): lr 0,0 -> nopr r0
22648         ("nop_lr0", "nop_lr1"): New insn definitions.
22650 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22652         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
22653         NDS32_V3PUSH_AVAILABLE_P macro.
22655 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
22656             Chung-Ju Wu  <jasonwucj@gmail.com>
22658         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
22659         (nds32*-*-*): Add float and fpu_config into supported_defaults.
22660         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
22661         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
22662         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
22663         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
22664         * config/nds32/constraints.md: New constraints and checking for hard
22665         float configuration.
22666         * config/nds32/iterators.md: New mode iterator and attribute for hard
22667         float configuration.
22668         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
22669         patterns.
22670         * config/nds32/nds32-fpu.md: New file.
22671         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
22672         deal with hard float code generation.
22673         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
22674         ARCH_V3S.
22675         (abi_type, float_reg_number): New enum type.
22676         * config/nds32/nds32-predicates.c: New predicates for hard float.
22677         * config/nds32/nds32-protos.h: Declare functions for hard float.
22678         * config/nds32/nds32.c: Implementation for hard float configuration.
22679         * config/nds32/nds32.h: Definitions for hard float configuration.
22680         * config/nds32/nds32.md: Include hard float machine description and
22681         modify patterns for hard float configuration.
22682         * config/nds32/nds32.opt: New options for hard float configuration.
22683         * config/nds32/predicates.md: New predicates for hard float
22684         configuration.
22686 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22688         * common/config/nds32/nds32-common.c
22689         (nds32_option_optimization_table): Enable -mreleax-hint by default.
22691 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
22693         PR middle-end/85195
22694         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
22695         CONSTRUCTOR_ELT (ctor, ...)->value.
22697 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
22699         PR target/85193
22700         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
22702 2018-04-05  Tom de Vries  <tom@codesourcery.com>
22704         PR target/85204
22705         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
22706         cond jump.
22708 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22709             Kito Cheng  <kito.cheng@gmail.com>
22711         * config/nds32/constraints.md (U33): Fine-tune checking condition.
22712         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
22713         * config/nds32/nds32.h (nds32_16bit_address_type): Add
22714         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
22716 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22717             Kito Cheng  <kito.cheng@gmail.com>
22719         * config/nds32/constraints.md (Ufe): New memory constraint.
22720         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
22721         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
22722         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
22723         operands.
22724         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
22725         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
22727 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22729         * config/nds32/nds32.md: Use optimize_size in the condition for
22730         alu-shift instructions.
22732 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22734         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
22736 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22738         * config/nds32/nds32.md (negsi2): Refine pattern.
22740 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22741             Chung-Ju Wu  <jasonwucj@gmail.com>
22743         * config/nds32/iterators.md (shift_rotate): New code iterator.
22744         (shift): New code attribute.
22745         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
22746         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
22747         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
22748         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
22749         bit-wise operations.
22750         (andsi3, *andsi3): Ditto.
22751         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
22752         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
22753         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
22754         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
22755         nds32_ior_operand, nds32_xor_operand): New predicates.
22757 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22759         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
22760         (addsi3, subsi3): ... this.
22762 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22764         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
22766 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22768         * config/nds32/nds32.md: Adjust indention.
22770 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22772         * config/nds32/nds32.md (feature): New attribute.
22774 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22776         * config/nds32/nds32.md (subtype): New attribute.
22778 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22780         PR target/85203
22781         * config/arm/arm-builtins.c (arm_expand_builtin): Change
22782         expansion to perform a bitwise AND of the argument followed by a
22783         boolean negation of the result.
22785 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
22787         PR rtl-optimization/84878
22788         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
22789         the basic block.  Assert the use reference is not artificial and that
22790         it has an associated insn.
22792 2018-04-04  Michael Matz  <matz@suse.de>
22794         * builtins.c (compute_objsize): Pass correct operand
22795         to array_at_struct_end_p.
22797 2018-04-04  Richard Biener  <rguenther@suse.de>
22799         PR lto/85176
22800         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
22801         from contexts for DINFO_LEVEL_TERSE and below.
22803 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22805         * config/nds32/nds32-doubleword.md (move_<mode>): Require
22806         resiter_operand condition.
22807         * config/nds32/nds32.md (*move<mode>): Ditto.
22809 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22810             Monk Chiang  <sh.chiang04@gmail.com>
22812         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
22814 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22816         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
22818 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22819             Kito Cheng  <kito.cheng@gmail.com>
22821         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
22822         nds32_cond_code_str, output_cond_branch,
22823         output_cond_branch_compare_zero, nds32_expand_cbranch,
22824         nds32_expand_cstore, nds32_expand_movcc,
22825         nds32_output_cbranchsi4_equality_zero,
22826         nds32_output_cbranchsi4_equality_reg,
22827         nds32_output_cbranchsi4_equality_reg_or_const_int,
22828         nds32_output_cbranchsi4_greater_less_zero: New functions.
22829         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
22830         nds32_expand_cstore, nds32_expand_movcc,
22831         nds32_output_cbranchsi4_equality_zero,
22832         nds32_output_cbranchsi4_equality_reg,
22833         nds32_output_cbranchsi4_equality_reg_or_const_int,
22834         nds32_output_cbranchsi4_greater_less_zero): Declare.
22835         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
22836         nds32_rimm11s_operand): New predicates.
22837         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
22838         * config/nds32/nds32.md: Rewrite all the branch and conditional move
22839         patterns.
22841 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22843         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
22844         * config/nds32/nds32.md: Ditto.
22845         * config/nds32/pipelines.md: Ditto.
22847 2018-04-04  Richard Biener  <rguenther@suse.de>
22849         PR tree-optimization/85168
22850         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
22851         propagating abnormals.
22853 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22855         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
22857 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22858             Kito Cheng  <kito.cheng@gmail.com>
22860         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
22861         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
22862         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
22863         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
22864         * config/nds32/nds32.md (sibcall_internal): New.
22865         (sibcall_register): Remove.
22866         (sibcall_immediate): Remove.
22867         (sibcall_value_internal): New.
22868         (sibcall_value_register): Remove.
22869         (sibcall_value_immediate): Remove.
22870         * config/nds32/predicates.md (nds32_general_register_operand): New.
22871         (nds32_call_address_operand): New.
22873 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22875         PR rtl-optimization/85167
22876         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
22877         bb_defs if *split_p, instead preinitialize it to NULL.
22879         PR tree-optimization/85156
22880         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
22881         evaluating the argument multiple times.
22883 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
22885         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
22886         than vector.
22887         (_mm_cvtpd_ps): Likewise.
22888         (_mm_cvttpd_epi32): Likewise.
22889         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
22890         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
22891         vector, pixel, and bool following altivec.h include.
22893 2018-04-03  Martin Sebor  <msebor@redhat.com>
22895         * doc/extend.texi (Common Function Attributes): Clarify.
22896         (const attribute): Likewise.
22897         (pure attribute): Likewise.
22899 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22901         PR target/85169
22902         * config/i386/i386.c (ix86_expand_vector_set): Use
22903         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
22905 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
22907         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
22908         instructions when changing rounding bits to preserve precision bits
22909         in the x87 control word.
22911 2018-04-03  Martin Liska  <mliska@suse.cz>
22913         PR tree-optimization/82491
22914         * rtl.h (strip_offset_and_add): Replace += suboffset with
22915         poly_uint64 () + suboffset.
22917 2018-03-29  Martin Liska  <mliska@suse.cz>
22918             Martin Jambor  <mjambor@suse.cz>
22920         PR ipa/84947
22921         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
22922         param_type is not an integral or pointer type.
22924 2018-04-03  Richard Biener  <rguenther@suse.de>
22926         * sese.h (recompute_all_dominators): Remove.
22928 2018-04-02  Martin Sebor  <msebor@redhat.com>
22930         * doc/invoke.texi (-Wrestrict): Fix typos.
22932 2018-04-02  Jim Wilson  <jimw@sifive.com>
22934         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
22935         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
22936         (<optab>di3, <optab>si3_extend): Likewise.
22937         (<optab>si3_mask, <optab>si3_mask_1): New.
22938         (<optab>di3_mask, <optab>di3_mask_1): New.
22939         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
22940         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
22941         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
22943 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
22945         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
22946         example.
22948 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
22950         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
22951         (nds32_canonicalize_comparison): New function.
22953 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22954             Kito Cheng  <kito.cheng@gmail.com>
22955             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22957         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
22958         * config/nds32/constants.md (unspec_volatile_element): Add
22959         UNSPEC_VOLATILE_RELAX_GROUP.
22960         * config/nds32/nds32-relax-opt.c: New file.
22961         * config/nds32/nds32-predicates.c
22962         (nds32_symbol_load_store_p): New function.
22963         * config/nds32/nds32-protos.h
22964         (nds32_symbol_load_store_p): Declare function.
22965         (make_pass_nds32_relax_opt): Declare new rtl pass function.
22966         * config/nds32/nds32.c
22967         (nds32_register_pass): New function to register pass.
22968         (nds32_register_passes): New function to register passes.
22969         * config/nds32/nds32.md (relax_group): New pattern.
22970         * config/nds32/nds32.opt (mrelax-hint): New option.
22971         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
22973 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
22975         * config/nds32/t-nds32: Modify files dependency.
22977 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22979         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
22980         (PROFILE_HOOK): Define its implementation.
22982 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22984         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
22985         type and 32-bit size.
22987 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
22989         PR middle-end/85090
22990         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
22991         (V_128_256): New mode iterator.
22992         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
22993         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
22994         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
22995         of V.
22996         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
22997         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
22999 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
23001         PR target/83315
23002         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
23003         NaN inputs correctly.
23005 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
23007         PR target/80546
23008         * config/rs6000/vsx.md (??r): New mode attribute.
23009         (*vsx_mov<mode>_64bit): Use it.
23010         (*vsx_mov<mode>_32bit): Likewise.
23012 2018-03-30  Martin Sebor  <msebor@redhat.com>
23014         PR tree-optimization/84818
23015         * builtins.c (check_access): Use warning_n.
23017 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23019         PR target/83822
23020         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
23021         condition.
23022         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
23023         condition.
23025 2018-03-30  Julia Koval  <julia.koval@intel.com>
23027         PR target/84413
23028         * x86-tune.def (movx, partial_reg_dependency): Enable for
23029         m_SKYLAKE_AVX512.
23031 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
23033         PR inline-asm/84985
23034         * lra-constraints.c (process_alt_operands): Move setting
23035         this_alternative_matches below.
23037 2018-03-29  Martin Liska  <mliska@suse.cz>
23039         PR lto/84995.
23040         * doc/invoke.texi: Document how LTO works with debug info.
23041         Describe auto-load support of binutils.  Mention 'x86-64'
23042         as valid option value of -march option.
23044 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
23046         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
23048         PR c/85094
23049         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
23050         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
23051         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
23052         checking.
23054 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
23056         PR target/84912
23057         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
23058         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
23059         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
23060         for RS6000_BTM_POWERPC64.
23061         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
23062         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
23063         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
23064         definition.
23065         (DIVDE): Use it.
23066         (DIVDEU): Likewise.
23068 2018-03-28  Carl Love  <cel@us.ibm.com>
23070         Revert
23071         2017-09-27  Carl Love  <cel@us.ibm.com>
23073         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
23074         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
23075         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
23076         fctiw instruction.
23078 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23080         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
23081         instead of __vector bool.
23082         (_mm_max_pu8): Likewise.
23083         (_mm_min_pi16): Likewise.
23085 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
23087         PR target/84912
23088         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
23089         (DIVWEUO): Likewise.
23090         (DIVDEO): Likewise.
23091         (DIVDEUO): Likewise.
23092         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
23093         DIVWEUO and DIVDEUO.
23094         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
23095         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
23096         (div_extend): Likewise.
23097         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
23098         builtin function.
23099         (__builtin_divweuo): Likewise.
23100         (__builtin_divdeo): Likewise.
23101         (__builtin_divdeuo): Likewise.
23103 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
23105         PR target/85095
23106         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
23107         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
23109         PR tree-optimization/82004
23110         * gimple-match-head.c (optimize_pow_to_exp): New function.
23111         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
23112         Don't fold to exp if optimize_pow_to_exp is false.
23114 2018-03-28  Martin Liska  <mliska@suse.cz>
23116         PR other/84819
23117         * calls.c (initialize_argument_information): Fix trailing space.
23118         * common.opt: Fix typo and provide better explanation for
23119         -fsanitize-coverage option.
23120         * config/i386/i386.opt: Fix typo.
23122 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
23123             Martin Liska  <mliska@suse.cz>
23125         PR sanitizer/85081
23126         * gimplify.c (asan_poison_variable): Don't do the check for
23127         gimplify_omp_ctxp here.
23128         (gimplify_decl_expr): Do it here.
23129         (gimplify_target_expr): Likewise.
23131 2018-03-28  Martin Liska  <mliska@suse.cz>
23133         PR target/84988
23134         * config/i386/i386.c (ix86_function_arg_advance): Do not call
23135         chkp_type_bounds_count if MPX is not enabled.
23137 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
23139         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
23141 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
23143         PR target/84914
23144         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
23145         function to create the function decl for complex long double
23146         multiply and divide for -mabi=ieeelongdouble.
23147         (init_float128_ieee): Call it.
23149 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
23151         PR target/85044
23152         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
23153         -fcf-protection=branch -mibt.
23154         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
23156 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
23158         PR target/81863
23159         * config/arm/arm.c (arm_valid_symbolic_address): Handle
23160         arm_word_relocations.
23162 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
23164         PR target/85056
23165         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
23166         extern array declarations.
23168 2018-03-27  Richard Biener  <rguenther@suse.de>
23170         PR middle-end/84067
23171         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
23172         explicit single_use checks.
23174 2018-03-27  Richard Biener  <rguenther@suse.de>
23176         PR tree-optimization/85082
23177         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
23178         Valueize the VUSE.
23180 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
23182         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
23183         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
23184         Turn on fasynchronous-unwind-tables and funwind-tables.
23186 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
23188         PR target/85073
23189         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
23190         (*bmi_blsr_<mode>_ccz): Ditto.
23192 2018-03-26  Tom de Vries  <tom@codesourcery.com>
23194         PR tree-optimization/85063
23195         * omp-general.c (offloading_function_p): New function.  Factor out
23196         of ...
23197         * omp-offload.c (pass_omp_target_link::gate): ... here.
23198         * omp-general.h (offloading_function_p): Declare.
23199         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
23200         with attribute omp declare target for offloading functions.
23202 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
23204         PR tree-optimization/84005
23205         * tree-data-ref.h (get_base_for_alignment): Declare.
23206         * tree-data-ref.c (get_base_for_alignment_1): New function.
23207         (get_base_for_alignment): Likewise.
23208         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
23209         get_base_for_alignment to find a suitable base object, instead
23210         of always using drb->base_address.
23212 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
23214         PR inline-asm/85022
23215         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
23216         known size by default.
23218 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
23220         PR inline-asm/85030
23221         * lra-constraints.c (process_alt_operands): Don't match BLKmode
23222         and non BLKmode operands.
23224 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23226         PR target/85026
23227         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
23228         Clean up attributes.
23230 2018-03-23  Richard Biener  <rguenther@suse.de>
23232         PR debug/85020
23233         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
23234         we are going to emit early debug for LTO.
23236 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
23238         PR inline-asm/85034
23239         * function.c (match_asm_constraints_1): Don't optimize if input
23240         doesn't satisfy general_operand predicate for output's mode.
23242         PR inline-asm/85022
23243         * alias.c (write_dependence_p): Don't require for x_canonicalized
23244         non-VOIDmode if x has VOIDmode.
23246         PR sanitizer/85029
23247         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
23248         just don't try to optimize it rather than assert it never happens.
23250 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23252         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
23253         macro expansions for definition of ST_INTERNAL_<mode> and
23254         LD_INTERNAL_<mode> builtins.
23255         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
23256         Remove prototype.
23257         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
23258         function.
23259         (altivec_expand_st_builtin): Likewise.
23260         (altivec_expand_builtin): Remove calls to deleted functions.
23261         (rs6000_address_for_altivec): Delete this function.
23262         * config/rs6000/vector.md: Remove expands for
23263         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
23265 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
23267         PR target/84826
23268         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
23269         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
23270         re-computing once computed.
23271         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
23272         (arm_init_machine_status): Initialize
23273         machine->static_chain_stack_bytes.
23275 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23277         PR target/84760
23278         * doc/extend.texi: Add four new prototypes for vec_ld.
23279         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
23280         definitions for more logical presentation.
23281         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
23282         entries for V1TI variants of __builtin_altivec_ld builtin.
23283         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
23284         handling of V1TI variant of LVX icode pattern.
23285         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
23286         (rs6000_gimple_fold_builtin): Likewise.
23287         (altivec_init_builtins): Add code to define
23288         __builtin_altivec_lvx_v1ti function.
23290 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
23292         PR inline-asm/84941
23293         * function.c (match_asm_constraints_1): Don't do the optimization
23294         if input isn't a REG, SUBREG, MEM or constant.
23296 2018-03-22  Tom de Vries  <tom@codesourcery.com>
23298         PR tree-optimization/84956
23299         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
23300         bb_has_abnormal_pred.
23302 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
23304         PR sanitizer/85018
23305         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
23306         DECL_INITIAL (decl) to decl at the end.
23307         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
23308         adjust the comment.
23310 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
23312         * doc/extend.texi (__builtin_tgmath): Document when complex
23313         integer types are treated as _Complex _Float64.
23315 2018-03-21  Tom de Vries  <tom@codesourcery.com>
23317         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
23319 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23321         PR tree-optimization/84960
23322         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
23323         if it is ENTRY block, move them into single succ of ENTRY in that case.
23325 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
23327         PR tree-optimization/84811
23328         * poly-int.h (poly_span_traits): Remove the T3 parameter and
23329         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
23330         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
23331         (known_subrange_p): Update accordingly.  Cast each value involved
23332         in the size comparison, rather than casting the result of the
23333         subtraction.
23335 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23337         PR tree-optimization/84982
23338         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
23339         by flipping the least significant bit rather than all bits from
23340         bitpos to bitpos + bitsize - 1.
23342 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23344         * doc/extend.texi (Deprecated Features): Remove mention of
23345         long-deleted deprecations.
23347 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23349         PR jit/84288
23350         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
23351         * configure: Regenerate.
23353 2018-03-21  Tom de Vries  <tom@codesourcery.com>
23355         PR tree-optimization/83126
23356         * tree-parloops.c (num_phis): New function.
23357         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
23359 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23361         * doc/extend.texi (Deprecated Features): Update deprecated flags,
23362         mention anon-struct/union members and trailing attributes.
23364 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
23366         PR tree-optimization/84969
23367         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
23368         builtin memset partitions if they set different rhs values.
23370 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23372         PR rtl-optimization/84989
23373         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
23374         VEC_DUPLICATE with scalar result mode.
23376 2018-03-21  Martin Liska  <mliska@suse.cz>
23378         PR ipa/84963
23379         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
23380         not intended return statement.
23382 2018-03-21  Martin Liska  <mliska@suse.cz>
23384         PR target/84988
23385         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
23386         (chkp_find_bound_slots_1): Limit number of iterations.
23388 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
23390         PR target/84838
23391         * Minor grammar fixes for x86 options.
23393 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23395         PR debug/84875
23396         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
23397         holding REG_CFA_RESTORE notes, instead turn them into a USE.
23399 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
23401         PR target/83789
23402         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
23403         (altivec_lvx_<mode>_1op): Likewise.
23404         (altivec_stvx_<mode>_2op): Likewise.
23405         (altivec_stvx_<mode>_1op): Likewise.
23406         (altivec_lvx_<VM2:mode>): New define_expand.
23407         (altivec_stvx_<VM2:mode>): Likewise.
23408         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
23409         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23410         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
23411         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23412         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
23413         (rs6000_gen_lvx): Likewise.
23414         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
23415         (altivec_expand_stv_builtin): Likewise.
23416         (altivec_expand_builtin): Likewise.
23417         * config/rs6000/vector.md: Likewise.
23419 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23421         PR target/82518
23422         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
23423         BYTES_BIG_ENDIAN.
23425 2018-03-20  Richard Biener  <rguenther@suse.de>
23427         PR target/84986
23428         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
23429         sign-conversions as zero, fall back to standard scalar_stmt
23430         cost for the rest.
23432 2018-03-20  Martin Liska  <mliska@suse.cz>
23434         PR ipa/84825
23435         * predict.c (rebuild_frequencies): Handle case when we have
23436         PROFILE_ABSENT, but flag_guess_branch_prob is false.
23438 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23440         PR target/84990
23441         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
23442         flag_section_anchors.
23443         * varasm.c (use_blocks_for_decl_p): Remove hack for
23444         dw2_force_const_mem.
23446         PR target/84845
23447         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
23448         to ...
23449         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
23450         be created, use lowpart_subreg of operands[0] rather than operands[0]
23451         itself.
23452         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
23453         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
23454         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
23455         and n constraint instead of aarch64_shift_imm_di and Usd.
23456         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
23457         (*aarch64_<optab>_reg_minus<mode>3): ... this.
23459 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
23461         PR target/82989
23462         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
23463         to favor GPR over NEON registers.
23464         (<shift>di3_neon): Likewise.
23466 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23468         PR target/84952
23469         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
23470         (nvptx_process_pars): Emit bar.sync asap and alap.
23472 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23474         PR target/84954
23475         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
23476         seen_label if seen_label is already set.
23478 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23480         PR target/84945
23481         * config/i386/i386.c (fold_builtin_cpu): For features above 31
23482         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
23483         Use 1U instead of 1.  Formatting fixes.
23485         PR c/84953
23486         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
23487         instead of TREE_TYPE (s1) for the return value.
23489 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
23491         PR tree-optimization/84946
23492         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
23493         bitsize + bitsize in poly_uint64 rather than poly_int64.
23495         PR sanitizer/78651
23496         * dwarf2asm.c: Include fold-const.c.
23497         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
23498         of decl rather than decl itself.
23500         PR rtl-optimization/84643
23501         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
23503 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
23505         PR sanitizer/78651
23506         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
23507         calling assemble_variable.
23509 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
23511         PR target/81647
23512         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
23513         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
23515 2018-03-19  Jim Wilson  <jimw@sifive.com>
23517         PR bootstrap/84856
23518         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
23519         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
23520         (riscv_first_stack_step): Likewise.
23521         (riscv_option_override): Use STACK_BOUNDARY instead of
23522         MIN_STACK_BOUNDARY.
23523         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
23524         MIN_STACK_BOUNDARY.
23525         (BIGGEST_ALIGNMENT): Set to 128.
23526         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
23527         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
23528         STACK_BOUNDARY.
23530 2018-03-19  Richard Biener  <rguenther@suse.de>
23532         PR tree-optimization/84933
23533         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
23534         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
23536 2018-03-19  Richard Biener  <rguenther@suse.de>
23538         PR tree-optimization/84859
23539         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
23540         (cond_if_else_store_replacement): Perform sinking operation on
23541         single-store BBs regardless of MAX_STORES_TO_SINK setting.
23542         Generalize what a BB with a single eligible store is.
23544 2018-03-19  Richard Biener  <rguenther@suse.de>
23546         PR tree-optimization/84929
23547         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
23548         chrec_is_positive against non-chrec arg.
23550 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
23552         PR target/84711
23553         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
23555 2018-03-18  Martin Liska  <mliska@suse.cz>
23557         PR rtl-optimization/84635
23558         * regrename.c (build_def_use): Use matches_mode only when
23559         matches >= 0.
23561 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
23563         PR tree-optimization/84913
23564         * tree-vect-loop.c (vectorizable_reduction): Don't try to
23565         vectorize chains of COND_EXPRs.
23567 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23569         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
23571 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23573         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
23575 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23577         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
23579 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23580             Kito Cheng  <kito.cheng@gmail.com>
23582         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
23583         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
23584         (nds32_adjust_reg_alloc_order): New function.
23585         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
23587 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
23589         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
23590         nds32_print_operand, nds32_print_operand_address): Use
23591         HOST_WIDE_INT_PRINT_DEC instead.
23593 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23595         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
23597 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
23599         PR target/84902
23600         * config/i386/i386.c (initial_ix86_tune_features,
23601         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
23602         unsigned long long.
23603         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
23604         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
23605         rather than 1u << ix86_tune.  Formatting fix.
23606         (ix86_option_override_internal): Change ix86_arch_mask from
23607         unsigned int to unsigned HOST_WIDE_INT, initialize to
23608         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
23609         (ix86_function_specific_restore): Likewise.
23611 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23613         PR target/84899
23614         * postreload.c (reload_combine_recognize_pattern): Perform
23615         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
23616         truncate_int_for_mode the result for the destination's mode.
23618         PR c/84909
23619         * hsa-gen.c (mem_type_for_type): Fix comment typo.
23620         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
23621         Likewise.
23622         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23623         Likewise.
23625 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
23627         PR target/84876
23628         * lra-assigns.c (lra_split_hard_reg_for): Don't use
23629         regno_allocno_class_array and sorted_pseudos.
23630         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
23631         insns where regno is used.
23633 2018-03-16  Martin Liska  <mliska@suse.cz>
23635         PR ipa/84833
23636         * multiple_target.c (create_dispatcher_calls): Redirect
23637         reference in the symbol table.
23639 2018-03-16  Martin Liska  <mliska@suse.cz>
23641         PR ipa/84722
23642         * multiple_target.c (create_dispatcher_calls): Redirect also
23643         an alias.
23645 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23647         PR c++/79937
23648         PR c++/82410
23649         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
23650         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
23651         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
23653 2018-03-16  Julia Koval  <julia.koval@intel.com>
23655         * doc/invoke.texi (Skylake Server): Add CLWB.
23656         Cannonlake): Remove CLWB.
23658 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23660         PR tree-optimization/84841
23661         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
23662         1 << 3.
23663         (FLOAT_ONE_CONST_TYPE): Define.
23664         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
23665         (sort_by_operand_rank): Put entries with higher constant_type last
23666         rather than first to match comments.
23668 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
23670         * config/nios2/nios2.md (movsi_internal): Fix thinko in
23671         split predicate.
23673 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23675         PR c++/79085
23676         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
23677         check and use address of target always.
23679 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
23681         PR target/84574
23682         * config/i386/i386.c (indirect_thunk_needed): Update comments.
23683         (indirect_thunk_bnd_needed): Likewise.
23684         (indirect_thunks_used): Likewise.
23685         (indirect_thunks_bnd_used): Likewise.
23686         (indirect_return_needed): New.
23687         (indirect_return_bnd_needed): Likewise.
23688         (output_indirect_thunk_function): Add a bool argument for
23689         function return.
23690         (output_indirect_thunk_function): Don't generate alias for
23691         function return thunk.
23692         (ix86_code_end): Call output_indirect_thunk_function to generate
23693         function return thunks.
23694         (ix86_output_function_return): Set indirect_return_bnd_needed
23695         and indirect_return_needed instead of indirect_thunk_bnd_needed
23696         and indirect_thunk_needed.
23698 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
23700         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
23701         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
23702         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
23704 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
23705             Paul Hua <paul.hua.gm@gmail.com>
23707         PR c/84852
23708         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
23710 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
23712         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
23713         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
23714         resp. SFmode cases.
23716 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
23718         PR target/84711
23719         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
23720         instead of GET_MODE_SIZE when comparing Units.
23722 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
23724         PR target/68256
23725         * varasm.c (hash_section): Return an unchangeble hash value
23726         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
23727         Return !aarch64_can_use_per_function_literal_pools_p ().
23729 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23731         PR target/84860
23732         * optabs.c (emit_conditional_move): Pass address of cmode's copy
23733         rather than address of cmode as last argument to prepare_cmp_insn.
23735 2018-03-15  Julia Koval  <julia.koval@intel.com>
23737         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
23738         F_AVX512VNNI, F_AVX512BITALG): New.
23740 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
23742         PR target/83451
23743         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
23744         insn for floating-point loads and stores.
23746 2018-03-14  Carl Love  <cel@us.ibm.com>
23748         * config/rs6000/rs6000-c.c: Add macro definitions for
23749         ALTIVEC_BUILTIN_VEC_PERMXOR.
23750         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
23751         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
23752         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
23753         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
23754         UNSPEC_VPERMXOR.
23755         * config/doc/extend.texi: Add prototypes for vec_permxor.
23757 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
23759         PR c/84852
23760         * diagnostic-show-locus.c (class layout_point): Convert m_line
23761         from int to linenum_type.
23762         (line_span::comparator): Use linenum "compare" function when
23763         comparing line numbers.
23764         (test_line_span): New function.
23765         (layout_range::contains_point): Convert param "row" from int to
23766         linenum_type.
23767         (layout_range::intersects_line_p): Likewise.
23768         (layout::will_show_line_p): Likewise.
23769         (layout::print_source_line): Likewise.
23770         (layout::should_print_annotation_line_p): Likewise.
23771         (layout::print_annotation_line): Likewise.
23772         (layout::print_leading_fixits): Likewise.
23773         (layout::annotation_line_showed_range_p): Likewise.
23774         (struct line_corrections): Likewise for field m_row.
23775         (line_corrections::line_corrections): Likewise for param "row".
23776         (layout::print_trailing_fixits): Likewise.
23777         (layout::get_state_at_point): Likewise.
23778         (layout::get_x_bound_for_row): Likewise.
23779         (layout::print_line): Likewise.
23780         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
23781         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
23782         * input.c (selftest::test_linenum_comparisons): New function.
23783         (selftest::input_c_tests): Call it.
23784         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
23785         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
23786         * selftest.h (ASSERT_GT): New macro.
23787         (ASSERT_GT_AT): New macro.
23788         (ASSERT_LT): New macro.
23789         (ASSERT_LT_AT): New macro.
23791 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
23793         PR rtl-optimization/84780
23794         * combine.c (distribute_links): Don't make a link based on pc_rtx.
23796 2018-03-14  Martin Liska  <mliska@suse.cz>
23798         * tree.c (record_node_allocation_statistics): Use
23799         get_stats_node_kind.
23800         (get_stats_node_kind): New function extracted from
23801         record_node_allocation_statistics.
23802         (free_node): Use get_stats_node_kind.
23804 2018-03-14  Richard Biener  <rguenther@suse.de>
23806         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
23807         that the value-set of ANTIC_IN doesn't grow.
23809         Revert
23810         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23811         member.
23812         (BB_VISITED_WITH_VISITED_SUCCS): New define.
23813         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23815 2018-03-14  Julia Koval  <julia.koval@intel.com>
23817         * config.gcc (icelake-client, icelake-server): New.
23818         (icelake): Remove.
23819         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
23820         (initial_ix86_arch_features): Ditto.
23821         (PTA_SKYLAKE): Add SGX.
23822         (PTA_ICELAKE): Remove.
23823         (PTA_ICELAKE_CLIENT): New.
23824         (PTA_ICELAKE_SERVER): New.
23825         (ix86_option_override_internal): Split up icelake on icelake client and
23826         icelake server.
23827         (get_builtin_code_for_version): Ditto.
23828         (fold_builtin_cpu): Ditto.
23829         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
23830         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
23831         * config/i386/i386.h (processor_type): Ditto.
23832         * doc/invoke.texi: Ditto.
23834 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
23836         PR sanitizer/83392
23837         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
23838         INTEGER_CST offset, add it together with bitpos / 8 and
23839         sign extend based on POINTER_SIZE.
23841         PR target/84844
23842         Revert
23843         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
23845         PR target/78090
23846         * config/i386/constraints.md (Yc): New register constraint.
23847         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
23848         Use Yc constraint for alternative 2 of operand 0.  Remove
23849         preferred_for_speed attribute.
23851 2018-03-14  Richard Biener  <rguenther@suse.de>
23853         PR tree-optimization/84830
23854         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
23855         with the old one to avoid oscillations.
23857 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
23859         PR target/83712
23860         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
23861         pseudos.
23862         (assign_by_spills): Return a flag of reload assignment failure.
23863         Do not process the reload assignment failures.  Do not spill other
23864         reload pseudos if they has the same reg class.  Update n if
23865         necessary.
23866         (lra_assign): Add a return arg.  Set up from the result of
23867         assign_by_spills call.
23868         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23869         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23870         usage_insns if it is not NULL.
23871         (spill_hard_reg_in_range): New function.
23872         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23873         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23874         function prototypes.
23875         (lra_assign): Change prototype.
23876         * lra.c (lra): Add code to deal with fails by splitting hard reg
23877         live ranges.
23879 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
23881         * config/riscv/riscv.opt (mrelax): New option.
23882         * config/riscv/riscv.c (riscv_file_start): Emit ".option
23883         "norelax" when riscv_mrelax is disabled.
23884         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
23886 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23888         PR target/84743
23889         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
23890         reassociation for int modes.
23892 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23894         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
23895         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
23896         for big-endian.
23897         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
23898         * config/aarch64/aarch64-sve.md
23899         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
23900         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
23901         (*extend<mode><Vwide>2): Rename to...
23902         (aarch64_sve_extend<mode><Vwide>2): ...this.
23903         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
23904         renaming the old pattern to...
23905         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
23906         unsigned packs.
23907         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
23908         define_expand, renaming the old pattern to...
23909         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
23910         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
23911         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
23912         account when deciding which SVE instruction the optab should use.
23913         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
23915 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23917         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
23918         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
23919         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
23920         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
23921         (tlsdesc_small_<mode>): Turn a define_expand and use
23922         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
23923         (tlsdesc_small_advsimd_<mode>): ...this.
23924         (tlsdesc_small_sve_<mode>): New pattern.
23926 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23928         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
23929         (UNSPEC_UMUL_HIGHPART): New constants.
23930         (MUL_HIGHPART): New int iteraor.
23931         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
23932         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
23933         define_expand.
23934         (*<su>mul<mode>3_highpart): New define_insn.
23936 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
23938         PR lto/84805
23939         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23940         incomplete types.
23942 2018-03-13  Martin Liska  <mliska@suse.cz>
23944         PR ipa/84658.
23945         * (sem_item_optimizer::sem_item_optimizer): Initialize new
23946         vector.
23947         (sem_item_optimizer::~sem_item_optimizer): Release it.
23948         (sem_item_optimizer::merge_classes): Register variable aliases.
23949         (sem_item_optimizer::fixup_pt_set): New function.
23950         (sem_item_optimizer::fixup_points_to_sets): Likewise.
23951         * ipa-icf.h: Declare new variables and functions.
23953 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
23955         PR middle-end/84834
23956         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
23957         integer_pow2p@2 and test integer_pow2p in condition.
23958         (A < 0 ? C : 0): Similarly for @1.
23960         PR middle-end/84831
23961         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
23962         characters starting at p contain '\0' character, don't look beyond
23963         that.
23965         PR target/84827
23966         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
23967         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
23969         PR target/84828
23970         * reg-stack.c (change_stack): Change update_end var from int to
23971         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
23972         also call set_block_for_insn on the newly added insns and rescan.
23974         PR target/84786
23975         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
23976         on the last operand.
23978         PR c++/84704
23979         * tree.c (stabilize_reference_1): Return save_expr (e) for
23980         STATEMENT_LIST even if it doesn't have side-effects.
23982 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
23984         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
23986 2018-03-12  Renlin Li  <renlin.li@arm.com>
23988         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
23989         aarch64_output_scalar_simd_mov_immediate.
23991 2018-03-12  Martin Sebor  <msebor@redhat.com>
23993         PR tree-optimization/83456
23994         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
23995         for perfectly overlapping calls to memcpy.
23996         (gimple_fold_builtin_memory_chk): Same.
23997         (gimple_fold_builtin_strcpy): Handle no-warning.
23998         (gimple_fold_builtin_stxcpy_chk): Same.
23999         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
24001 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
24003         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
24004         parameter.  Use it for SFmode.
24005         (rs6000_function_arg_advance_1): Adjust.
24006         (rs6000_function_arg): Adjust.
24007         (rs6000_gimplify_va_arg): Pass false for that new parameter.
24009 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
24011         PR rtl-optimization/84169
24012         PR rtl-optimization/84780
24013         * combine.c (can_combine_p): Check for a 2-insn combination whether
24014         the destination register is used between the two insns, too.
24016 2018-03-12  Richard Biener  <rguenther@suse.de>
24018         PR tree-optimization/84803
24019         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
24020         for refs DR analysis didn't process.
24022 2018-03-12  Richard Biener  <rguenther@suse.de>
24024         PR tree-optimization/84777
24025         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
24026         force-vectorize loops ignore whether we are optimizing for size.
24028 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
24030         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
24031         (TARGET_MD_ASM_ADJUST): Define.
24033 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
24034             Kito Cheng  <kito.cheng@gmail.com>
24035             Chung-Ju Wu  <jasonwucj@gmail.com>
24037         * config/nds32/nds32.c (nds32_compute_stack_frame,
24038         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
24039         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
24040         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
24041         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
24042         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
24043         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
24044         * config/nds32/nds32.md (prologue, epilogue): Use macro
24045         NDS32_V3PUSH_AVAILABLE_P to do checking.
24047 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
24049         PR debug/58150
24050         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
24051         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
24052         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
24053         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
24054         addition of most attributes on !orig_type_die or the attribute not
24055         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
24057 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
24058             Chung-Ju Wu  <jasonwucj@gmail.com>
24060         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
24061         __NDS32_VH__ macro.
24062         * config/nds32/nds32.opt (mvh): New option.
24064 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
24065             Chung-Ju Wu  <jasonwucj@gmail.com>
24067         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
24068         function.
24069         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
24070         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
24071         definition.
24073 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
24074             Chung-Ju Wu  <jasonwucj@gmail.com>
24076         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
24077         function.
24078         * config/nds32/nds32-multiple.md (strlensi): New pattern.
24079         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
24081 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
24082             Kito Cheng  <kito.cheng@gmail.com>
24083             Chung-Ju Wu  <jasonwucj@gmail.com>
24085         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
24086         UNSPEC_FFMISM and UNSPEC_FLMISM.
24087         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
24088         for ffb, ffmism and flmism.
24089         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
24090         (unspec_ffmism): Ditto.
24091         (unspec_flmism): Ditto.
24092         (nds32_expand_builtin_impl): Check if string extension is available.
24093         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
24094         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
24096 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
24098         Reverting patch:
24099         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
24101         PR target/83712
24102         * lra-assigns.c (assign_by_spills): Return a flag of reload
24103         assignment failure.  Do not process the reload assignment
24104         failures.  Do not spill other reload pseudos if they has the same
24105         reg class.
24106         (lra_assign): Add a return arg.  Set up from the result of
24107         assign_by_spills call.
24108         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24109         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
24110         usage_insns if it is not NULL.
24111         (spill_hard_reg_in_range): New function.
24112         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24113         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24114         function prototypes.
24115         (lra_assign): Change prototype.
24116         * lra.c (lra): Add code to deal with fails by splitting hard reg
24117         live ranges.
24119 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
24121         PR target/84807
24122         * config/i386/i386.opt: Replace Enforcment with Enforcement.
24124 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
24126         PR debug/84620
24127         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
24128         (dw_val_node): Add val_symbolic_view.
24129         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
24130         (symview_upper_bound): New.
24131         (new_line_info_table): Initialize symviews_since_reset.
24132         (dwarf2out_source_line): Count symviews_since_reset and set
24133         symview_upper_bound.
24134         (dw_val_equal_p): Handle symview.
24135         (add_AT_symview): New.
24136         (print_dw_val): Handle symview.
24137         (attr_checksum, attr_checksum_ordered): Likewise.
24138         (same_dw_val_p, size_of_die): Likewise.
24139         (value_format, output_die): Likewise.
24140         (add_high_low_attributes): Use add_AT_symview for entry_view.
24141         (dwarf2out_finish): Reset symview_upper_bound, clear
24142         zero_view_p.
24144 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
24146         PR target/83969
24147         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
24148         Add strict argument and use it.
24149         (rs6000_split_multireg_move): Update for new strict argument.
24150         (mem_operand_gpr): Disallow all non-offsettable addresses.
24151         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
24153 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
24155         PR target/84772
24156         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
24157         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
24158         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
24160         PR c++/84767
24161         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
24162         decl, use remap_type if we want to use the type.
24164 2018-03-09  Martin Sebor  <msebor@redhat.com>
24166         PR tree-optimization/84526
24167         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
24168         Remove dead code.
24169         (builtin_access::generic_overlap): Be prepared to handle non-array
24170         base objects.
24172 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
24174         PR rtl-optimization/84682
24175         * lra-constraints.c (process_address_1): Check is_address flag
24176         for address constraints.
24177         (process_alt_operands): Likewise.
24178         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
24179         preprocess_constraints.
24180         * recog.h (preprocess_constraints): Add oploc parameter.
24181         Adjust callers.
24182         * recog.c (preprocess_constraints): Test address_operand for
24183         CT_ADDRESS constraints.
24185 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
24187         PR target/83712
24188         * lra-assigns.c (assign_by_spills): Return a flag of reload
24189         assignment failure.  Do not process the reload assignment
24190         failures.  Do not spill other reload pseudos if they has the same
24191         reg class.
24192         (lra_assign): Add a return arg.  Set up from the result of
24193         assign_by_spills call.
24194         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24195         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
24196         usage_insns if it is not NULL.
24197         (spill_hard_reg_in_range): New function.
24198         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24199         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24200         function prototypes.
24201         (lra_assign): Change prototype.
24202         * lra.c (lra): Add code to deal with fails by splitting hard reg
24203         live ranges.
24205 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24207         PR target/83193
24208         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
24209         Accept complain bool parameter.  Only emit errors if it is true.
24210         (arm_parse_cpu_option_name): Likewise.
24211         (arm_target_thumb_only): Adjust callers of the above.
24212         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
24213         prototype to take a default true bool parameter.
24214         (arm_parse_arch_option_name): Likewise.
24216 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
24217             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
24219         PR jit/64089
24220         PR jit/84288
24221         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
24222         * configure: Regenerate.
24223         * configure.ac ("linker --version-script option"): New.
24224         ("linker soname option"): New.
24226 2018-03-09  Richard Biener  <rguenther@suse.de>
24228         PR tree-optimization/84775
24229         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
24230         immediate uses of predicate stmts and mark them modified.
24232         Revert
24233         PR tree-optimization/84178
24234         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24235         to caller.
24236         (version_loop_for_if_conversion): Delay update_ssa call.
24237         (tree_if_conversion): Delay update_ssa until after predicate
24238         insertion.
24240 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
24242         PR target/84763
24243         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
24244         when the function accesses prior frames.
24246 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
24248         PR debug/84456
24249         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
24250         gen_llsym, otherwise call maybe_gen_llsym.
24252         PR inline-asm/84742
24253         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
24254         has ',' character inside of it.
24256 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24258         PR target/84748
24259         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
24260         as clobbering CC_REGNUM.
24262 2018-03-08  Richard Biener  <rguenther@suse.de>
24264         PR middle-end/84552
24265         * tree-scalar-evolution.c: Include tree-into-ssa.h.
24266         (follow_copies_to_constant): Do not follow SSA names registered
24267         for update.
24269 2018-03-08  Richard Biener  <rguenther@suse.de>
24271         PR tree-optimization/84178
24272         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24273         to caller.
24274         (version_loop_for_if_conversion): Delay update_ssa call.
24275         (tree_if_conversion): Delay update_ssa until after predicate
24276         insertion.
24278 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
24280         PR tree-optimization/84178
24281         * tree-if-conv.c (release_bb_predicate): Remove the
24282         the assertion that the stmts have NULL use_ops.
24283         Discard the statements, asserting that they haven't
24284         yet been added to a BB.
24286 2018-03-08  Richard Biener  <rguenther@suse.de>
24288         PR tree-optimization/84746
24289         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
24290         (phi_translate): Pass in destination ANTIC_OUT set.
24291         (phi_translate_1): Likewise.  For a simplified result lookup
24292         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
24293         (phi_translate_set): Adjust.
24294         (do_pre_regular_insertion): Likewise.
24295         (do_pre_partial_partial_insertion): Likewise.
24297 2018-03-08  Martin Liska  <mliska@suse.cz>
24299         PR gcov-profile/84735
24300         * doc/gcov.texi: Document usage of profile files.
24301         * gcov-io.h: Document changes in the format.
24303 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
24305         PR debug/84404
24306         PR debug/84408
24307         * dwarf2out.c (struct dw_line_info_table): Update comments for
24308         view == -1.
24309         (FORCE_RESET_NEXT_VIEW): New.
24310         (FORCE_RESETTING_VIEW_P): New.
24311         (RESETTING_VIEW_P): Check for -1 too.
24312         (ZERO_VIEW_P): Likewise.
24313         (new_line_info_table): Force-reset next view.
24314         (dwarf2out_begin_function): Likewise.
24315         (dwarf2out_source_line): Simplify zero_view_p initialization.
24316         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
24317         view directly.  Omit view when omitting .loc at line 0.
24319 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
24321         PR tree-optimization/84740
24322         * tree-switch-conversion.c (process_switch): Call build_constructors
24323         only if info.phi_count is non-zero.
24325         PR tree-optimization/84739
24326         * tree-tailcall.c (find_tail_calls): Check call arguments against
24327         DECL_ARGUMENTS (current_function_decl) rather than
24328         DECL_ARGUMENTS (func) when checking for tail recursion.
24330 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24332         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
24333         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
24334         Volker Reichelt's entry and add entries for people that perform
24335         GCC fuzzy testing and report numerous bugs.
24337 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
24339         PR target/82411
24340         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
24341         readonly data in sdata, if that is disabled.
24342         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
24343         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
24344         -mreadonly-in-sdata option.
24346 2018-03-07  Martin Sebor  <msebor@redhat.com>
24348         PR tree-optimization/84468
24349         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
24350         basic block when looking for nul assignment.
24352 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
24354         PR target/84277
24355         * except.h (output_function_exception_table): Adjust prototype.
24356         * except.c (output_function_exception_table): Remove FNNAME parameter
24357         and add SECTION parameter.  Ouput one part of the table at a time.
24358         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
24359         the first part of the exception table and emit unwind directives.
24360         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
24361         (i386_pe_seh_cold_init): Likewise.
24362         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
24363         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
24364         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
24365         (ix86_output_call_insn): Emit a nop in one more case for SEH.
24366         * config/i386/winnt.c: Include except.h.
24367         (struct seh_frame_state): Add reg_offset, after_prologue and
24368         in_cold_section fields.
24369         (i386_pe_seh_end_prologue): Set seh->after_prologue.
24370         (i386_pe_seh_cold_init): New function.
24371         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
24372         to seh->in_cold_section.
24373         (seh_emit_push): Record the offset of the push.
24374         (seh_emit_save): Record the offet of the save.
24375         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
24376         Test seh->after_prologue to disregard the epilogue.
24377         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
24378         (i386_pe_end_cold_function): New function.
24380 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24382         PR fortran/84565
24383         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
24384         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
24386         PR c++/84704
24387         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
24388         on tmp_var.
24389         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
24390         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
24392         PR middle-end/84723
24393         * multiple_target.c: Include tree-inline.h and intl.h.
24394         (expand_target_clones): Diagnose and fail if node->definition and
24395         !tree_versionable_function_p (node->decl).
24397 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
24399         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
24400         sprint_ul.
24401         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
24402         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
24403         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
24405 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
24407         PR target/84710
24408         * combine.c (try_combine): Use reg_or_subregno instead of handling
24409         just paradoxical SUBREGs and REGs.
24411 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
24413         * config/arc/arc.c (arc_finalize_pic): Remove function.
24414         (arc_must_save_register): We use single base PIC register, remove
24415         checks to save/restore the PIC register.
24416         (arc_expand_prologue): Likewise.
24417         * config/arc/arc-protos.h (arc_set_default_type_attributes):
24418         Remove.
24419         (arc_verify_short): Likewise.
24420         (arc_attr_type): Likewise.
24421         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
24422         (walk_stores): Likewise.
24423         (arc_address_cost): Make it static.
24424         (arc_verify_short): Likewise.
24425         (branch_dest): Likewise.
24426         (arc_attr_type): Likewise.
24427         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
24428         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
24429         (arc_final_prescan_insn): Remove inserting the nops due to
24430         hardware hazards.  It is done in reorg step.
24431         (insn_length_variant_t): Remove.
24432         (insn_length_parameters_t): Likewise.
24433         (arc_insn_length_parameters): Likewise.
24434         (arc_get_insn_variants): Likewise.
24435         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
24437 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
24439         PR inline-asm/84683
24440         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
24441         assertion failure.
24443         PR tree-optimization/84687
24444         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
24445         on new_node->decl.
24446         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
24448 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24450         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
24451         Rename to ppc_speculation_barrier.
24452         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
24453         __builtin_ppc_speculation_barrier.
24455 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
24457         PR target/84700
24458         * combine.c (combine_simplify_rtx): Don't try to simplify if
24459         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
24460         are equal to x.
24462 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
24464         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
24465         to 32 bytes when compiling for POWER9.
24467 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
24469         PR target/84564
24470         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
24471         regparm >= 3 with no arg reg available also for calls with
24472         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
24474         PR target/84524
24475         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
24476         orig,vex.
24477         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
24479 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
24481         PR target/84264
24482         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
24484 2018-03-05  Richard Biener  <rguenther@suse.de>
24486         PR tree-optimization/84486
24487         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
24488         When inserting a __builtin_assume_aligned call set the LHS
24489         SSA name alignment info accordingly.
24491 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
24493         PR tree-optimization/84114
24494         * config/aarch64/aarch64.c (aarch64_reassociation_width)
24495         Avoid reassociation of FLOAT_MODE addition.
24497 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
24499         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
24500         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
24501         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
24502         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
24503         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
24504         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
24505         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
24506         and -mwbnoinvd.
24507         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
24508         __builtin_ia32_wbinvd): New builtins.
24509         (SPECIAL_ARGS2): New.
24510         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
24511         (SPECIAL_ARGS2): New.
24512         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
24513         (ix86_valid_target_attribute_inner_p): Ditto.
24514         (ix86_init_mmx_sse_builtins): Add special_args2.
24515         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
24516         TARGET_WBNOINVD_P): New.
24517         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
24518         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
24519         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
24520         * config/i386/immintrin.h (_wbinvd): New intrinsic.
24521         * config/i386/pconfigintrin.h: New file.
24522         * config/i386/wbnoinvdintrin.h: Ditto.
24523         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
24524         wbnoinvdintrin.h.
24525         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
24527 2018-03-05  Richard Biener  <rguenther@suse.de>
24529         PR tree-optimization/84670
24530         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
24531         member.
24532         (BB_VISITED_WITH_VISITED_SUCCS): New define.
24533         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
24534         (compute_antic_aux): Only assert the number of values in ANTIC_IN
24535         doesn't grow if all successors (recursively) were visited at least
24536         once.
24538 2018-03-05  Richard Biener  <rguenther@suse.de>
24540         PR tree-optimization/84650
24541         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
24542         if executed in the loop pipeline.
24544 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
24546         * doc/configfiles.texi (Configuration Files): Move info about
24547         conditionalizing $target-protos.h to...
24548         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
24549         differs from $target-protos.h.
24551 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
24552             Chung-Ju Wu  <jasonwucj@gmail.com>
24554         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
24555         * config/nds32/nds32-multiple.md (setmemsi): Define.
24556         * config/nds32/nds32-memory-manipulation.c
24557         (nds32_gen_dup_4_byte_to_word_value): New.
24558         (emit_setmem_word_loop): New.
24559         (emit_setmem_byte_loop): New.
24560         (nds32_expand_setmem_loop): New.
24561         (nds32_expand_setmem_loop_v3m): New.
24562         (nds32_expand_setmem_unroll): New.
24563         (nds32_expand_setmem): New.
24565 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24566             Chung-Ju Wu  <jasonwucj@gmail.com>
24568         * config/nds32/nds32-memory-manipulation.c
24569         (nds32_emit_load_store): New.
24570         (nds32_emit_post_inc_load_store): New.
24571         (nds32_emit_mem_move): New.
24572         (nds32_emit_mem_move_block): New.
24573         (nds32_expand_movmemsi_loop_unknown_size): New.
24574         (nds32_expand_movmemsi_loop_known_size): New.
24575         (nds32_expand_movmemsi_loop): New.
24576         (nds32_expand_movmemsi_unroll): New.
24577         (nds32_expand_movmemqi): Rename ...
24578         (nds32_expand_movmemsi): ... to this.
24579         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
24580         (movmemsi): ... to this.
24581         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
24582         (nds32_expand_movmemsi): ... to this.
24584 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24585             Monk Chiang  <sh.chiang04@gmail.com>
24586             Chung-Ju Wu  <jasonwucj@gmail.com>
24588         * config/nds32/nds32-protos.h
24589         (nds32_expand_load_multiple): New arguments.
24590         (nds32_expand_store_multiple): Ditto.
24591         (nds32_valid_multiple_load_store): Rename ...
24592         (nds32_valid_multiple_load_store_p): ... to this.
24593         * config/nds32/nds32-memory-manipulation.c
24594         (nds32_expand_load_multiple): Refine implementation.
24595         (nds32_expand_store_multiple): Ditto.
24596         * config/nds32/nds32-multiple.md
24597         (load_multiple): Update nds32_expand_load_multiple interface.
24598         (store_multiple): Update nds32_expand_store_multiple interface.
24599         * config/nds32/nds32-predicates.c
24600         (nds32_valid_multiple_load_store): Rename ...
24601         (nds32_valid_multiple_load_store_p): ... to this and refine
24602         implementation.
24603         * config/nds32/predicates.md
24604         (nds32_load_multiple_and_update_address_operation): New predicate.
24605         (nds32_store_multiple_and_update_address_operation): New predicate.
24607 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24608             Chung-Ju Wu  <jasonwucj@gmail.com>
24610         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
24611         (combo): New attribute.
24612         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
24614 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
24616         * config/nds32/nds32.opt: Change -mcmodel= default value.
24618 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
24619             Monk Chiang  <sh.chiang04@gmail.com>
24620             Chung-Ju Wu  <jasonwucj@gmail.com>
24622         * config/nds32/constants.md (unspec_element): New enum.
24623         * config/nds32/constraints.md (Umw): New constraint.
24624         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
24625         * config/nds32/nds32-intrinsic.md: Likewise.
24626         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
24627         (nds32_valid_smw_lwm_base_p): New.
24628         (nds32_output_smw_single_word): New.
24629         (nds32_output_lmw_single_word): New.
24630         (nds32_expand_unaligned_load): New.
24631         (nds32_expand_unaligned_store): New.
24632         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
24633         (nds32_output_smw_single_word): Declare.
24634         (nds32_output_lmw_single_word): Declare.
24635         (nds32_expand_unaligned_load): Declare.
24636         (nds32_expand_unaligned_store): Declare.
24637         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
24638         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
24639         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
24640         NDS32_BUILTIN_UASTORE_DW.
24641         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
24642         predicate.
24644 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
24645             Kito Cheng  <kito.cheng@gmail.com>
24646             Chung-Ju Wu  <jasonwucj@gmail.com>
24648         * config/nds32/nds32-intrinsic.c
24649         (nds32_expand_builtin_null_ftype_reg): Delete.
24650         (nds32_expand_builtin_reg_ftype_imm): Ditto.
24651         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
24652         (nds32_read_argument): New.
24653         (nds32_legitimize_target): Ditto.
24654         (nds32_legitimize_argument): Ditto.
24655         (nds32_check_constant_argument): Ditto.
24656         (nds32_expand_unop_builtin): Ditto.
24657         (nds32_expand_unopimm_builtin): Ditto.
24658         (nds32_expand_binop_builtin): Ditto.
24659         (nds32_builtin_decl_impl): Ditto.
24660         (builtin_description): Ditto.
24661         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
24662         (nds32_init_builtins_impl): Ditto.
24663         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
24664         (nds32_builtin_decl): New.
24665         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
24666         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
24668 2018-03-02  Jeff Law  <law@redhat.com>
24670         * reorg.c (stop_search_p): Handle DEBUG_INSN.
24671         (redundant_insn, fill_simple_delay_slots): Likewise.
24672         (fill_slots_from_thread): Likewise.
24673         * resource.c (mark_referenced_resources): Likewise.
24674         (mark_set_resources, find_dead_or_set_registers): Likewise.
24676 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24678         * substring-locations.h (format_warning_va): Formatting fix for
24679         ATTRIBUTE_GCC_DIAG.
24680         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
24681         argument.
24682         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
24683         * substring-locations.c: Include intl.h.
24684         (format_warning_va): Turned into small wrapper around
24685         format_warning_n_va, renamed to ...
24686         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
24687         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
24688         use ngettext.
24689         (format_warning_at_substring_n): New function.
24690         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
24691         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
24692         format_warning_at_substring with just a shorter name instead of
24693         const function pointer.
24694         (fmtwarn_n): New function.
24695         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
24696         appropriate, get rid of all the fmtstr temporaries, move conditionals
24697         with G_() wrapped string literals directly into fmtwarn arguments,
24698         cast dir.len to (int), formatting fixes.
24700 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
24702         * doc/invoke.texi: Remove "Cilk Plus" references.
24704 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24705             Richard Biener  <rguenther@suse.de>
24707         PR ipa/84628
24708         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
24709         for error or warning attributes if CALL_FROM_THUNK_P is set.
24710         Formatting fixes.
24712 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24714         PR target/56540
24715         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
24716         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
24718         PR target/56540
24719         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
24720         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
24722         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
24723         instead of -1U in last predictors element's probability member.
24725 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
24727         PR ipa/83983
24728         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
24729         arguments if they are comparable.
24731 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24733         PR tree-optimization/84634
24734         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
24735         masks and masked_loop_p with a single loop_masks, making sure it's
24736         null for bb vectorization.
24738 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24740         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
24741         (vect_analyze_data_ref_access): Use loop->safe_len rather than
24742         loop->force_vectorize to check whether there is no alias.
24744 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24746         PR target/84614
24747         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
24748         prototypes.
24749         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
24750         comments.
24751         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
24752         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
24753         instead of a loop around prev_real_insn.
24754         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
24755         prev_real_insn.
24757         PR inline-asm/84625
24758         * config/i386/i386.c (ix86_print_operand): Use conditional
24759         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
24760         zero vector.
24762 2018-03-02  Richard Biener  <rguenther@suse.de>
24764         PR tree-optimization/84427
24765         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
24766         (bitmap_set_subtract_values): Rewrite to handle multiple
24767         exprs per value.
24768         (clean): Likewise.
24769         (prune_clobbered_mems): Likewise.
24770         (phi_translate): Take edge instead of pred/phiblock.
24771         (phi_translate_1): Likewise.
24772         (phi_translate_set): Likewise.  Insert all translated
24773         exprs for a value into the set, keeping possibly multiple
24774         expressions per value.
24775         (compute_antic_aux): Adjust for phi_translate changes.
24776         When intersecting union the expressions and prune those
24777         not in the final value set, keeping possibly multiple
24778         expressions per value.  Do not use value-insertion
24779         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
24780         all expressions.  Add verification that the value-sets
24781         only shrink during iteration.
24782         (compute_partial_antic_aux): Adjust for the phi_translate changes.
24783         (do_pre_regular_insertion): Likewise.
24784         (do_pre_partial_partial_insertion): Likewise.
24786 2018-03-02  Richard Biener  <rguenther@suse.de>
24788         PR target/82005
24789         * config/darwin.c (saved_debug_info_level): New static global.
24790         (darwin_asm_lto_start): Disable debug info generation for LTO out.
24791         (darwin_asm_lto_end): Restore debug info generation settings.
24793 2018-03-01  Martin Liska  <mliska@suse.cz>
24795         PR sanitizer/82484
24796         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
24797         volatile arguments.
24799 2018-03-01  Richard Biener  <rguenther@suse.de>
24801         PR debug/84645
24802         * dwarf2out.c (gen_variable_die): Properly handle late VLA
24803         type annotation with LTO when debug was disabled at compile-time.
24805 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
24807         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
24808         XINT with INTVAL.
24809         (mips_final_postscan_insn): Likewise.
24811 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
24813         PR rtl-optimization/84528
24814         * alias.c (init_alias_target): Add commentary.
24815         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
24816         a unique base value if the frame pointer is not eliminated
24817         to the stack pointer.
24819 2018-03-01  Tom de Vries  <tom@codesourcery.com>
24821         PR rtl-optimization/83327
24822         * lra-int.h (hard_regs_spilled_into): Declare.
24823         * lra.c (hard_regs_spilled_into): Define.
24824         (init_reg_info): Init hard_regs_spilled_into.
24825         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
24826         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
24827         (process_bb_lives): Handle hard_regs_spilled_into.
24828         (lra_create_live_ranges_1): Before doing liveness propagation, clear
24829         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
24831 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
24833         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
24834         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
24835         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
24836         * config/rs6000/aix72.h: New file.
24838 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
24840         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
24841         instead of warning_at with conditional singular and plural messages
24842         where possible.
24844         PR target/52991
24845         * stor-layout.c (update_alignment_for_field): For
24846         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
24847         && !DECL_PACKED (field), do the alignment update, just use
24848         only desired_align instead of MAX (type_align, desired_align)
24849         as the alignment.
24850         (place_field): Don't do known_align < desired_align handling
24851         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
24852         is non-NULL, instead do it after rli->prev_field handling and
24853         only if not within a bitfield word.  For DECL_PACKED (field)
24854         use type_align of BITS_PER_UNIT.
24856 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
24858         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
24859         superfluous parentheses and trailing spaces.
24861 2018-02-28  Richard Biener  <rguenther@suse.de>
24863         PR tree-optimization/84584
24864         * graphite-scop-detection.c (scop_detection::add_scop): Discard
24865         SCoPs with fake exit edge.
24867 2018-02-28  Martin Liska  <mliska@suse.cz>
24869         PR testsuite/84597
24870         * timevar.c (timer::print): Fix format to properly print 100%
24871         values.
24873 2018-02-28  Richard Biener  <rguenther@suse.de>
24875         PR middle-end/84607
24876         * genmatch.c (capture_info::walk_match): Do not mark
24877         captured expressions without operands as expr_p given
24878         they act more like predicates and should be subject to
24879         "lost tail" side-effect preserving.
24881 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
24883         PR rtl-optimization/81611
24884         * auto-inc-dec.c (attempt_change): Move dead note from
24885         mem_insn if it's the next use of regno
24886         (find_address): Take address use of reg holding
24887         non-incremented value.  Add parm to limit search to the named
24888         reg only.
24889         (merge_in_block): Attempt to use a mem insn that is the next
24890         use of the original regno.
24892 2018-02-27  Martin Sebor  <msebor@redhat.com>
24894         PR c++/83871
24895         * doc/invoke.texi (-Wmissing-attributes): New option.
24896         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
24898 2018-02-27  Martin Sebor  <msebor@redhat.com>
24900         PR translation/84207
24901         * diagnostic-core.h (warning_n, error_n, inform_n): Change
24902         n argument to unsigned HOST_WIDE_INT.
24903         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
24904         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
24905         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
24906         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
24908 2018-02-27  Richard Biener  <rguenther@suse.de>
24910         PR tree-optimization/84512
24911         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
24912         Do not use the estimate returned from record_stmt_cost for
24913         the scalar iteration cost but sum properly using add_stmt_cost.
24915 2018-02-27  Richard Biener  <rguenther@suse.de>
24917         PR tree-optimization/84466
24918         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24919         Adjust last change to less strictly validate use operands.
24921 2018-02-27  Martin Liska  <mliska@suse.cz>
24923         PR gcov-profile/84548
24924         * gcov.c (process_file): Allow partial overlap and consider it
24925         also as group functions.
24926         (output_lines): Properly calculate range of lines for a group.
24928 2018-02-27  Martin Liska  <mliska@suse.cz>
24930         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
24931         'ggc' suffixes.  Change first column width.
24932         (timer::print): Fix formatting of the column.
24934 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
24936         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
24937         preserve inline entry blocks for the sake of debug inline
24938         entry point markers alone.
24939         (remove_unused_locals): Suggest in comments a better place to
24940         force the preservation of inline entry blocks that are
24941         otherwise unused, but do not preserve them.
24943 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24945         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
24947 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24949         PR target/84039
24950         * config/i386/constraints.md (Bs): Replace
24951         ix86_indirect_branch_register with
24952         TARGET_INDIRECT_BRANCH_REGISTER.
24953         (Bw): Likewise.
24954         * config/i386/i386.md (indirect_jump): Likewise.
24955         (tablejump): Likewise.
24956         (*sibcall_memory): Likewise.
24957         (*sibcall_value_memory): Likewise.
24958         Peepholes of indirect call and jump via memory: Likewise.
24959         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
24960         (*sibcall_value_GOT_32): Likewise.
24961         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24962         (GOT_memory_operand): Likewise.
24963         (call_insn_operand): Likewise.
24964         (sibcall_insn_operand): Likewise.
24965         (GOT32_symbol_operand): Likewise.
24966         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
24968 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24970         PR rtl-optimization/83496
24971         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
24972         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
24973         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
24974         redundant insn, if any.
24975         (relax_delay_slots): Likewise.
24976         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
24978 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
24980         PR tree-optimization/83965
24981         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
24982         that grouped statements are part of a reduction chain.  Return
24983         true if the statement is not marked as a reduction itself but
24984         is part of a group.
24985         (vect_recog_dot_prod_pattern): Don't check whether the statement
24986         is part of a group here.
24987         (vect_recog_sad_pattern): Likewise.
24988         (vect_recog_widen_sum_pattern): Likewise.
24990 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24992         PR debug/84545
24993         * final.c (rest_of_clean_state): Also look for calls inside sequences.
24995 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24997         PR target/84530
24998         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
24999         the bool argument.
25000         (ix86_output_indirect_function_return): New prototype.
25001         (ix86_split_simple_return_pop_internal): Likewise.
25002         * config/i386/i386.c (indirect_return_via_cx): New.
25003         (indirect_return_via_cx_bnd): Likewise.
25004         (indirect_thunk_name): Handle return va CX_REG.
25005         (output_indirect_thunk_function): Create alias for
25006         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
25007         (ix86_output_indirect_jmp): Remove the bool argument.
25008         (ix86_output_indirect_function_return): New function.
25009         (ix86_split_simple_return_pop_internal): Likewise.
25010         * config/i386/i386.md (*indirect_jump): Don't pass false
25011         to ix86_output_indirect_jmp.
25012         (*tablejump_1): Likewise.
25013         (simple_return_pop_internal): Change it to define_insn_and_split.
25014         Call ix86_split_simple_return_pop_internal to split it for
25015         -mfunction-return=.
25016         (simple_return_indirect_internal): Call
25017         ix86_output_indirect_function_return instead of
25018         ix86_output_indirect_jmp.
25020 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
25022         PR bootstrap/84405
25023         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
25024         memset and value initialization afterwards.
25026 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
25028         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
25030 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25032         PR target/84521
25033         * common/config/aarch64/aarch64-common.c
25034         (aarch_option_optimization_table[]): Switch
25035         off fomit-frame-pointer
25037 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
25038             Chung-Ju Wu  <jasonwucj@gmail.com>
25040         * config/nds32/nds32-multiple.md (load_multiple): Disallow
25041         volatile memory.
25042         (store_multiple): Ditto.
25044 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
25046         * config.gcc: Add --with-cpu support for nds32 target.
25047         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
25048         * config/nds32/nds32.opt: Add -mcpu= option.
25050 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
25052         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
25053         isel=yes): Warn for these deprecated options.
25055 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
25057         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
25058         ISA_2_5_MASKS_EMBEDDED.
25060 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
25062         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
25063         p->max as pointers rather than using iterative_hash_expr.
25065 2018-02-23  Carl Love  <cel@us.ibm.com>
25067         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
25068         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
25069         BU_P8V_OVERLOAD_2.
25070         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
25071         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
25072         P8V_BUILTIN_VEC_VUNSIGNED2.
25074 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
25076         PR target/81572
25077         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
25078         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
25079         LRA_UNKNOWN_ALT.
25080         * lra-constraints.c (curr_insn_transform): Set up
25081         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
25082         LRA_UNKNOWN_ALT.
25083         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
25084         * lra-eliminations.c (spill_pseudos): Ditto.
25085         (process_insn_for_elimination): Ditto.
25086         * lra-lives.c (reg_early_clobber_p): Use the new macros.
25087         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
25088         LRA_NON_CLOBBERED_ALT.
25090 2018-02-22  Martin Sebor  <msebor@redhat.com>
25092         PR tree-optimization/84480
25093         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
25094         to maybe_diag_stxncpy_trunc.  Call it.
25095         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
25096         from gimple_fold_builtin_strcpy.  Print inlining stack.
25097         (handle_builtin_stxncpy): Print inlining stack.
25098         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
25100 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
25102         PR target/84176
25103         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
25104         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
25105         and -fcheck-pointer-bounds are used together.
25106         (indirect_thunk_prefix): New enum.
25107         (indirect_thunk_need_prefix): New function.
25108         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
25109         "_nt" instead of "_bnd" for NOTRACK prefix.
25110         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
25111         (output_indirect_thunk_function): Likewise.
25112         (): Likewise.
25113         (ix86_code_end): Update output_indirect_thunk_function calls.
25114         (ix86_output_indirect_branch_via_reg): Replace
25115         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
25116         (ix86_output_indirect_branch_via_push): Likewise.
25117         (ix86_output_function_return): Likewise.
25118         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
25119         incompatible with -fcf-protection=branch and
25120         -fcheck-pointer-bounds.
25122 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
25124         PR target/83335
25125         * config/aarch64/aarch64.c (aarch64_print_address_internal):
25126         Change gcc_assert call to output_operand_lossage.
25128 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
25130         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
25132 2018-02-22  DJ Delorie  <dj@redhat.com>
25133             Sebastian Perta  <sebastian.perta@renesas.com>
25134             Oleg Endo  <olegendo@gcc.gnu.org>
25136         * config/rx/rx.c (rx_rtx_costs): New function.
25137         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
25139 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25141         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
25143 2018-02-22  Martin Liska  <mliska@suse.cz>
25145         PR driver/83193
25146         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
25147         Add "native" as a possible value.
25149 2018-02-22  Martin Liska  <mliska@suse.cz>
25151         PR driver/83193
25152         * config/i386/i386.c (ix86_option_override_internal):
25153         Add "native" as a possible value for -march and -mtune.
25155 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
25157         PR target/84502
25158         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
25159         to all type variants.
25161         PR tree-optimization/84503
25162         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
25163         width as info->bitpos + info->bitsize - start.
25164         (merged_store_group::merge_overlapping): Simplify width computation.
25165         (check_no_overlap): New function.
25166         (imm_store_chain_info::try_coalesce_bswap): Compute expected
25167         start + width and last_order of the group, fail if check_no_overlap
25168         fails.
25169         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
25170         to group if check_no_overlap fails.
25172 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
25174         * config/rs6000/altivec.md: Delete contraint arguments to
25175         define_expand, define_split, and define_peephole2, and in
25176         define_insn_and_split if always unused.
25177         * config/rs6000/darwin.md: Ditto.
25178         * config/rs6000/dfp.md: Ditto.
25179         * config/rs6000/rs6000.md: Ditto.
25180         * config/rs6000/sync.md: Ditto.
25181         * config/rs6000/vector.md: Ditto.
25182         * config/rs6000/vsx.md: Ditto.
25184 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
25186         * config/rs6000/altivec.md: Write output control strings as braced
25187         blocks instead of double-quoted strings.
25188         * config/rs6000/darwin.md: Ditto.
25189         * config/rs6000/rs6000.md: Ditto.
25190         * config/rs6000/vector.md: Ditto.
25191         * config/rs6000/vsx.md: Ditto.
25193 2018-02-21  Jason Merrill  <jason@redhat.com>
25195         PR c++/84314 - ICE with templates and fastcall attribute.
25196         * attribs.c (build_type_attribute_qual_variant): Remove assert.
25198 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
25200         * ipa-cp.c (determine_versionability): Fix comment typos.
25202 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
25204         PR c/84229
25205         * ipa-cp.c (determine_versionability): Do not version functions caling
25206         va_arg_pack.
25208 2018-02-21  Martin Liska  <mliska@suse.cz>
25210         PR driver/83193
25211         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
25212         Add "native" as a possible value.
25213         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
25214         the macro when native cpu detection is available.
25216 2018-02-21  Martin Liska  <mliska@suse.cz>
25218         PR driver/83193
25219         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
25220         Add "native" as a possible value.
25221         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
25222         when native cpu detection is available.
25224 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
25225             Martin Sebor  <msebor@redhat.com>
25227         PR tree-optimization/84478
25228         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
25229         false.
25230         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
25231         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
25232         support which is conservatively correct, for 2 only stay conservative
25233         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
25234         argument to the 2 argument get_range_strlen, adjust 6 arg
25235         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
25236         false.
25237         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
25238         (gimple_fold_builtin_strlen): Pass true as last argument to
25239         get_range_strlen.
25241 2018-02-20  Martin Sebor  <msebor@redhat.com>
25243         PR middle-end/84095
25244         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
25245         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
25246         (builtin_memref::builtin_memref): Factor out parts into
25247         set_base_and_offset and call it.
25249 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
25251         PR middle-end/84406
25252         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
25253         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
25254         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
25255         search at the associated MODE_INT.
25257 2018-02-20  Jeff Law  <law@redhat.com>
25259         PR middle-end/82123
25260         PR tree-optimization/81592
25261         PR middle-end/79257
25262         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
25263         for range data rather than using global data.
25264         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
25265         range data rather than using global data.
25266         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
25267         pass it to children as needed.
25268         (struct directive::fmtresult): Similarly.
25269         (struct directive::set_width): Similarly.
25270         (struct directive::set_precision): Similarly.
25271         (format_integer, format_directive, parse_directive): Similarly.
25272         (format_none): Accept unnamed vr_values parameter.
25273         (format_percent, format_floating, format_character): Similarly.
25274         (format_string, format_plain): Similarly.
25275         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
25276         the EVRP range analyzer for range data rather than using global data.
25277         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
25278         gimple-ssa-evrp-analyze.h
25279         (class sprintf_dom_walker): Add after_dom_children member function.
25280         Add evrp_range_analyzer member.
25281         (sprintf_dom_walker::before_dom_children): Call into the EVRP
25282         range analyzer as needed.
25283         (sprintf_dom_walker::after_dom_children): New member function.
25284         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
25285         if not optimizing.
25286         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
25287         (evrp_range_analyzer::pop_to_marker): Likewise.
25289 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
25291         PR tree-optimization/84419
25292         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
25293         with the required type if its current type is compatible but
25294         different.
25296 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
25298         PR middle-end/82004
25299         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
25300         after vectorization.
25302 2018-02-20  Martin Liska  <mliska@suse.cz>
25304         PR driver/83193
25305         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
25306         possible values if we don't have a hint.
25308 2018-02-20  Martin Liska  <mliska@suse.cz>
25310         PR c/84310
25311         PR target/79747
25312         * final.c (shorten_branches): Build align_tab array with one
25313         more element.
25314         * opts.c (finish_options): Add alignment option limit check.
25315         (MAX_CODE_ALIGN): Likewise.
25316         (MAX_CODE_ALIGN_VALUE): Likewise.
25317         * doc/invoke.texi: Document maximum allowed option value for
25318         all -falign-* options.
25320 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
25322         PR target/84146
25323         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
25324         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
25325         * var-tracking.c (emit_note_insn_var_location): Remove all references
25326         to NOTE_INSN_CALL_ARG_LOCATION.
25327         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
25328         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
25329         Use copy_rtx_if_shared.
25330         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
25331         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
25332         (dwarf2out_var_location): Remove handling of
25333         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
25334         on call_insn.
25335         * final.c (final_scan_insn): Remove all references to
25336         NOTE_INSN_CALL_ARG_LOCATION.
25337         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
25338         before dumping final insns.
25339         * except.c (emit_note_eh_region_end): Remove all references to
25340         NOTE_INSN_CALL_ARG_LOCATION.
25341         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
25342         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
25343         * config/arc/arc.c (hwloop_optimize): Likewise.
25344         * config/arm/arm.c (create_fix_barrier): Likewise.
25345         * config/s390/s390.c (s390_chunkify_start): Likewise.
25346         * config/sh/sh.c (find_barrier): Likewise.
25347         * config/i386/i386.c (rest_of_insert_endbranch,
25348         ix86_seh_fixup_eh_fallthru): Likewise.
25349         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
25350         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
25351         * config/frv/frv.c (frv_function_prologue): Likewise.
25352         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
25353         reg note.
25354         (note_outside_basic_block_p): Remove all references to
25355         NOTE_INSN_CALL_ARG_LOCATION.
25356         * gengtype.c (adjust_field_rtx_def): Likewise.
25357         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
25358         Likewise.
25359         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
25360         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
25362         PR c++/84444
25363         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
25364         is ADDR_EXPR.
25366         PR tree-optimization/84452
25367         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
25368         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
25369         is NULL.
25371 2018-02-19  Martin Liska  <mliska@suse.cz>
25373         PR sanitizer/82183
25374         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
25376 2018-02-19  Martin Liska  <mliska@suse.cz>
25377             Richard Sandiford  <richard.sandiford@linaro.org>
25379         PR tree-optimization/82491
25380         * gimple-fold.c (get_base_constructor): Make earlier bail out
25381         to prevent ubsan.
25383 2018-02-19  Carl Love  <cel@us.ibm.com>
25385         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
25386         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
25387         BU_P8V_OVERLOAD_1.
25388         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
25389         P8V_BUILTIN_VEC_NEG.
25391 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
25393         * config/rl78/rl78.md (movdf): New define expand.
25395 2018-02-19  Martin Liska  <mliska@suse.cz>
25397         PR other/80589
25398         * doc/invoke.texi: Fix typo.
25399         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
25401 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
25403         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
25404         handle rs6000_single_float and rs6000_double_float specially for
25405         e500 family CPUs.
25407 2018-02-16  Jeff Law  <law@redhat.com>
25409         * config/rx/rx.c (add_pop_cfi_notes): New function.;
25410         (pop_regs): Use it.
25412 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
25414         PR ipa/84425
25415         * ipa-inline.c (inline_small_functions): Fix a typo.
25417 2018-02-16  Nathan Sidwell  <nathan@acm.org>
25419         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
25421 2018-02-16  Carl Love  <cel@us.ibm.com>
25423         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
25424         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
25425         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
25426         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
25427         expansion to P8V_BUILTIN_VEC_FLOAT2.
25429 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
25431         PR rtl-optimization/70023
25432         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
25433         src_regno into account.
25435 2018-02-16  Carl Love  <cel@us.ibm.com>
25437         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
25438         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
25439         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
25440         * config/rs6000/rs6000.c: Remove case statements for
25441         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
25442         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
25443         and P9V_BUILTIN_VEC_VINSERT4B.
25444         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
25445         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
25446         * config/rs6000/vsx.md:
25447         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
25448         vec_insert4b.
25450 2018-02-16  Carl Love  <cel@us.ibm.com>
25452         * config/rs6000/altivec.h: Add builtin names vec_extract4b
25453         vec_insert4b.
25454         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
25455         definitions.
25456         * config/rs6000/rs6000-c.c: Add the definitions for
25457         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
25458         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
25459         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
25460         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
25461         definition for insert4b and define insn *insert3b_internal.
25462         * doc/extend.texi: Add documentation for vec_extract4b.
25464 2018-02-16  Nathan Sidwell  <nathan@acm.org>
25466         * doc/extend.texi (Backwards Compatibility): Mention friend
25467         injection.  Note for-scope is deprecated.
25468         * doc/invoke.texi (-ffriend-injection): Deprecate.
25470 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
25472         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
25473         that moved to I2, also allow destinations that are a paradoxical
25474         subreg (instead of a normal reg).
25476 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
25478         PR target/83831
25479         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
25480         to QImode.
25482 2018-02-16  Richard Biener  <rguenther@suse.de>
25484         PR tree-optimization/84037
25485         PR tree-optimization/84016
25486         PR target/82862
25487         * config/i386/i386.c (ix86_builtin_vectorization_cost):
25488         Adjust vec_construct for the fact we need additional higher latency
25489         128bit inserts for AVX256 and AVX512 vector builds.
25490         (ix86_add_stmt_cost): Scale vector construction cost for
25491         elementwise loads.
25493 2018-02-16  Richard Biener  <rguenther@suse.de>
25495         PR tree-optimization/84417
25496         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
25497         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
25498         (non_rewritable_lvalue_p): Likewise, use poly-ints.
25500 2018-02-16  Martin Liska  <mliska@suse.cz>
25502         PR sanitizer/84307
25503         * internal-fn.def (ASAN_CHECK): Set proper flags.
25504         (ASAN_MARK): Likewise.
25506 2018-02-16  Julia Koval  <julia.koval@intel.com>
25508         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
25509         from PTA_CANNONLAKE.
25511 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
25513         PR target/84272
25514         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
25515         Use ++iter rather than iter++ for std::list iterators.
25516         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
25517         defer deleting them until all nodes in the forest are processed.  Do
25518         free even leaf nodes.  Change to_process into auto_vec.
25520         PR bootstrap/84405
25521         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
25522         * vec.h (vec_default_construct): Use memset instead of placement new
25523         if BROKEN_VALUE_INITIALIZATION is defined.
25524         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
25525         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
25526         is defined.
25528         PR rtl-optimization/83723
25529         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
25530         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
25531         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
25532         recursive calls.
25533         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
25534         callers.
25535         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
25537 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
25539         PR rtl-optimization/81443
25540         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
25541         from inner REGs to paradoxical SUBREGs.
25543 2018-02-16  Richard Biener  <rguenther@suse.de>
25545         PR tree-optimization/84399
25546         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
25547         For operands we can analyze at their definition make sure we can
25548         analyze them at each use as well.
25550 2018-02-16  Richard Biener  <rguenther@suse.de>
25552         PR tree-optimization/84190
25553         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
25554         volatile accesses if the decl isn't volatile.
25556 2018-02-15  Jason Merrill  <jason@redhat.com>
25558         PR c++/84314 - ICE with templates and fastcall attribute.
25559         * attribs.c (build_type_attribute_qual_variant): Don't clobber
25560         TYPE_CANONICAL on an existing type.
25562 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
25564         PR tree-optimization/84383
25565         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
25566         dstoff nor call operand_equal_p if dstbase is NULL.
25568         PR tree-optimization/84334
25569         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
25570         also a CONSTANT_CLASS_P, punt.
25572 2018-02-14  Jim Wilson  <jimw@sifive.com>
25574         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
25575         first SMALL_OPERAND check.  New local min_second_step.  Move assert
25576         to where locals are set.  Add TARGET_RVC support.
25577         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
25579 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
25581         * doc/invoke.texi: Correct -Wformat-overflow code sample.
25583 2018-02-14  Martin Sebor  <msebor@redhat.com>
25585         PR tree-optimization/83698
25586         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
25587         arrays constrain the offset range to their bounds.
25588         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
25589         (builtin_access::overlap): Avoid setting the size of overlap if it's
25590         already been set.
25591         (maybe_diag_overlap): Also consider arrays when deciding what values
25592         of offsets to include in diagnostics.
25594 2018-02-14  Martin Sebor  <msebor@redhat.com>
25596         PR c/84108
25597         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
25598         that correspond to the kind of a declaration.
25600 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
25602         PR target/83984
25603         * config/pa/pa.md: Load address of PIC label using the linkage table
25604         if the label is nonlocal.
25606 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25608         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
25609         warning message if user requests -maltivec=be.
25610         * doc/invoke.texi: Document deprecation of -maltivec=be.
25612 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
25614         PR target/84220
25615         * config/rs6000/rs6000-c.c: Update definitions for
25616         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
25617         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
25619 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25621         PR target/84239
25622         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
25623         add _get_ssp intrinsics. Remove argument from
25624         __builtin_ia32_rdssp[d|q].
25625         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
25626         * config/i386/i386-builtin.def: Remove argument from
25627         __builtin_ia32_rdssp[d|q].
25628         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
25629         ix86_expand_special_args_builtin for _rdssp[d|q].
25630         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
25631         Clear register before usage.
25632         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
25633         Add documentation for new _get_ssp and _inc_ssp intrinsics.
25635 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
25637         PR tree-optimization/84357
25638         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
25639         operand 1 of an ARRAY_REF too.
25641 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
25643         PR target/83831
25644         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
25645         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25646         declarations.
25647         (set_of_reg): New struct.
25648         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
25649         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
25650         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25651         functions.
25652         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
25653         Split into bitclr, bitset, bitinvert patterns if appropriate.
25654         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
25655         use rx_fuse_in_memory_bitop.
25656         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
25657         to named insn, correct maximum insn length.
25659 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
25661         PR target/79242
25662         * machmode.def: Define a complex mode for PARTIAL_INT.
25663         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
25664         MODE_PARTIAL_INT.
25665         * doc/rtl.texi: Document CSPImode.
25666         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
25667         handling.
25668         (msp430_hard_regno_nregs_with_padding): Likewise.
25670 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
25672         PR target/84279
25673         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
25675 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
25677         PR rtl-optimization/84169
25678         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
25679         we generated a parallel as new i3 and we split that to new i2 and i3
25680         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
25681         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
25682         those to i2, not i1.  Partially rewrite this scan code.
25684 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25686         PR c/82210
25687         * stor-layout.c (place_field): For variable length fields, adjust
25688         offset_align afterwards not just based on the field's alignment,
25689         but also on the size.
25691         PR middle-end/84309
25692         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
25693         of exps and logs in the use_exp2 case.
25695 2018-02-13  Jeff Law  <law@redhat.com>
25697         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
25698         entry for "vector".
25700         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
25701         ARGS as unused.
25703 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
25705         PR debug/84342
25706         PR debug/84319
25707         * common.opt (gas-loc-support, gas-locview-support): New.
25708         (ginline-points, ginternal-reset-location-views): New.
25709         * doc/invoke.texi: Document them.  Use @itemx where intended.
25710         (gvariable-location-views): Adjust.
25711         * target.def (reset_location_view): New.
25712         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
25713         (TARGET_RESET_LOCATION_VIEW): New.
25714         * doc/tm.texi: Rebuilt.
25715         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
25716         (dwarf2out_default_as_locview_support): New.
25717         (output_asm_line_debug_info): Use option variables.
25718         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
25719         (output_loc_list): Likewise.
25720         (add_high_low_attributes): Check option variables.
25721         Don't output entry view attribute in strict mode.
25722         (gen_inlined_subroutine_die): Check option variables.
25723         (dwarf2out_inline_entry): Likewise.
25724         (init_sections_and_labels): Likewise.
25725         (dwarf2out_early_finish): Likewise.
25726         (maybe_reset_location_view): New, from...
25727         (dwarf2out_var_location): ... here.  Call it.
25728         * debug.h (dwarf2out_default_as_loc_support): Declare.
25729         (dwarf2out_default_as_locview_support): Declare.
25730         * hooks.c (hook_int_rtx_insn_0): New.
25731         * hooks.h (hook_int_rtx_insn_0): Declare.
25732         * toplev.c (process_options): Take -gas-loc-support and
25733         -gas-locview-support from dwarf2out.  Enable
25734         -gvariable-location-views by default only with locview
25735         assembler support.  Enable -ginternal-reset-location-views by
25736         default only if the target defines the corresponding hook.
25737         Enable -ginline-points by default if location views are
25738         enabled; force it disabled if statement frontiers are
25739         disabled.
25740         * tree-inline.c (expand_call_inline): Check option variables.
25741         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
25743 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
25745         PR tree-optimization/84321
25746         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
25747         handling.  Also check whether the anti-range contains any values
25748         that satisfy the mask; switch to a VR_RANGE if not.
25750 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
25752         PR sanitizer/84340
25753         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
25755 2018-02-13  Martin Jambor  <mjambor@suse.cz>
25757         PR c++/83990
25758         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
25759         of call statements, also set location of a load to a temporary.
25761 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
25763         * config/rl78/rl78.c (add_vector_labels): New function.
25764         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
25765         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
25766         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
25767         which checks that no arguments are passed.
25768         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
25769         * doc/extend.texi: Documentation for the new attribute.
25771 2018-02-13  Andreas Schwab  <schwab@suse.de>
25773         * config/riscv/linux.h (CPP_SPEC): Define.
25775 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25777         PR target/84335
25778         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
25779         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
25780         OPTION_MASK_ISA_AES as first argument to def_builtin_const
25781         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
25782         instead of OPTION_MASK_ISA_PCLMUL as first argument to
25783         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
25784         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
25785         temporarily for AES and PCLMUL builtins.
25787         PR tree-optimization/84339
25788         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
25789         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
25790         Formatting fixes.
25792         PR middle-end/84309
25793         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
25794         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
25795         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
25796         inline function.
25797         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
25798         inline function.
25799         * omp-simd-clone.h: New file.
25800         * omp-simd-clone.c: Include omp-simd-clone.h.
25801         (expand_simd_clones): No longer static.
25802         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
25803         cgraph.h and omp-simd-clone.h.
25804         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
25805         (vect_recog_widen_shift_pattern): Formatting fix.
25806         (vect_pattern_recog_1): Don't check optab for calls.
25808         PR target/84336
25809         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
25810         operands[2] into a REG before using gen_lowpart on it.
25812 2018-02-12  Jeff Law  <law@redhat.com>
25814         PR target/83760
25815         * config/sh/sh.c (find_barrier): Consider a sibling call
25816         a barrier as well.
25818         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
25819         successfully back substituting a reg.
25821 2018-02-12  Richard Biener  <rguenther@suse.de>
25823         PR tree-optimization/84037
25824         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
25825         parameter, move visited init to caller.
25826         (vect_slp_analyze_operations): Separate cost from validity
25827         check, initialize visited once for all instances.
25828         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
25829         for all instances.
25830         * tree-vect-stmts.c (vect_model_simple_cost): Make early
25831         out an assert.
25832         (vect_model_promotion_demotion_cost): Likewise.
25833         (vectorizable_bswap): Guard cost modeling with !slp_node
25834         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
25835         SLP stmts.
25836         (vectorizable_call): Likewise.
25837         (vectorizable_conversion): Likewise.
25838         (vectorizable_assignment): Likewise.
25839         (vectorizable_shift): Likewise.
25840         (vectorizable_operation): Likewise.
25841         (vectorizable_store): Likewise.
25842         (vectorizable_load): Likewise.
25843         (vectorizable_condition): Likewise.
25844         (vectorizable_comparison): Likewise.
25846 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
25848         PR sanitizer/84307
25849         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
25850         (ASAN_MARK): Fix fnspec to account for return value, change pointer
25851         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
25853 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
25855         PR middle-end/83665
25856         * params.def (inline-min-speedup): Increase from 8 to 15.
25857         (max-inline-insns-auto): Decrease from 40 to 30.
25858         * ipa-split.c (consider_split): Add some buffer for function to
25859         be considered inlining candidate.
25860         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
25861         default values.
25863 2018-02-12  Richard Biener  <rguenther@suse.de>
25865         PR tree-optimization/84037
25866         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
25867         matched stmts if we cannot swap the non-matched ones.
25869 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25871         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
25872         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
25873         _mm_maskz_scalef_round_ss): New intrinsics.
25874         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
25875         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
25876         __builtin_ia32_scalefss_round): Remove.
25877         (__builtin_ia32_scalefsd_mask_round,
25878         __builtin_ia32_scalefss_mask_round): New intrinsics.
25879         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
25880         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
25881         ((match_operand:VF_128 2 "<round_nimm_predicate>"
25882         "<round_constraint>")): Changed to ...
25883         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
25884         "<round_scalar_constraint>")): ... this.
25885         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
25886         %0, %1, %2<round_op3>}"): Changed to ...
25887         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
25888         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
25889         %2<round_scalar_mask_op3>}"): ... this.
25890         * config/i386/subst.md (round_scalar_nimm_predicate): New.
25892 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25894         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
25895         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
25896         (_mm_maskz_sqrt_round_ss): New intrinsics.
25897         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
25898         (__builtin_ia32_sqrtsd_mask_round)
25899         (__builtin_ia32_sqrtss_mask_round): New builtins.
25900         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
25901         (__builtin_ia32_sqrtss_round): Remove.
25902         (__builtin_ia32_sqrtsd_mask_round)
25903         (__builtin_ia32_sqrtss_mask_round): New builtins.
25904         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
25905         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
25906         ((match_operand:VF_128 1 "vector_operand"
25907         "xBm,<round_constraint>")): Changed to ...
25908         ((match_operand:VF_128 1 "vector_operand"
25909         "xBm,<round_scalar_constraint>")): ... this.
25910         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
25911         %0, %2, %<iptr>1<round_op3>}): Changed to ...
25912         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
25913         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
25914         %<iptr>1<round_scalar_mask_op3>}): ... this.
25915         ((set_attr "prefix" "<round_prefix>")): Changed to ...
25916         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
25918 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
25920         PR target/84266
25921         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
25922         Cast vec_cmpeq result to correct type.
25923         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
25924         Cast vec_cmpgt result to correct type.
25926 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
25928         * final.c (final_scan_insn_1): Renamed from...
25929         (final_scan_insn): ... this.  New wrapper, to recover
25930         seen from the outermost call in recursive ones.
25931         * config/sparc/sparc.c (output_return): Drop seen from call.
25932         (output_sibcall): Likewise.
25933         * config/visium/visium.c (output_branch): Likewise.
25935 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
25937         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
25938         function label.
25940 2018-02-10  Alan Modra  <amodra@gmail.com>
25942         PR target/84300
25943         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
25944         Specify LR as an input.
25946 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
25948         PR sanitizer/83987
25949         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
25950         remove_member_access_dummy_vars): New functions.
25951         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
25952         lower_omp_1, execute_lower_omp): Use them.
25954         PR rtl-optimization/84308
25955         * shrink-wrap.c (spread_components): Release todo vector.
25957 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
25959         PR rtl-optimization/57193
25960         * ira-color.c (struct allocno_color_data): Add member
25961         conflict_allocno_hard_prefs.
25962         (update_conflict_allocno_hard_prefs): New.
25963         (bucket_allocno_compare_func): Add a preference based on
25964         conflict_allocno_hard_prefs.
25965         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
25966         (color_allocnos): Remove a dead code.  Initiate
25967         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
25969 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25971         PR target/84226
25972         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
25973         constraint from =wa to wa.  Avoid a subreg on the output operand,
25974         instead use a pseudo and subreg it in a move.
25975         (p9_xxbrd_<mode>): Changed to ...
25976         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
25977         (p9_xxbrd_v2df): New expander.
25978         (p9_xxbrw_<mode>): Changed to ...
25979         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
25980         (p9_xxbrw_v4sf): New expander.
25982 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25984         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
25986 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
25988         PR target/83926
25989         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
25990         multiply in 32-bit mode.
25991         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
25992         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
25993         mode.
25995 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25997         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
25998         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
25999         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
26000         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
26002 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
26004         PR lto/84213
26005         * dwarf2out.c (is_trivial_indirect_ref): New function.
26006         (dwarf2out_late_global_decl): Do not generate a location
26007         attribute for variables that have a non-trivial DECL_VALUE_EXPR
26008         and that are not defined in the current unit.
26010 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
26012         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
26013         instead of a libcall for UNORDERED.
26015 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
26017         PR target/82641
26018         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
26019         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
26021 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26023         PR target/PR84295
26024         * config/s390/s390.c (s390_set_current_function): Invoke
26025         s390_indirect_branch_settings also if fndecl didn't change.
26027 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
26029         * config/rs6000/rs6000.md (blockage): Set length to zero.
26031 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
26033         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
26035 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
26037         PR sanitizer/84285
26038         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
26039         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
26040         -static-lib*san.
26042         PR debug/84252
26043         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
26044         PARALLEL incoming that failed vt_get_decl_and_offset check.
26046         PR middle-end/84237
26047         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
26048         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
26049         TREE_READONLY bit.
26050         (get_variable_section): For decls in named .bss* sections pass true as
26051         second argument to bss_initializer_p.
26053 2018-02-09  Marek Polacek  <polacek@redhat.com>
26054             Jakub Jelinek  <jakub@redhat.com>
26056         PR c++/83659
26057         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
26058         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
26059         Sync some changes from cxx_fold_indirect_ref.
26061 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
26063         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
26064         markers.
26065         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
26066         (BLOCK_INLINE_ENTRY_LABEL): New.
26067         (dwarf2out_var_location): Disregard inline entry markers.
26068         (inline_entry_data): New struct.
26069         (inline_entry_data_hasher): New hashtable type.
26070         (inline_entry_data_hasher::hash): New.
26071         (inline_entry_data_hasher::equal): New.
26072         (inline_entry_data_table): New variable.
26073         (add_high_low_attributes): Add DW_AT_entry_pc and
26074         DW_AT_GNU_entry_view attributes if a pending entry is found
26075         in inline_entry_data_table.  Add old entry_pc attribute only
26076         if debug nonbinding markers are disabled.
26077         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
26078         markers are enabled.
26079         (block_within_block_p, dwarf2out_inline_entry): New.
26080         (dwarf2out_finish): Check that no entries remained in
26081         inline_entry_data_table.
26082         * final.c (reemit_insn_block_notes): Handle inline entry notes.
26083         (final_scan_insn, notice_source_line): Likewise.
26084         (rest_of_clean_state): Skip inline entry markers.
26085         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
26086         markers.
26087         * gimple.c (gimple_build_debug_inline_entry): New.
26088         * gimple.h (enum gimple_debug_subcode): Add
26089         GIMPLE_DEBUG_INLINE_ENTRY.
26090         (gimple_build_debug_inline_entry): Declare.
26091         (gimple_debug_inline_entry_p): New.
26092         (gimple_debug_nonbind_marker_p): Adjust.
26093         * insn-notes.def (INLINE_ENTRY): New.
26094         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
26095         inline entry marker notes.
26096         (print_insn): Likewise.
26097         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
26098         (INSN_DEBUG_MARKER_KIND): Likewise.
26099         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
26100         * tree-inline.c (expand_call_inline): Build and insert
26101         debug_inline_entry stmt.
26102         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
26103         inline entry blocks early, if nonbind markers are enabled.
26104         (dump_scope_block): Dump fragment info.
26105         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
26106         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
26107         (gimple_build_debug_inline_entry): New.
26108         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
26109         Enable/disable inline entry points too.
26110         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
26111         (DEBUG_INSN): Describe inline entry markers.
26113         * common.opt (gvariable-location-views): New.
26114         (gvariable-location-views=incompat5): New.
26115         * config.in: Rebuilt.
26116         * configure: Rebuilt.
26117         * configure.ac: Test assembler for view support.
26118         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
26119         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
26120         * dwarf2out.c (var_loc_view): New typedef.
26121         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
26122         (dwarf2out_locviews_in_attribute): New.
26123         (dwarf2out_locviews_in_loclist): New.
26124         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
26125         (enum dw_line_info_opcode): Add LI_adv_address.
26126         (struct dw_line_info_table): Add view.
26127         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
26128         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
26129         (zero_view_p): New variable.
26130         (ZERO_VIEW_P): New macro.
26131         (output_asm_line_debug_info): New.
26132         (struct var_loc_node): Add view.
26133         (add_AT_view_list, AT_loc_list): New.
26134         (add_var_loc_to_decl): Add view param.  Test it against last.
26135         (new_loc_list): Add view params.  Record them.
26136         (AT_loc_list_ptr): Handle loc and view lists.
26137         (view_list_to_loc_list_val_node): New.
26138         (print_dw_val): Handle dw_val_class_view_list.
26139         (size_of_die): Likewise.
26140         (value_format): Likewise.
26141         (loc_list_has_views): New.
26142         (gen_llsym): Set vl_symbol too.
26143         (maybe_gen_llsym, skip_loc_list_entry): New.
26144         (dwarf2out_maybe_output_loclist_view_pair): New.
26145         (output_loc_list): Output view list or entries too.
26146         (output_view_list_offset): New.
26147         (output_die): Handle dw_val_class_view_list.
26148         (output_dwarf_version): New.
26149         (output_compilation_unit_header): Use it.
26150         (output_skeleton_debug_sections): Likewise.
26151         (output_rnglists, output_line_info): Likewise.
26152         (output_pubnames, output_aranges): Update version comments.
26153         (output_one_line_info_table): Output view numbers in asm comments.
26154         (dw_loc_list): Determine current endview, pass it to new_loc_list.
26155         Call maybe_gen_llsym.
26156         (loc_list_from_tree_1): Adjust.
26157         (add_AT_location_description): Create view list attribute if
26158         needed, check it's absent otherwise.
26159         (convert_cfa_to_fb_loc_list): Adjust.
26160         (maybe_emit_file): Call output_asm_line_debug_info for test.
26161         (dwarf2out_var_location): Reset views as needed.  Precompute
26162         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
26163         attribute.  Set view.
26164         (new_line_info_table): Reset next view.
26165         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
26166         (dwarf2out_source_line): Likewise.  Output view resets and labels to
26167         the assembler, or select appropriate line info opcodes.
26168         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
26169         (optimize_string_length): Catch it.  Adjust.
26170         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
26171         dw_val_class_view_list, and remove it if no longer needed.
26172         (hash_loc_list): Hash view numbers.
26173         (loc_list_hasher::equal): Compare them.
26174         (optimize_location_lists): Check whether a view list symbol is
26175         needed, and whether the locview attribute is present, and
26176         whether they match.  Remove the locview attribute if no longer
26177         needed.
26178         (index_location_lists): Call skip_loc_list_entry for test.
26179         (dwarf2out_finish): Call output_asm_line_debug_info for test.
26180         Use output_dwarf_version.
26181         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
26182         (struct dw_val_node): Add val_view_list.
26183         * final.c (SEEN_NEXT_VIEW): New.
26184         (set_next_view_needed): New.
26185         (clear_next_view_needed): New.
26186         (maybe_output_next_view): New.
26187         (final_start_function): Rename to...
26188         (final_start_function_1): ... this.  Take pointer to FIRST,
26189         add SEEN parameter.  Emit param bindings in the initial view.
26190         (final_start_function): Reintroduce SEEN-less interface.
26191         (final): Rename to...
26192         (final_1): ... this.  Take SEEN parameter.  Output final pending
26193         next view at the end.
26194         (final): Reintroduce seen-less interface.
26195         (final_scan_insn): Output pending next view before switching
26196         sections or ending a block.  Mark the next view as needed when
26197         outputting variable locations.  Notify debug backend of section
26198         changes, and of location view changes.
26199         (rest_of_handle_final): Adjust.
26200         * toplev.c (process_options): Autodetect value for debug variable
26201         location views option.  Warn on incompat5 without -gdwarf-5.
26202         * doc/invoke.texi (gvariable-location-views): New.
26203         (gvariable-location-views=incompat5): New.
26204         (gno-variable-location-views): New.
26206 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
26208         PR tree-optimization/84136
26209         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
26210         that the result of find_edge is non-NULL.
26212 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
26214         PR target/83008
26215         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
26216         storing integer register in SImode.  Fix cost of 256 and 512
26217         byte aligned SSE register store.
26219 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
26221         * config/i386/i386.c (ix86_multiplication_cost): Fix
26222         multiplication cost for TARGET_AVX512DQ.
26224 2018-02-08  Marek Polacek  <polacek@redhat.com>
26226         PR tree-optimization/84238
26227         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
26228         get_range_strlen.
26230 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26232         PR tree-optimization/84265
26233         * tree-vect-stmts.c (vectorizable_store): Don't treat
26234         VMAT_CONTIGUOUS accesses as grouped.
26235         (vectorizable_load): Likewise.
26237 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26239         PR tree-optimization/81635
26240         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
26241         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
26242         (test_round_for_mask): New functions.
26243         (wide_int_cc_tests): Call test_round_for_mask.
26244         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
26245         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
26246         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
26247         range returned by get_range_info.
26249 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
26251         PR ipa/81360
26252         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
26253         * symtab.c: Include builtins.h
26254         (symtab_node::output_to_lto_symbol_table_p): Move here
26255         from lto-streamer-out.c:output_symbol_p.
26256         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
26257         (output_symbol_p): Move all logic to symtab.c
26258         (produce_symtab): Update.
26260 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26262         * config/s390/s390-opts.h (enum indirect_branch): Define.
26263         * config/s390/s390-protos.h (s390_return_addr_from_memory)
26264         (s390_indirect_branch_via_thunk)
26265         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
26266         (enum s390_indirect_branch_type): Define.
26267         * config/s390/s390.c (struct s390_frame_layout, struct
26268         machine_function): Remove.
26269         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
26270         (indirect_branch_table_label_no, indirect_branch_table_name):
26271         Define variables.
26272         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
26273         (enum s390_indirect_branch_option): Define.
26274         (s390_return_addr_from_memory): New function.
26275         (s390_handle_string_attribute): New function.
26276         (s390_attribute_table): Add new attribute handler.
26277         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
26278         (s390_indirect_branch_via_thunk): New function.
26279         (s390_indirect_branch_via_inline_thunk): New function.
26280         (s390_function_ok_for_sibcall): When jumping via thunk disallow
26281         sibling call optimization for non z10 compiles.
26282         (s390_emit_call): Force indirect branch target to be a single
26283         register.  Add r1 clobber for non-z10 compiles.
26284         (s390_emit_epilogue): Emit return jump via return_use expander.
26285         (s390_reorg): Handle JUMP_INSNs as execute targets.
26286         (s390_option_override_internal): Perform validity checks for the
26287         new command line options.
26288         (s390_indirect_branch_attrvalue): New function.
26289         (s390_indirect_branch_settings): New function.
26290         (s390_set_current_function): Invoke s390_indirect_branch_settings.
26291         (s390_output_indirect_thunk_function):  New function.
26292         (s390_code_end): Implement target hook.
26293         (s390_case_values_threshold): Implement target hook.
26294         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
26295         macros.
26296         * config/s390/s390.h (struct s390_frame_layout)
26297         (struct machine_function): Move here from s390.c.
26298         (TARGET_INDIRECT_BRANCH_NOBP_RET)
26299         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
26300         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
26301         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
26302         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
26303         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
26304         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
26305         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
26306         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
26307         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
26308         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
26309         (mnemonic attribute): Add values which aren't recognized
26310         automatically.
26311         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
26312         pattern for branch conversion.  Fix mnemonic attribute.
26313         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
26314         indirect branch via thunk if requested.
26315         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
26316         ("*indirect_jump"): Disable for branch conversion using out of
26317         line thunks.
26318         ("indirect_jump_via_thunk<mode>_z10")
26319         ("indirect_jump_via_thunk<mode>")
26320         ("indirect_jump_via_inlinethunk<mode>_z10")
26321         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
26322         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
26323         ("casesi_jump_via_inlinethunk<mode>_z10")
26324         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
26325         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
26326         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
26327         ("*indirect2_jump"): Disable for branch conversion.
26328         ("casesi_jump"): Turn into expander and expand patterns for branch
26329         conversion.
26330         ("return_use"): New expander.
26331         ("*return"): Emit return via thunk and rename it to ...
26332         ("*return<mode>"): ... this one.
26333         * config/s390/s390.opt: Add new options and and enum for the
26334         option values.
26336 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26338         * lra-constraints.c (match_reload): Unconditionally use
26339         gen_lowpart_SUBREG, rather than selecting between that
26340         and equivalent gen_rtx_SUBREG code.
26342 2018-02-08  Richard Biener  <rguenther@suse.de>
26344         PR tree-optimization/84233
26345         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
26346         changed flag instead of boguously re-using phi_inserted.
26348 2018-02-08  Martin Jambor  <mjambor@suse.cz>
26350         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
26351         static local variables.
26353 2018-02-08  Richard Biener  <rguenther@suse.de>
26355         PR tree-optimization/84278
26356         * tree-vect-stmts.c (vectorizable_store): When looking for
26357         smaller vector types to perform grouped strided loads/stores
26358         make sure the mode is supported by the target.
26359         (vectorizable_load): Likewise.
26361 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26363         * config/aarch64/aarch64.c (aarch64_components_for_bb):
26364         Increase LDP/STP opportunities by adding adjacent callee-saves.
26366 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26368         PR rtl-optimization/84068
26369         PR rtl-optimization/83459
26370         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
26372 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
26374         PR tree-optimization/84224
26375         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
26376         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
26377         non-zero arguments.
26379 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
26381         PR target/84113
26382         * config/rs6000/altivec.md (*restore_world): Remove LR use.
26383         * config/rs6000/predicates.md (restore_world_operation): Adjust op
26384         count, remove one USE.
26386 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
26388         * doc/install.texi (Configuration): Document the
26389         --with-long-double-format={ibm,ieee} PowerPC configuration
26390         options.
26392         PR target/84154
26393         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
26394         Convert from define_expand to be define_insn_and_split.  Rework
26395         float/double/_Float128 conversions to QI/HI/SImode to work with
26396         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
26397         conversions to QI/HImode types did a store and then a load to
26398         truncate the value.  For conversions to VSX registers, don't split
26399         the insn, instead emit the code directly.  Use the code iterator
26400         any_fix to combine signed and unsigned conversions.
26401         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
26402         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
26403         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26404         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
26405         (fix_<mode>di2_hw): Likewise.
26406         (fixuns_<mode>di2_hw): Likewise.
26407         (fix_<mode>si2_hw): Likewise.
26408         (fixuns_<mode>si2_hw): Likewise.
26409         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
26410         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26411         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
26412         fix<uns>_trunc<SFDF:mode>si2_p8.
26413         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
26414         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
26415         (fix<uns>_<mode>_mem): Likewise.
26416         (fctiw<u>z_<mode>_mem): Likewise.
26417         (fix<uns>_<mode>_mem): Likewise.
26418         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
26419         the register allocator from doing a direct move to the GPRs to do
26420         a store, and instead use the ISA 3.0 store byte/half-word from
26421         vector register instruction.  For IEEE 128-bit floating point,
26422         also optimize stores of 32-bit ints.
26423         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
26425 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
26427         * genextract.c (push_pathstr_operand): New function to support
26428         [a-zA-Z].
26429         (walk_rtx): Call push_pathstr_operand.
26430         (print_path): Support [a-zA-Z].
26432 2018-02-07  Richard Biener  <rguenther@suse.de>
26434         PR tree-optimization/84037
26435         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
26436         (cse_and_gimplify_to_preheader): Declare.
26437         (vect_get_place_in_interleaving_chain): Likewise.
26438         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26439         ivexpr_map.
26440         (_loop_vec_info::~_loop_vec_info): Delete it.
26441         (cse_and_gimplify_to_preheader): New function.
26442         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
26443         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
26444         (vectorizable_load): Likewise.  For grouped stores always base
26445         the IV on the first element.
26446         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
26447         condition before gimplifying.
26449 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
26451         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
26452         *DIV_EXPR and *MOD_EXPR.
26454 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
26456         PR target/84248
26457         * config/i386/i386.c (ix86_option_override_internal): Mask out
26458         the CF_SET bit when checking -fcf-protection.
26460 2018-02-07  Tom de Vries  <tom@codesourcery.com>
26462         PR libgomp/84217
26463         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
26464         enough.
26466 2018-02-07  Richard Biener  <rguenther@suse.de>
26468         PR tree-optimization/84204
26469         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
26470         this place.
26472         PR tree-optimization/84205
26473         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
26474         special-case isl_ast_op_zdiv_r.
26476         PR tree-optimization/84223
26477         * graphite-scop-detection.c (gather_bbs::before_dom_children):
26478         Only add conditions from within the region.
26479         (gather_bbs::after_dom_children): Adjust.
26481 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
26483         PR target/84209
26484         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
26485         * config/avr/avr.md: Only post-reload split REG-REG moves if
26486         either register is GENERAL_REG_P.
26488 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
26490         PR tree-optimization/84235
26491         * tree-ssa-scopedtables.c
26492         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
26493         if the subtraction is performed in floating point type where NaNs are
26494         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
26495         build 1.  Formatting fix.
26497 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
26499         PR target/84146
26500         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
26501         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
26502         and skip it regardless of bb boundaries.  Use CALL_P macro,
26503         don't test INSN_P (insn) together with CALL_P or JUMP_P check
26504         unnecessarily, formatting fix.
26506 2018-02-06  Michael Collison  <michael.collison@arm.com>
26508         * config/arm/thumb2.md:
26509         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
26510         (*thumb_mov_notscc): Ditto.
26512 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
26514         PR target/84154
26515         * config/rs6000/rs6000.md (su code attribute): Use "u" for
26516         unsigned_fix, not "s".
26518 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26520         * configure.ac (gcc_fn_eh_frame_ro): New function.
26521         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
26522         correct .eh_frame permissions.
26523         * configure: Regenerate.
26525 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
26527         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
26528         irrelevant options.
26530 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26532         * config/rs6000/rs6000.c (rs6000_option_override_internal):
26533         Display warning message for -mno-speculate-indirect-jumps.
26535 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
26537         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
26538         Undocumented.
26539         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
26541 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
26543         PR tree-optimization/84225
26544         * tree-eh.c (find_trapping_overflow): Only call
26545         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
26547 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26549         PR target/84145
26550         * config/i386/i386.c: Reimplement the check of possible options
26551         -mibt/-mshstk conbination. Change error messages.
26552         * doc/invoke.texi: Fix a typo: remove extra '='.
26554 2018-02-06  Marek Polacek  <polacek@redhat.com>
26556         PR tree-optimization/84228
26557         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
26559 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
26561         PR target/82641
26562         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
26563         emitted arch directives.
26564         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
26565         __ARM_FEATURE_COPROC before changing architectures.
26567 2018-02-06  Richard Biener  <rguenther@suse.de>
26569         * config/i386/i386.c (print_reg): Fix typo.
26570         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
26572 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
26574         * configure: Regenerate.
26576 2018-02-05  Martin Sebor  <msebor@redhat.com>
26578         PR tree-optimization/83369
26579         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
26580         inlining context.
26582 2018-02-05  Martin Liska  <mliska@suse.cz>
26584         * doc/invoke.texi: Cherry-pick upstream r323995.
26586 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
26588         * ira.c (ira_init_register_move_cost): Adjust comment.
26590 2018-02-05  Martin Liska  <mliska@suse.cz>
26592         PR gcov-profile/84137
26593         * doc/gcov.texi: Fix typo in documentation.
26595 2018-02-05  Martin Liska  <mliska@suse.cz>
26597         PR gcov-profile/83879
26598         * doc/gcov.texi: Document necessity of --dynamic-list-data when
26599         using dlopen functionality.
26601 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
26603         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
26604         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
26605         _mm_maskz_range_ss, _mm_mask_range_round_ss,
26606         _mm_maskz_range_round_ss): New intrinsics.
26607         (__builtin_ia32_rangesd128_round)
26608         (__builtin_ia32_rangess128_round): Remove.
26609         (__builtin_ia32_rangesd128_mask_round,
26610         __builtin_ia32_rangess128_mask_round): New builtins.
26611         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
26612         __builtin_ia32_rangess128_round): Remove.
26613         (__builtin_ia32_rangesd128_mask_round,
26614         __builtin_ia32_rangess128_mask_round): New builtins.
26615         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
26616         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
26617         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
26618         "<round_saeonly_constraint>")): Changed to ...
26619         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
26620         "<round_saeonly_scalar_constraint>")): ... this.
26621         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
26622         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
26623         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
26624         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
26625         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
26627 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
26629         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
26630         options.
26631         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
26632         Remove all values except native, 8540 and 8548.
26634 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
26636         * config/i386/i386.c (ix86_output_function_return): Pass
26637         INVALID_REGNUM, instead of -1, as invalid register number to
26638         indirect_thunk_name and output_indirect_thunk.
26640 2018-02-02  Julia Koval  <julia.koval@intel.com>
26642         * config.gcc: Add -march=icelake.
26643         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
26644         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
26645         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
26646         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
26647         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
26648         (processor_target_table): Add icelake.
26649         (ix86_option_override_internal): Handle new PTAs.
26650         (get_builtin_code_for_version): Handle icelake.
26651         (M_INTEL_COREI7_ICELAKE): New.
26652         (fold_builtin_cpu): Handle icelake.
26653         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
26654         * doc/invoke.texi: Add -march=icelake.
26656 2018-02-02  Julia Koval  <julia.koval@intel.com>
26658         * config/i386/i386.c (ix86_option_override_internal): Change flags type
26659         to wide_int_bitmask.
26660         * wide-int-bitmask.h: New.
26662 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26664         PR target/84066
26665         * config/i386/i386.md: Replace Pmode with word_mode in
26666         builtin_setjmp_setup and builtin_longjmp to support x32.
26668 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
26670         PR target/56010
26671         PR target/83743
26672         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
26673         #include "opts.h".
26674         (rs6000_supported_cpu_names): New static variable.
26675         (linux_cpu_translation_table): Likewise.
26676         (elf_platform) <cpu>: Define new static variable and use it.
26677         Translate kernel AT_PLATFORM name to canonical name if needed.
26678         Error if platform name is unknown.
26680 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
26682         PR target/84089
26683         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
26685 2018-02-01  Jeff Law  <law@redhat.com>
26687         PR target/84128
26688         * config/i386/i386.c (release_scratch_register_on_entry): Add new
26689         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
26690         the scratch if RELEASE_VIA_POP is false.
26691         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
26692         If we have to save a temporary register, decrement SIZE appropriately.
26693         Pass new arguments to release_scratch_register_on_entry.
26694         (ix86_adjust_stack_and_probe): Likewise.
26695         (ix86_emit_probe_stack_range): Pass new arguments to
26696         release_scratch_register_on_entry.
26698 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
26700         PR rtl-optimization/84157
26701         * combine.c (change_zero_ext): Use REG_P predicate in
26702         front of HARD_REGISTER_P predicate.
26704 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
26706         * config/avr/avr.c (avr_option_override): Move disabling of
26707         -fdelete-null-pointer-checks to...
26708         * common/config/avr/avr-common.c (avr_option_optimization_table):
26709         ...here.
26711 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26713         PR tree-optimization/81635
26714         * tree-data-ref.c (split_constant_offset_1): For types that
26715         wrap on overflow, try to use range info to prove that wrapping
26716         cannot occur.
26718 2018-02-01  Renlin Li  <renlin.li@arm.com>
26720         PR target/83370
26721         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
26722         TAILCALL_ADDR_REGS.
26723         (aarch64_register_move_cost): Likewise.
26724         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
26725         TAILCALL_ADDR_REGS.
26726         (REG_CLASS_NAMES): Likewise.
26727         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
26728         TAILCALL_ADDR_REGS. Remove IP registers.
26729         * config/aarch64/aarch64.md (Ucs): Update register constraint.
26731 2018-02-01  Richard Biener  <rguenther@suse.de>
26733         * domwalk.h (dom_walker::dom_walker): Add additional constructor
26734         for specifying RPO order and allow NULL for that.
26735         * domwalk.c (dom_walker::dom_walker): Likewise.
26736         (dom_walker::walk): Handle NULL RPO order.
26737         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
26738         in RPO order.
26739         (rewrite_update_dom_walker): Likewise.
26740         (mark_def_dom_walker): Likewise.
26742 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26744         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
26745         (aarch64_maybe_expand_sve_subreg_move): Declare.
26746         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
26747         * config/aarch64/predicates.md (aarch64_any_register_operand): New
26748         predicate.
26749         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
26750         that are semantically a reverse operation.
26751         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
26752         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
26753         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
26754         functions.
26755         (aarch64_can_change_mode_class): For big-endian, forbid changes
26756         between two SVE modes if they have different element sizes.
26758 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26760         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
26761         the TImode handling for big-endian targets.
26763 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26765         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
26766         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
26767         not just bytes.
26768         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
26769         Remove BSWAP handing for big-endian targets and use the form of
26770         LD1RQ appropariate for the mode.
26772 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26774         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
26775         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
26776         duplicated element.
26778 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26780         PR tearget/83845
26781         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
26782         check for operands that need to go through aarch64_sve_reload_be.
26784 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
26786         PR tree-optimization/81661
26787         PR tree-optimization/84117
26788         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
26789         * tree-eh.c: Include gimplify.h.
26790         (find_trapping_overflow, replace_trapping_overflow,
26791         rewrite_to_non_trapping_overflow): New functions.
26792         * tree-vect-loop.c: Include tree-eh.h.
26793         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
26794         * tree-data-ref.c: Include tree-eh.h.
26795         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
26797 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
26799         PR rtl-optimization/84123
26800         * combine.c (change_zero_ext): Check if hard register satisfies
26801         can_change_dest_mode before calling gen_lowpart_SUBREG.
26803 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
26805         PR target/82444
26806         * ira.c (ira_init_register_move_cost): Remove assert.
26808 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26810         PR rtl-optimization/84071
26811         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
26812         * doc/tm.texi: Regenerate.
26814 2018-01-31  Richard Biener  <rguenther@suse.de>
26816         PR tree-optimization/84132
26817         * tree-data-ref.c (analyze_miv_subscript): Properly
26818         check whether evolution_function_is_affine_multivariate_p
26819         before calling gcd_of_steps_may_divide_p.
26821 2018-01-31  Julia Koval  <julia.koval@intel.com>
26823         PR target/83618
26824         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
26825         * config/i386/i386.md (rdpid_rex64) New.
26826         (rdpid): Make 32bit only.
26828 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
26830         PR lto/84105
26831         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
26832         an IDENTIFIER_NODE for FUNCTION_TYPE's.
26834 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26836         Revert
26837         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26839         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26841 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26843         PR rtl-optimization/84071
26844         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
26845         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
26847 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26849         * config/arc/arc.c (arc_handle_aux_attribute): New function.
26850         (arc_attribute_table): Add 'aux' attribute.
26851         (arc_in_small_data_p): Consider aux like variables.
26852         (arc_is_aux_reg_p): New function.
26853         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
26854         (arc_get_aux_arg): New function.
26855         (prepare_move_operands): Handle aux-register access.
26856         (arc_handle_aux_attribute): New function.
26857         * doc/extend.texi (ARC Variable attributes): Add subsection.
26859 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26861         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
26862         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
26863         (arc_attribute_table): Add 'uncached' attribute.
26864         (arc_print_operand): Print '.di' flag for uncached memory
26865         accesses.
26866         (arc_in_small_data_p): Do not consider for small data the uncached
26867         types.
26868         (arc_is_uncached_mem_p): New function.
26869         * config/arc/predicates.md (compact_store_memory_operand): Check
26870         for uncached memory accesses.
26871         (nonvol_nonimm_operand): Likewise.
26872         * doc/extend.texi (ARC Type Attribute): New subsection.
26874 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
26876         PR c/84100
26877         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
26878         falign-loops=): Add Optimization flag.
26880 2018-01-30  Jeff Law  <law@redhat.com>
26882         PR target/84064
26883         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
26884         INT_REGISTERS_SAVED.  Check it prior to calling
26885         get_scratch_register_on_entry.
26886         (ix86_adjust_stack_and_probe): Similarly.
26887         (ix86_emit_probe_stack_range): Similarly.
26888         (ix86_expand_prologue): Corresponding changes.
26890 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26892         PR target/40411
26893         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
26894         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
26896 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
26898         PR target/84112
26899         * lra-constraints.c (curr_insn_transform): Process AND in the
26900         address.
26902 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
26904         PR rtl-optimization/83986
26905         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
26906         dependence against last_pending_memory_flush in addition to
26907         pending_jump_insns.
26909 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
26911         PR tree-optimization/81611
26912         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
26913         copies.
26915 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26917         PR target/83758
26918         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
26919         a reg rtx.
26921 2018-01-30  Richard Biener  <rguenther@suse.de>
26922             Jakub Jelinek  <jakub@redhat.com>
26924         PR tree-optimization/84111
26925         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
26926         inner loops added during recursion, as they don't have up-to-date
26927         SSA form.
26929 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26931         PR ipa/81360
26932         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
26933         (can_inline_edge_by_limits_p): ... here.
26934         (can_early_inline_edge_p, check_callers,
26935         update_caller_keys, update_callee_keys, recursive_inlining,
26936         add_new_edges_to_heap, speculation_useful_p,
26937         inline_small_functions,
26938         inline_small_functions, flatten_function,
26939         inline_to_all_callers_1): Update.
26941 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26943         * profile-count.c (profile_count::combine_with_ipa_count): Handle
26944         zeros correctly.
26946 2018-01-30  Richard Biener  <rguenther@suse.de>
26948         PR tree-optimization/83008
26949         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
26950         invariant and constant vector uses in stmts when they need
26951         more than one stmt.
26953 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26955         PR bootstrap/84017
26956         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
26957         * configure: Regenerate.
26959 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26961         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
26962         pattern.
26963         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
26964         Use gen_rtx_REG rather than gen_lowpart.
26966 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26968         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
26969         rather than 0 when creating partial subregs.
26971 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26973         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
26974         of usage.
26976 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
26978         PR target/81550
26979         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
26980         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
26981         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
26982         flags.  This restores the settings used before the 2017-07-24.
26983         Turning off pre increment/decrement/modify allows IVOPTS to
26984         optimize DF/SF loops where the index is an int.
26986 2018-01-29  Richard Biener  <rguenther@suse.de>
26987             Kelvin Nilsen  <kelvin@gcc.gnu.org>
26989         PR bootstrap/80867
26990         * tree-vect-stmts.c (vectorizable_call): Don't call
26991         targetm.vectorize_builtin_md_vectorized_function if callee is
26992         NULL.
26994 2018-01-22  Carl Love  <cel@us.ibm.com>
26996         * doc/extend.tex: Fix typo in second arg in
26997         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
26999 2018-01-29  Richard Biener  <rguenther@suse.de>
27001         PR tree-optimization/84086
27002         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
27003         (flush_ssaname_freelist): When SSA names were released reset
27004         the SCEV hash table.
27006 2018-01-29  Richard Biener  <rguenther@suse.de>
27008         PR tree-optimization/84057
27009         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
27010         removed paths when removing edges.
27012 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
27014         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
27015         -mfunction-return=@var{choice}.
27017 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27019         PR diagnostic/84034
27020         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
27021         Handle CR like TAB.
27022         (layout::print_source_line): Likewise.
27023         (test_get_line_width_without_trailing_whitespace): Add test cases.
27025 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
27027         PR middle-end/84040
27028         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
27029         debug insns.
27031 2018-01-26  Jim Wilson  <jimw@sifive.com>
27033         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
27035         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
27036         specified.
27038 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27040         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
27041         and CMP + SUB-immediate -> SUBS.
27043 2018-01-26  Martin Sebor  <msebor@redhat.com>
27045         PR tree-optimization/83896
27046         * tree-ssa-strlen.c (get_string_len): Rename...
27047         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
27048         Avoid assuming length is constant.
27049         (handle_char_store): Use HOST_WIDE_INT for string length.
27051 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
27053         PR target/81763
27054         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
27055         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
27057 2018-01-26  Richard Biener  <rguenther@suse.de>
27059         PR rtl-optimization/84003
27060         * dse.c (record_store): Only record redundant stores when
27061         the earlier store aliases at least all accesses the later one does.
27063 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
27065         PR rtl-optimization/83985
27066         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
27067         REG_CFA_RESTORE insns.
27068         (delete_unmarked_insns): Don't ignore separate shrink wrapping
27069         REG_CFA_RESTORE insns here.
27071         PR c/83989
27072         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
27073         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
27075 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27077         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
27078         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
27079         (arc_init): Likewise.
27080         (arc_override_options): Likewise.
27081         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
27082         value.
27083         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
27084         support.
27085         * config/arc/arc.h (TARGET_DBNZ): Define.
27086         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
27087         properly set the tune attribute.
27088         (dbnz): Use TARGET_DBNZ guard.
27089         * config/arc/arc.opt (mtune): Add core3 option.
27091 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27093         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
27094         recognize new pic like addresses.
27095         (arc_delegitimize_address): Clean up.
27097 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27099         * config/arc/arc-arches.def: Option mrf16 valid for all
27100         architectures.
27101         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
27102         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
27103         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
27104         * config/arc/arc-tables.opt: Regenerate.
27105         * config/arc/arc.c (arc_conditional_register_usage): Handle
27106         reduced register file case.
27107         (arc_file_start): Set must have build attributes.
27108         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
27109         mrf16 option value.
27110         * config/arc/arc.opt (mrf16): Add new option.
27111         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
27112         * config/arc/genmultilib.awk: Handle new mrf16 option.
27113         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
27114         * config/arc/t-multilib: Regenerate.
27115         * doc/invoke.texi (ARC Options): Document mrf16 option.
27117 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27119         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
27120         * config/arc/arc.c (arc_handle_secure_attribute): New function.
27121         (arc_attribute_table): Add 'secure_call' attribute.
27122         (arc_print_operand): Print secure call operand.
27123         (arc_function_ok_for_sibcall): Don't optimize tail calls when
27124         secure.
27125         (arc_is_secure_call_p): New function.  * config/arc/arc.md
27126         (call_i): Add support for sjli instruction.
27127         (call_value_i): Likewise.
27128         * config/arc/constraints.md (Csc): New constraint.
27130 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27131             John Eric Martin  <John.Martin@emmicro-us.com>
27133         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
27134         * config/arc/arc.c (_arc_jli_section): New struct.
27135         (arc_jli_section): New type.
27136         (rc_jli_sections): New static variable.
27137         (arc_handle_jli_attribute): New function.
27138         (arc_attribute_table): Add jli_always and jli_fixed attribute.
27139         (arc_file_end): New function.
27140         (TARGET_ASM_FILE_END): Define.
27141         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
27142         (arc_add_jli_section): New function.
27143         (jli_call_scan): Likewise.
27144         (arc_reorg): Call jli_call_scan.
27145         (arc_output_addsi): Remove 'S' from printing asm operand.
27146         (arc_is_jli_call_p): New function.
27147         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
27148         operand.
27149         (movhi_insn): Likewise.
27150         (movsi_insn): Likewise.
27151         (movsi_set_cc_insn): Likewise.
27152         (loadqi_update): Likewise.
27153         (load_zeroextendqisi_update): Likewise.
27154         (load_signextendqisi_update): Likewise.
27155         (loadhi_update): Likewise.
27156         (load_zeroextendhisi_update): Likewise.
27157         (load_signextendhisi_update): Likewise.
27158         (loadsi_update): Likewise.
27159         (loadsf_update): Likewise.
27160         (movsicc_insn): Likewise.
27161         (bset_insn): Likewise.
27162         (bxor_insn): Likewise.
27163         (bclr_insn): Likewise.
27164         (bmsk_insn): Likewise.
27165         (bicsi3_insn): Likewise.
27166         (cmpsi_cc_c_insn): Likewise.
27167         (movsi_ne): Likewise.
27168         (movsi_cond_exec): Likewise.
27169         (clrsbsi2): Likewise.
27170         (norm_f): Likewise.
27171         (normw): Likewise.
27172         (swap): Likewise.
27173         (divaw): Likewise.
27174         (flag): Likewise.
27175         (sr): Likewise.
27176         (kflag): Likewise.
27177         (ffs): Likewise.
27178         (ffs_f): Likewise.
27179         (fls): Likewise.
27180         (call_i): Remove 'S' asm letter, add jli instruction.
27181         (call_value_i): Likewise.
27182         * config/arc/arc.op (mjli-always): New option.
27183         * config/arc/constraints.md (Cji): New constraint.
27184         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
27185         operand.
27186         (subsf3_fpx): Likewise.
27187         (mulsf3_fpx): Likewise.
27188         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
27189         asm operand.
27190         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
27191         function attrbutes.
27192         * doc/invoke.texi (ARC): Document mjli-always option.
27194 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
27196         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
27197         avoid addition with 0 and use incw and decw where possible.
27199 2018-01-26  Richard Biener  <rguenther@suse.de>
27201         PR tree-optimization/81082
27202         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
27203         association if it requires casting to unsigned.
27204         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
27205         from fold_plusminus_mult_expr to catch important cases late when
27206         range info is available.
27208 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27210         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
27211         * configure.ac (hidden_linkonce): New test.
27212         * configure: Regenerate.
27213         * config.in: Regenerate.
27215 2018-01-26  Julia Koval  <julia.koval@intel.com>
27217         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
27218         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
27219         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
27220         _mm_mask_bitshuffle_epi64_mask): Fix type.
27221         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
27222         USI_FTYPE_V4DI_V4DI_USI): Remove.
27223         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
27224         __builtin_ia32_vpshufbitqmb256_mask,
27225         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
27226         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
27227         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
27229 2018-01-26  Alan Modra  <amodra@gmail.com>
27231         PR target/84033
27232         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
27233         UNSPEC_VBPERMQ.  Sort other unspecs.
27235 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
27237         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
27239 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
27241         PR middle-end/83055
27242         * predict.c (drop_profile): Do not push/pop cfun; update also
27243         node->count.
27244         (handle_missing_profiles): Fix logic looking for zero profiles.
27246 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
27248         PR middle-end/83977
27249         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
27250         on functions with #pragma omp declare simd or functions with simd
27251         attribute.
27252         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
27253         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
27254         Remove trailing \n from warning_at calls.
27256 2018-01-25  Tom de Vries  <tom@codesourcery.com>
27258         PR target/84028
27259         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27260         for neutered workers.
27262 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
27264         PR target/68467
27265         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
27266         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
27268 2018-01-24  Jeff Law  <law@redhat.com>
27270         PR target/83994
27271         * i386.c (get_probe_interval): Move to earlier point.
27272         (ix86_compute_frame_layout): If -fstack-clash-protection and
27273         the frame is larger than the probe interval, then use pushes
27274         to save registers rather than reg->mem moves.
27275         (ix86_expand_prologue): Remove conditional for int_registers_saved
27276         assertion.
27278 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
27280         PR target/84014
27281         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
27282         min/max for never referenced object.
27284 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
27286         PR middle-end/83977
27287         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
27288         here.
27289         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
27290         attributes from DECL_ATTRIBUTES (decl) without affecting
27291         DECL_ATTRIBUTES (current_function_decl).
27292         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
27293         functions with non-NULL DECL_ABSTRACT_ORIGIN.
27295 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
27297         PR tree-optimization/83979
27298         * fold-const.c (fold_comparison): Use constant_boolean_node
27299         instead of boolean_{true,false}_node.
27301 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
27303         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
27304         with zero counts.
27306 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27308         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27309         Simplify the clause that sets the length attribute.
27310         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27311         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
27312         clause that sets the length attribute.
27313         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27315 2018-01-24  Tom de Vries  <tom@codesourcery.com>
27317         PR target/83589
27318         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
27319         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
27320         Add strict parameter.
27321         (prevent_branch_around_nothing): Insert dummy insn between branch to
27322         label and label with no ptx insn inbetween.
27323         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
27325 2018-01-24  Tom de Vries  <tom@codesourcery.com>
27327         PR target/81352
27328         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27329         for neutered threads in warp.
27330         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
27332 2018-01-24  Richard Biener  <rguenther@suse.de>
27334         PR tree-optimization/83176
27335         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
27336         operands.
27338 2018-01-24  Richard Biener  <rguenther@suse.de>
27340         PR tree-optimization/82819
27341         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
27342         code generating pluses that are no-ops in the target precision.
27344 2018-01-24  Richard Biener  <rguenther@suse.de>
27346         PR middle-end/84000
27347         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
27349 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27351         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
27352         to merge probabilities.
27353         * predict.c (probably_never_executed): Also mark as cold functions
27354         with global 0 profile and guessed local profile.
27355         * profile-count.c (profile_probability::combine_with_count): New
27356         member function.
27357         * profile-count.h (profile_probability::operator*,
27358         profile_probability::operator*=, profile_probability::operator/,
27359         profile_probability::operator/=): Reduce precision to adjusted
27360         and set value to guessed on contradictory divisions.
27361         (profile_probability::combine_with_freq): Remove.
27362         (profile_probability::combine_wiht_count): Declare.
27363         (profile_count::force_nonzero):: Set to adjusted.
27364         (profile_count::probability_in):: Set quality to adjusted.
27365         * tree-ssa-tail-merge.c (replace_block_by): Use
27366         combine_with_count.
27368 2018-01-23  Andrew Waterman  <andrew@sifive.com>
27369             Jim Wilson  <jimw@sifive.com>
27371         * config/riscv/riscv.c (riscv_stack_boundary): New.
27372         (riscv_option_override): Set riscv_stack_boundary.  Handle
27373         riscv_preferred_stack_boundary_arg.
27374         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
27375         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
27376         (STACK_BOUNDARY): Set to riscv_stack_boundary.
27377         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
27378         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
27379         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
27381 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
27383         PR target/83905
27384         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
27385         of struct ix86_frame.
27386         (ix86_expand_epilogue): Likewise.  Add a local variable for
27387         the reg_save_offset field in struct ix86_frame.
27389 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
27391         PR tree-optimization/82604
27392         * tree-loop-distribution.c (enum partition_kind): New enum item
27393         PKIND_PARTIAL_MEMSET.
27394         (partition_builtin_p): Support above new enum item.
27395         (generate_code_for_partition): Ditto.
27396         (compute_access_range): Differentiate cases that equality can be
27397         proven at all loops, the innermost loops or no loops.
27398         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
27399         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
27400         (finalize_partitions, distribute_loop): Don't fuse partition of
27401         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
27402         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
27403         parloop is enabled.
27405 2018-01-23  Martin Liska  <mliska@suse.cz>
27407         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
27408         order to ignore the predictor.
27409         (PRED_POLYMORPHIC_CALL): Likewise.
27410         (PRED_RECURSIVE_CALL): Likewise.
27412 2018-01-23  Martin Liska  <mliska@suse.cz>
27414         * tree-profile.c (tree_profiling): Print function header to
27415         aware reader which function we are working on.
27416         * value-prof.c (gimple_find_values_to_profile): Do not print
27417         not interesting value histograms.
27419 2018-01-23  Martin Liska  <mliska@suse.cz>
27421         * profile-count.h (enum profile_quality): Add
27422         profile_uninitialized as the first value. Do not number values
27423         as they are zero based.
27424         (profile_count::verify): Update sanity check.
27425         (profile_probability::verify): Likewise.
27427 2018-01-23  Nathan Sidwell  <nathan@acm.org>
27429         * doc/invoke.texi (ffor-scope): Deprecate.
27431 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27433         PR tree-optimization/83510
27434         * domwalk.c (set_all_edges_as_executable): New function.
27435         (dom_walker::dom_walker): Convert bool param
27436         "skip_unreachable_blocks" to enum reachability.  Move setup of
27437         edge flags to set_all_edges_as_executable and only do it when
27438         reachability is REACHABLE_BLOCKS.
27439         * domwalk.h (enum dom_walker::reachability): New enum.
27440         (dom_walker::dom_walker): Convert bool param
27441         "skip_unreachable_blocks" to enum reachability.
27442         (set_all_edges_as_executable): New decl.
27443         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
27444         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
27445         "reachability".
27446         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
27447         but converting true to REACHABLE_BLOCKS.
27448         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
27449         * tree-vrp.c
27450         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
27451         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
27452         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
27453         REACHABLE_BLOCKS.
27454         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
27455         if check_all_array_refs will be called.
27457 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27459         * tree.c (selftest::test_location_wrappers): Add more test
27460         coverage.
27462 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27464         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
27465         (selftest::test_bit_in_range): Likewise.
27467 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27469         PR testsuite/83888
27470         * doc/sourcebuild.texi (vect_float): Say that the selector
27471         only describes the situation when -funsafe-math-optimizations is on.
27472         (vect_float_strict): Document.
27474 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27476         PR tree-optimization/83965
27477         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
27478         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
27479         instead of checking only for a reduction.
27480         (vect_recog_widen_sum_pattern): Likewise.
27482 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27484         * predict.c (probably_never_executed): Only use precise profile info.
27485         (compute_function_frequency): Skip after inlining hack since we now
27486         have quality checking.
27488 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27490         * profile-count.h (profile_probability::very_unlikely,
27491         profile_probability::unlikely, profile_probability::even): Set
27492         precision to guessed.
27494 2018-01-23  Richard Biener  <rguenther@suse.de>
27496         PR tree-optimization/83963
27497         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
27498         Properly terminate dominator walk when crossing the exit edge not
27499         when visiting its source block.
27501 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
27503         PR c++/83918
27504         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
27505         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
27507 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
27509         PR tree-optimization/83957
27510         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
27511         semicolon after for body surrounded by braces.
27513         PR tree-optimization/83081
27514         * profile-count.h (profile_probability::split): New method.
27515         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
27516         Use profile_probability::split.
27517         (do_compare_rtx_and_jump): Fix adjustment of probabilities
27518         when splitting a single conditional jump into 2.
27520 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
27522         PR tree-optimization/69452
27523         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
27524         decl.
27526 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27528         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
27529         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
27530         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
27532 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27534         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
27535         declaration.
27536         * config/rl78/rl78.md (movdi): New define_expand.
27537         * config/rl78/rl78.c (rl78_split_movdi): New function.
27539 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
27541         PR target/83862
27542         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
27543         no longer used.
27544         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
27545         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
27546         128-bit to produce an UNSPEC move to get the double word with the
27547         signbit and then a shift directly to do signbit.
27548         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
27549         implementation with a new version that just does either a direct
27550         move or a regular move.  Move memory interface to separate insns.
27551         Move insns so they are next to the expander.
27552         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
27553         with signbit move.  Split big and little endian case.
27554         (signbit<mode>2_dm_mem_le): Likewise.
27555         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
27556         (signbit<mode>2_dm2): Likewise.
27558 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27560         * config/rl78/rl78.md (anddi3): New define_expand.
27562 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27564         * config/rl78/rl78.md (umindi3): New define_expand.
27566 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27568         * config/rl78/rl78.md (smindi3): New define_expand.
27570 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27572         * config/rl78/rl78.md (smaxdi3): New define_expand.
27574 2018-01-22  Carl Love  <cel@us.ibm.com>
27576         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
27577         LVX_V1TI): Add macro expansion.
27578         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
27579         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
27580         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
27581         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
27582         Change check to determine if the instruction is a byte reversing
27583         entry.  Fix typo in comment.
27584         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
27585         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
27586         Add def_builtin calls for new builtins.
27587         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
27588         Add define_insn expansion.
27590 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27592         * config/rl78/rl78.md (umaxdi3): New define_expand.
27594 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27596         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
27597         for non-QImode registers.
27599 2018-01-22  Richard Biener  <rguenther@suse.de>
27601         PR tree-optimization/83963
27602         * graphite-scop-detection.c (scop_detection::get_sese): Delay
27603         including the loop exit block.
27604         (scop_detection::merge_sese): Likewise.
27605         (scop_detection::add_scop): Do it here instead.
27607 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27609         * doc/sourcebuild.texi (arm_softfloat): Document.
27611 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
27613         PR gcc/77734
27614         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
27615         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
27616         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
27618 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27619             David Edelsohn  <dje.gcc@gmail.com>
27621         PR target/83946
27622         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27623         Change "crset eq" to "crset 2".
27624         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27625         (*call_indirect_aix<mode>_nospec): Likewise.
27626         (*call_value_indirect_aix<mode>_nospec): Likewise.
27627         (*call_indirect_elfv2<mode>_nospec): Likewise.
27628         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
27629         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
27630         change assembly output from . to $.
27631         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27632         (indirect_jump<mode>_nospec): Change assembly output from . to $.
27633         (*tablejump<mode>_internal1_nospec): Likewise.
27635 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
27637         PR target/80870
27638         * config/sh/sh_optimize_sett_clrt.cc:
27639         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
27641 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
27643         PR tree-optimization/83940
27644         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
27645         offset_dt to vect_constant_def rather than vect_unknown_def_type.
27646         (vect_check_load_store_mask): Add a mask_dt_out parameter and
27647         use it to pass back the definition type.
27648         (vect_check_store_rhs): Likewise rhs_dt_out.
27649         (vect_build_gather_load_calls): Add a mask_dt argument and use
27650         it instead of a call to vect_is_simple_use.
27651         (vectorizable_store): Update calls to vect_check_load_store_mask
27652         and vect_check_store_rhs.  Use the dt returned by the latter instead
27653         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
27654         instead of calls to vect_is_simple_use.  Pass the scalar rather
27655         than the vector operand to vect_is_simple_use when handling
27656         second and subsequent copies of an rhs value.
27657         (vectorizable_load): Update calls to vect_check_load_store_mask
27658         and vect_build_gather_load_calls.  Use the cached mask_dt and
27659         gs_info.offset_dt instead of calls to vect_is_simple_use.
27661 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
27663         PR middle-end/83945
27664         * tree-emutls.c: Include gimplify.h.
27665         (lower_emutls_2): New function.
27666         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
27667         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
27668         it before further processing.
27670         PR target/83930
27671         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
27672         UINTVAL (trueop1) instead of INTVAL (op1).
27674 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
27676         PR debug/81570
27677         PR debug/83728
27678         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
27679         INCOMING_FRAME_SP_OFFSET if not defined.
27680         (scan_trace): Add ENTRY argument.  If true and
27681         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
27682         emit a note to adjust the CFA offset.
27683         (create_cfi_notes): Adjust scan_trace callers.
27684         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
27685         INCOMING_FRAME_SP_OFFSET in the CIE.
27686         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
27687         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
27688         Likewise.
27689         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
27690         * doc/tm.texi: Regenerated.
27692 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27694         PR rtl-optimization/83147
27695         * lra-constraints.c (remove_inheritance_pseudos): Use
27696         lra_substitute_pseudo_within_insn.
27698 2018-01-19  Tom de Vries  <tom@codesourcery.com>
27699             Cesar Philippidis  <cesar@codesourcery.com>
27701         PR target/83920
27702         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
27704 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
27706         PR target/83790
27707         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
27708         spaces for function labels.
27710 2018-01-19  Martin Liska  <mliska@suse.cz>
27712         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
27713         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
27714         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
27715         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
27716         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
27717         (PRED_CONST_RETURN): Change from 69 to 65.
27718         (PRED_NULL_RETURN): Change from 91 to 71.
27719         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
27720         (PRED_LOOP_GUARD): Change from 66 to 73.
27722 2018-01-19  Martin Liska  <mliska@suse.cz>
27724         * predict.c (predict_insn_def): Add new assert.
27725         (struct branch_predictor): Change type to signed integer.
27726         (test_prediction_value_range): Amend test to cover
27727         PROB_UNINITIALIZED.
27728         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
27729         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
27730         (PRED_LOOP_ITERATIONS_MAX): Likewise.
27731         (PRED_LOOP_IV_COMPARE): Likewise.
27732         * predict.h (PROB_UNINITIALIZED): Define new constant.
27734 2018-01-19  Martin Liska  <mliska@suse.cz>
27736         * predict.c (dump_prediction): Add new format for
27737         analyze_brprob.py script which is enabled with -details
27738         suboption.
27739         * profile-count.h (precise_p): New function.
27741 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27743         PR tree-optimization/83922
27744         * tree-vect-loop.c (vect_verify_full_masking): Return false if
27745         there are no statements that need masking.
27746         (vect_active_double_reduction_p): New function.
27747         (vect_analyze_loop_operations): Use it when handling phis that
27748         are not in the loop header.
27750 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27752         PR tree-optimization/83914
27753         * tree-vect-loop.c (vectorizable_induction): Don't convert
27754         init_expr or apply the peeling adjustment for inductions
27755         that are nested within the vectorized loop.
27757 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27759         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
27760         instead of NEG.
27762 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27764         PR sanitizer/81715
27765         PR testsuite/83882
27766         * function.h (gimplify_parameters): Add gimple_seq * argument.
27767         * function.c: Include gimple.h and options.h.
27768         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
27769         for the added local temporaries if needed.
27770         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
27771         if there are any parameter cleanups, wrap whole body into a
27772         try/finally with the cleanups.
27774 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
27776         PR target/82964
27777         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27778         Use GET_MODE_CLASS for scalar floating point.
27780 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27782         PR ipa/82256
27783         patch by PaX Team
27784         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
27785         Fix call of call_cgraph_insertion_hooks.
27787 2018-01-18  Martin Sebor  <msebor@redhat.com>
27789         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
27791 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27793         PR ipa/83619
27794         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
27795         frequencies.
27797 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
27799         PR other/70268
27800         * common.opt: (-ffile-prefix-map): New option.
27801         * opts.c (common_handle_option): Defer it.
27802         * opts-global.c (handle_common_deferred_options): Handle it.
27803         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
27804         * file-prefix-map.h: New file.
27805         (remap_debug_filename, add_debug_prefix_map): ...here.
27806         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
27807         * final.c (debug_prefix_map, add_debug_prefix_map
27808         remap_debug_filename): Move to...
27809         * file-prefix-map.c: New file.
27810         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
27811         generalize, get rid of alloca(), use strrchr() instead of strchr().
27812         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
27813         Implement in terms of add_prefix_map().
27814         (remap_macro_filename, remap_debug_filename): Implement in term of
27815         remap_filename().
27816         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
27817         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
27818         * dbxout.c: Include file-prefix-map.h.
27819         * varasm.c: Likewise.
27820         * vmsdbgout.c: Likewise.
27821         * xcoffout.c: Likewise.
27822         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
27823         * doc/cppopts.texi (-fmacro-prefix-map): Document.
27824         * doc/invoke.texi (-ffile-prefix-map): Document.
27825         (-fdebug-prefix-map): Update description.
27827 2018-01-18  Martin Liska  <mliska@suse.cz>
27829         * config/i386/i386.c (indirect_thunk_name): Document that also
27830         lfence is emitted.
27831         (output_indirect_thunk): Document why both instructions
27832         (pause and lfence) are generated.
27834 2018-01-18  Richard Biener  <rguenther@suse.de>
27836         PR tree-optimization/83887
27837         * graphite-scop-detection.c
27838         (scop_detection::get_nearest_dom_with_single_entry): Remove.
27839         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
27840         (scop_detection::merge_sese): Re-implement with a flood-fill
27841         algorithm that properly finds a SESE region if it exists.
27843 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27845         PR c/61240
27846         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
27847         pointer_diff optimizations use view_convert instead of convert.
27849 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27851         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27852         Generate different code for -mno-speculate-indirect-jumps.
27853         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27854         (*call_indirect_aix<mode>): Disable for
27855         -mno-speculate-indirect-jumps.
27856         (*call_indirect_aix<mode>_nospec): New define_insn.
27857         (*call_value_indirect_aix<mode>): Disable for
27858         -mno-speculate-indirect-jumps.
27859         (*call_value_indirect_aix<mode>_nospec): New define_insn.
27860         (*sibcall_nonlocal_sysv<mode>): Generate different code for
27861         -mno-speculate-indirect-jumps.
27862         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27864 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
27866         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
27867         long double type, set the flags for noting the default long double
27868         type, even if we don't pass or return a long double type.
27870 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
27872         PR ipa/83051
27873         * ipa-inline.c (flatten_function): Do not overwrite final inlining
27874         failure.
27876 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
27878         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
27879         support for merge[hl].
27880         (fold_mergehl_helper): New helper function.
27881         (tree-vector-builder.h): New #include for tree_vector_builder usage.
27882         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
27883         (altivec_vmrglw_direct): Add xxmrglw insn.
27885 2018-01-17  Andrew Waterman  <andrew@sifive.com>
27887         * config/riscv/riscv.c (riscv_conditional_register_usage): If
27888         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
27890 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
27892         PR lto/83121
27893         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
27894         call the lto_location_cache before reading the
27895         DECL_SOURCE_LOCATION of the types.
27897 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
27898             Richard Sandiford  <richard.sandiford@linaro.org>
27900         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
27901         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
27902         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
27903         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
27904         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
27905         Add declaration.
27906         * config/aarch64/constraints.md (aarch64_movti_operand):
27907         Limit immediates.
27908         * config/aarch64/predicates.md (Uti): Add new constraint.
27910 2018-01-17  Carl Love  <cel@us.ibm.com>
27912         * config/rs6000/vsx.md (define_expand xl_len_r,
27913         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
27914         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
27915         lxvll.
27916         (define_expand, define_insn): Move the shift left from  the
27917         define_insn to the define_expand for lxvl and stxvl instructions.
27918         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
27919         and XL_LEN_R definitions to PURE.
27921 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
27923         * config/i386/i386.c (indirect_thunk_name): Declare regno
27924         as unsigned int.  Compare regno with INVALID_REGNUM.
27925         (output_indirect_thunk): Ditto.
27926         (output_indirect_thunk_function): Ditto.
27927         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
27928         in the call to output_indirect_thunk_function.
27930 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
27932         PR middle-end/83884
27933         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
27934         rather than the size of inner_type to determine the stack slot size
27935         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
27937 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
27939         PR target/83546
27940         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
27941         to PTA_SILVERMONT.
27943 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
27945         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
27946         endian Linux systems to optionally enable multilibs for selecting
27947         the long double type if the user configured an explicit type.
27948         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
27949         have no long double multilibs if not defined.
27950         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
27951         warn if the user used -mabi={ieee,ibm}longdouble and we built
27952         multilibs for long double.
27953         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
27954         appropriate multilib option.
27955         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
27956         multilib options.
27957         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
27958         for building long double multilibs.
27959         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
27961 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
27963         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
27964         copies.
27966         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
27967         64 bits.
27968         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
27969         128 bits.
27971         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
27972         variables.
27974         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
27975         return value.
27977 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27979         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
27980         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
27982 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27984         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
27985         different rtl trees depending on TARGET_64BIT.
27986         (rs6000_gen_lvx): Likewise.
27988 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27990         * config/visium/visium.md (nop): Tweak comment.
27991         (hazard_nop): Likewise.
27993 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27995         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
27996         -mspeculate-indirect-jumps.
27997         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
27998         for -mno-speculate-indirect-jumps.
27999         (*call_indirect_elfv2<mode>_nospec): New define_insn.
28000         (*call_value_indirect_elfv2<mode>): Disable for
28001         -mno-speculate-indirect-jumps.
28002         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
28003         (indirect_jump): Emit different RTL for
28004         -mno-speculate-indirect-jumps.
28005         (*indirect_jump<mode>): Disable for
28006         -mno-speculate-indirect-jumps.
28007         (*indirect_jump<mode>_nospec): New define_insn.
28008         (tablejump): Emit different RTL for
28009         -mno-speculate-indirect-jumps.
28010         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
28011         (tablejumpsi_nospec): New define_expand.
28012         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
28013         (tablejumpdi_nospec): New define_expand.
28014         (*tablejump<mode>_internal1): Disable for
28015         -mno-speculate-indirect-jumps.
28016         (*tablejump<mode>_internal1_nospec): New define_insn.
28017         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
28018         option.
28020 2018-01-16  Artyom Skrobov tyomitch@gmail.com
28022         * caller-save.c (insert_save): Drop unnecessary parameter.  All
28023         callers updated.
28025 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
28026             Richard Biener  <rguenth@suse.de>
28028         PR libgomp/83590
28029         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
28030         return early, inline manually is_gimple_sizepos.  Make sure if we
28031         call gimplify_expr we don't end up with a gimple constant.
28032         * tree.c (variably_modified_type_p): Don't return true for
28033         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
28034         * gimplify.h (is_gimple_sizepos): Remove.
28036 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28038         PR tree-optimization/83857
28039         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
28040         vectorizable_live_operation for pure SLP statements.
28041         (vectorizable_live_operation): Handle PHIs.
28043 2018-01-16  Richard Biener  <rguenther@suse.de>
28045         PR tree-optimization/83867
28046         * tree-vect-stmts.c (vect_transform_stmt): Precompute
28047         nested_in_vect_loop_p since the scalar stmt may get invalidated.
28049 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
28051         PR c/83844
28052         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
28053         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
28054         If off is not INTEGER_CST, issue a may not be aligned warning
28055         rather than isn't aligned.  Use isn%'t rather than isn't.
28056         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
28057         into MULT_EXPR.
28058         <case MULT_EXPR>: Improve the case when bottom and one of the
28059         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
28060         operand, in that case check if the other operand is multiple of
28061         bottom divided by the INTEGER_CST operand.
28063 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28065         PR target/83858
28066         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
28067         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
28068         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
28069         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
28070         * config/pa/pa.c (pa_function_arg_advance): Likewise.
28071         (pa_function_arg, pa_arg_partial_bytes): Likewise.
28072         (pa_function_arg_size): New function.
28074 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28076         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
28077         in a separate statement.
28079 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28081         PR tree-optimization/83847
28082         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
28083         group gathers and scatters.
28085 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
28087         PR rtl-optimization/86620
28088         * params.def (max-sched-ready-insns): Bump minimum value to 1.
28090         PR rtl-optimization/83213
28091         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
28092         to last if both are JUMP_INSNs.
28094         PR tree-optimization/83843
28095         * gimple-ssa-store-merging.c
28096         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
28097         store_immediate_info for bswap/nop orig_stores.
28099 2018-01-15  Andrew Waterman  <andrew@sifive.com>
28101         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
28102         !TARGET_MUL.
28103         <UDIV>: Increase cost if !TARGET_DIV.
28105 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
28107         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
28108         (define_attr "cr_logical_3op"): New.
28109         (cceq_ior_compare): Adjust.
28110         (cceq_ior_compare_complement): Adjust.
28111         (*cceq_rev_compare): Adjust.
28112         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
28113         (is_cracked_insn): Adjust.
28114         (insn_must_be_first_in_group): Adjust.
28115         * config/rs6000/40x.md: Adjust.
28116         * config/rs6000/440.md: Adjust.
28117         * config/rs6000/476.md: Adjust.
28118         * config/rs6000/601.md: Adjust.
28119         * config/rs6000/603.md: Adjust.
28120         * config/rs6000/6xx.md: Adjust.
28121         * config/rs6000/7450.md: Adjust.
28122         * config/rs6000/7xx.md: Adjust.
28123         * config/rs6000/8540.md: Adjust.
28124         * config/rs6000/cell.md: Adjust.
28125         * config/rs6000/e300c2c3.md: Adjust.
28126         * config/rs6000/e500mc.md: Adjust.
28127         * config/rs6000/e500mc64.md: Adjust.
28128         * config/rs6000/e5500.md: Adjust.
28129         * config/rs6000/e6500.md: Adjust.
28130         * config/rs6000/mpc.md: Adjust.
28131         * config/rs6000/power4.md: Adjust.
28132         * config/rs6000/power5.md: Adjust.
28133         * config/rs6000/power6.md: Adjust.
28134         * config/rs6000/power7.md: Adjust.
28135         * config/rs6000/power8.md: Adjust.
28136         * config/rs6000/power9.md: Adjust.
28137         * config/rs6000/rs64.md: Adjust.
28138         * config/rs6000/titan.md: Adjust.
28140 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28142         * config/i386/predicates.md (indirect_branch_operand): Rewrite
28143         ix86_indirect_branch_register logic.
28145 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28147         * config/i386/constraints.md (Bs): Update
28148         ix86_indirect_branch_register check.  Don't check
28149         ix86_indirect_branch_register with GOT_memory_operand.
28150         (Bw): Likewise.
28151         * config/i386/predicates.md (GOT_memory_operand): Don't check
28152         ix86_indirect_branch_register here.
28153         (GOT32_symbol_operand): Likewise.
28155 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28157         * config/i386/predicates.md (constant_call_address_operand):
28158         Rewrite ix86_indirect_branch_register logic.
28159         (sibcall_insn_operand): Likewise.
28161 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28163         * config/i386/constraints.md (Bs): Replace
28164         ix86_indirect_branch_thunk_register with
28165         ix86_indirect_branch_register.
28166         (Bw): Likewise.
28167         * config/i386/i386.md (indirect_jump): Likewise.
28168         (tablejump): Likewise.
28169         (*sibcall_memory): Likewise.
28170         (*sibcall_value_memory): Likewise.
28171         Peepholes of indirect call and jump via memory: Likewise.
28172         * config/i386/i386.opt: Likewise.
28173         * config/i386/predicates.md (indirect_branch_operand): Likewise.
28174         (GOT_memory_operand): Likewise.
28175         (call_insn_operand): Likewise.
28176         (sibcall_insn_operand): Likewise.
28177         (GOT32_symbol_operand): Likewise.
28179 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
28181         PR middle-end/83837
28182         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
28183         type rather than type addr's type points to.
28184         (expand_omp_atomic_mutex): Likewise.
28185         (expand_omp_atomic): Likewise.
28187 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28189         PR target/83839
28190         * config/i386/i386.c (output_indirect_thunk_function): Use
28191         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
28192         for  __x86_return_thunk.
28194 2018-01-15  Richard Biener  <rguenther@suse.de>
28196         PR middle-end/83850
28197         * expmed.c (extract_bit_field_1): Fix typo.
28199 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28201         PR target/83687
28202         * config/arm/iterators.md (VF): New mode iterator.
28203         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
28204         Remove integer-related logic from pattern.
28205         (neon_vabd<mode>_3): Likewise.
28207 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
28209         PR middle-end/82694
28210         * common.opt (fstrict-overflow): No longer an alias.
28211         (fwrapv-pointer): New option.
28212         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
28213         also for pointer types based on flag_wrapv_pointer.
28214         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
28215         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
28216         opts->x_flag_wrapv got set.
28217         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
28218         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
28219         POINTER_TYPE_OVERFLOW_UNDEFINED.
28220         * match.pd: Likewise in address comparison pattern.
28221         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
28223 2018-01-15  Richard Biener  <rguenther@suse.de>
28225         PR lto/83804
28226         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
28227         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
28228         Reset type names to their identifier if their TYPE_DECL doesn't
28229         have linkage (and thus is used for ODR and devirt).
28230         (save_debug_info_for_decl): Remove.
28231         (save_debug_info_for_type): Likewise.
28232         (add_tree_to_fld_list): Adjust.
28233         * tree-pretty-print.c (dump_generic_node): Make dumping of
28234         type names more robust.
28236 2018-01-15  Richard Biener  <rguenther@suse.de>
28238         * BASE-VER: Bump to 8.0.1.
28240 2018-01-14  Martin Sebor  <msebor@redhat.com>
28242         PR other/83508
28243         * builtins.c (check_access): Avoid warning when the no-warning bit
28244         is set.
28246 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
28248         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
28249         * ira-color (allocno_hard_regs_compare): Likewise.
28251 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
28253         PR target/83013
28254         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
28255         Use .pushsection/.popsection.
28257 2018-01-14  Martin Sebor  <msebor@redhat.com>
28259         PR c++/81327
28260         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
28262 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
28264         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
28265         entry from extra_headers.
28266         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
28267         extra_headers, make the list bitwise identical to the i?86-*-* one.
28269 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28271         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
28272         -mcmodel=large with -mindirect-branch=thunk,
28273         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
28274         -mfunction-return=thunk-extern.
28275         * doc/invoke.texi: Document -mcmodel=large is incompatible with
28276         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
28277         -mfunction-return=thunk and -mfunction-return=thunk-extern.
28279 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28281         * config/i386/i386.c (print_reg): Print the name of the full
28282         integer register without '%'.
28283         (ix86_print_operand): Handle 'V'.
28284         * doc/extend.texi: Document 'V' modifier.
28286 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28288         * config/i386/constraints.md (Bs): Disallow memory operand for
28289         -mindirect-branch-register.
28290         (Bw): Likewise.
28291         * config/i386/predicates.md (indirect_branch_operand): Likewise.
28292         (GOT_memory_operand): Likewise.
28293         (call_insn_operand): Likewise.
28294         (sibcall_insn_operand): Likewise.
28295         (GOT32_symbol_operand): Likewise.
28296         * config/i386/i386.md (indirect_jump): Call convert_memory_address
28297         for -mindirect-branch-register.
28298         (tablejump): Likewise.
28299         (*sibcall_memory): Likewise.
28300         (*sibcall_value_memory): Likewise.
28301         Disallow peepholes of indirect call and jump via memory for
28302         -mindirect-branch-register.
28303         (*call_pop): Replace m with Bw.
28304         (*call_value_pop): Likewise.
28305         (*sibcall_pop_memory): Replace m with Bs.
28306         * config/i386/i386.opt (mindirect-branch-register): New option.
28307         * doc/invoke.texi: Document -mindirect-branch-register option.
28309 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28311         * config/i386/i386-protos.h (ix86_output_function_return): New.
28312         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
28313         set function_return_type.
28314         (indirect_thunk_name): Add ret_p to indicate thunk for function
28315         return.
28316         (output_indirect_thunk_function): Pass false to
28317         indirect_thunk_name.
28318         (ix86_output_indirect_branch_via_reg): Likewise.
28319         (ix86_output_indirect_branch_via_push): Likewise.
28320         (output_indirect_thunk_function): Create alias for function
28321         return thunk if regno < 0.
28322         (ix86_output_function_return): New function.
28323         (ix86_handle_fndecl_attribute): Handle function_return.
28324         (ix86_attribute_table): Add function_return.
28325         * config/i386/i386.h (machine_function): Add
28326         function_return_type.
28327         * config/i386/i386.md (simple_return_internal): Use
28328         ix86_output_function_return.
28329         (simple_return_internal_long): Likewise.
28330         * config/i386/i386.opt (mfunction-return=): New option.
28331         (indirect_branch): Mention -mfunction-return=.
28332         * doc/extend.texi: Document function_return function attribute.
28333         * doc/invoke.texi: Document -mfunction-return= option.
28335 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28337         * config/i386/i386-opts.h (indirect_branch): New.
28338         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
28339         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
28340         with local indirect jump when converting indirect call and jump.
28341         (ix86_set_indirect_branch_type): New.
28342         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
28343         (indirectlabelno): New.
28344         (indirect_thunk_needed): Likewise.
28345         (indirect_thunk_bnd_needed): Likewise.
28346         (indirect_thunks_used): Likewise.
28347         (indirect_thunks_bnd_used): Likewise.
28348         (INDIRECT_LABEL): Likewise.
28349         (indirect_thunk_name): Likewise.
28350         (output_indirect_thunk): Likewise.
28351         (output_indirect_thunk_function): Likewise.
28352         (ix86_output_indirect_branch_via_reg): Likewise.
28353         (ix86_output_indirect_branch_via_push): Likewise.
28354         (ix86_output_indirect_branch): Likewise.
28355         (ix86_output_indirect_jmp): Likewise.
28356         (ix86_code_end): Call output_indirect_thunk_function if needed.
28357         (ix86_output_call_insn): Call ix86_output_indirect_branch if
28358         needed.
28359         (ix86_handle_fndecl_attribute): Handle indirect_branch.
28360         (ix86_attribute_table): Add indirect_branch.
28361         * config/i386/i386.h (machine_function): Add indirect_branch_type
28362         and has_local_indirect_jump.
28363         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
28364         to true.
28365         (tablejump): Likewise.
28366         (*indirect_jump): Use ix86_output_indirect_jmp.
28367         (*tablejump_1): Likewise.
28368         (simple_return_indirect_internal): Likewise.
28369         * config/i386/i386.opt (mindirect-branch=): New option.
28370         (indirect_branch): New.
28371         (keep): Likewise.
28372         (thunk): Likewise.
28373         (thunk-inline): Likewise.
28374         (thunk-extern): Likewise.
28375         * doc/extend.texi: Document indirect_branch function attribute.
28376         * doc/invoke.texi: Document -mindirect-branch= option.
28378 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
28380         PR ipa/83051
28381         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
28383 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
28385         * ipa-inline.c (want_inline_small_function_p): Return false if
28386         inlining has already failed with CIF_FINAL_ERROR.
28387         (update_caller_keys): Call want_inline_small_function_p before
28388         can_inline_edge_p.
28389         (update_callee_keys): Likewise.
28391 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
28393         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
28394         New function.
28395         (rs6000_quadword_masked_address_p): Likewise.
28396         (quad_aligned_load_p): Likewise.
28397         (quad_aligned_store_p): Likewise.
28398         (const_load_sequence_p): Add comment to describe the outer-most loop.
28399         (mimic_memory_attributes_and_flags): New function.
28400         (rs6000_gen_stvx): Likewise.
28401         (replace_swapped_aligned_store): Likewise.
28402         (rs6000_gen_lvx): Likewise.
28403         (replace_swapped_aligned_load): Likewise.
28404         (replace_swapped_load_constant): Capitalize argument name in
28405         comment describing this function.
28406         (rs6000_analyze_swaps): Add a third pass to search for vector loads
28407         and stores that access quad-word aligned addresses and replace
28408         with stvx or lvx instructions when appropriate.
28409         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
28410         New function prototype.
28411         (rs6000_quadword_masked_address_p): Likewise.
28412         (rs6000_gen_lvx): Likewise.
28413         (rs6000_gen_stvx): Likewise.
28414         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
28415         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
28416         when memory address is aligned.
28417         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
28418         this split to select lvx instruction when memory address is aligned.
28419         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
28420         instruction when memory address is aligned.
28421         (*vsx_le_perm_load_v16qi): Likewise.
28422         (four unnamed splitters): Modify to select the stvx instruction
28423         when memory is aligned.
28425 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
28427         * predict.c (determine_unlikely_bbs): Handle correctly BBs
28428         which appears in the queue multiple times.
28430 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28431             Alan Hayward  <alan.hayward@arm.com>
28432             David Sherwood  <david.sherwood@arm.com>
28434         * tree-vectorizer.h (vec_lower_bound): New structure.
28435         (_loop_vec_info): Add check_nonzero and lower_bounds.
28436         (LOOP_VINFO_CHECK_NONZERO): New macro.
28437         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
28438         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
28439         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
28440         fields.  Make seg_len the distance travelled, not including the
28441         access size.
28442         (dr_direction_indicator): Declare.
28443         (dr_zero_step_indicator): Likewise.
28444         (dr_known_forward_stride_p): Likewise.
28445         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
28446         tree-ssanames.h.
28447         (runtime_alias_check_p): Allow runtime alias checks with
28448         variable strides.
28449         (operator ==): Compare access_size and align.
28450         (prune_runtime_alias_test_list): Rework for new distinction between
28451         the access_size and seg_len.
28452         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
28453         segment lengths.
28454         (get_segment_min_max): New function.
28455         (create_intersect_range_checks): Use it.
28456         (dr_step_indicator): New function.
28457         (dr_direction_indicator): Likewise.
28458         (dr_zero_step_indicator): Likewise.
28459         (dr_known_forward_stride_p): Likewise.
28460         * tree-loop-distribution.c (data_ref_segment_size): Return
28461         DR_STEP * (niters - 1).
28462         (compute_alias_check_pairs): Update call to the dr_with_seg_len
28463         constructor.
28464         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
28465         (vect_preserves_scalar_order_p): New function, split out from...
28466         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
28467         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
28468         (vect_vfa_access_size): New function.
28469         (vect_vfa_align): Likewise.
28470         (vect_compile_time_alias): Take access_size_a and access_b arguments.
28471         (dump_lower_bound): New function.
28472         (vect_check_lower_bound): Likewise.
28473         (vect_small_gap_p): Likewise.
28474         (vectorizable_with_step_bound_p): Likewise.
28475         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
28476         depencies if the vectorization factor is 1.  Convert the checks
28477         for nonzero steps into checks on the bounds of DR_STEP.  Try using
28478         a bunds check for variable steps if the minimum required step is
28479         relatively small. Update calls to the dr_with_seg_len
28480         constructor and to vect_compile_time_alias.
28481         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
28482         function.
28483         (vect_loop_versioning): Call it.
28484         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
28485         when retrying.
28486         (vect_estimate_min_profitable_iters): Account for any bounds checks.
28488 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28489             Alan Hayward  <alan.hayward@arm.com>
28490             David Sherwood  <david.sherwood@arm.com>
28492         * doc/sourcebuild.texi (vect_scatter_store): Document.
28493         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
28494         optabs.
28495         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
28496         Document.
28497         * genopinit.c (main): Add supports_vec_scatter_store and
28498         supports_vec_scatter_store_cached to target_optabs.
28499         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
28500         IFN_MASK_SCATTER_STORE.
28501         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
28502         functions.
28503         * internal-fn.h (internal_store_fn_p): Declare.
28504         (internal_fn_stored_value_index): Likewise.
28505         * internal-fn.c (scatter_store_direct): New macro.
28506         (expand_scatter_store_optab_fn): New function.
28507         (direct_scatter_store_optab_supported_p): New macro.
28508         (internal_store_fn_p): New function.
28509         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
28510         IFN_MASK_SCATTER_STORE.
28511         (internal_fn_mask_index): Likewise.
28512         (internal_fn_stored_value_index): New function.
28513         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
28514         for scatter stores.
28515         * optabs-query.h (supports_vec_scatter_store_p): Declare.
28516         * optabs-query.c (supports_vec_scatter_store_p): New function.
28517         * tree-vectorizer.h (vect_get_store_rhs): Declare.
28518         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
28519         true for scatter stores.
28520         (vect_gather_scatter_fn_p): Handle scatter stores too.
28521         (vect_check_gather_scatter): Consider using scatter stores if
28522         supports_vec_scatter_store_p.
28523         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
28524         scatter stores too.
28525         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28526         internal_fn_stored_value_index.
28527         (check_load_store_masking): Handle scatter stores too.
28528         (vect_get_store_rhs): Make public.
28529         (vectorizable_call): Use internal_store_fn_p.
28530         (vectorizable_store): Handle scatter store internal functions.
28531         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
28532         when deciding whether the end of the group has been reached.
28533         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
28534         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
28535         (mask_scatter_store<mode>): New insns.
28537 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28538             Alan Hayward  <alan.hayward@arm.com>
28539             David Sherwood  <david.sherwood@arm.com>
28541         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
28542         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
28543         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
28544         function.
28545         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
28546         Use vect_truncate_gather_scatter_offset if we can't treat the
28547         operation as a normal gather load or scatter store.
28548         (get_group_load_store_type): Take the gather_scatter_info
28549         as argument.  Try using a gather load or scatter store for
28550         single-element groups.
28551         (get_load_store_type): Update calls to get_group_load_store_type
28552         and vect_use_strided_gather_scatters_p.
28554 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28555             Alan Hayward  <alan.hayward@arm.com>
28556             David Sherwood  <david.sherwood@arm.com>
28558         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
28559         optional tree argument.
28560         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
28561         null target hooks.
28562         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
28563         but continue to use the current value as a fallback.
28564         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
28565         to compare the updates.
28566         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
28567         (get_load_store_type): Use it when handling a strided access.
28568         (vect_get_strided_load_store_ops): New function.
28569         (vect_get_data_ptr_increment): Likewise.
28570         (vectorizable_load): Handle strided gather loads.  Always pass
28571         a step to vect_create_data_ref_ptr and bump_vector_ptr.
28573 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28574             Alan Hayward  <alan.hayward@arm.com>
28575             David Sherwood  <david.sherwood@arm.com>
28577         * doc/md.texi (gather_load@var{m}): Document.
28578         (mask_gather_load@var{m}): Likewise.
28579         * genopinit.c (main): Add supports_vec_gather_load and
28580         supports_vec_gather_load_cached to target_optabs.
28581         * optabs-tree.c (init_tree_optimization_optabs): Use
28582         ggc_cleared_alloc to allocate target_optabs.
28583         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
28584         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
28585         functions.
28586         * internal-fn.h (internal_load_fn_p): Declare.
28587         (internal_gather_scatter_fn_p): Likewise.
28588         (internal_fn_mask_index): Likewise.
28589         (internal_gather_scatter_fn_supported_p): Likewise.
28590         * internal-fn.c (gather_load_direct): New macro.
28591         (expand_gather_load_optab_fn): New function.
28592         (direct_gather_load_optab_supported_p): New macro.
28593         (direct_internal_fn_optab): New function.
28594         (internal_load_fn_p): Likewise.
28595         (internal_gather_scatter_fn_p): Likewise.
28596         (internal_fn_mask_index): Likewise.
28597         (internal_gather_scatter_fn_supported_p): Likewise.
28598         * optabs-query.c (supports_at_least_one_mode_p): New function.
28599         (supports_vec_gather_load_p): Likewise.
28600         * optabs-query.h (supports_vec_gather_load_p): Declare.
28601         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
28602         and memory_type field.
28603         (NUM_PATTERNS): Bump to 15.
28604         * tree-vect-data-refs.c: Include internal-fn.h.
28605         (vect_gather_scatter_fn_p): New function.
28606         (vect_describe_gather_scatter_call): Likewise.
28607         (vect_check_gather_scatter): Try using internal functions for
28608         gather loads.  Recognize existing calls to a gather load function.
28609         (vect_analyze_data_refs): Consider using gather loads if
28610         supports_vec_gather_load_p.
28611         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
28612         (vect_get_gather_scatter_offset_type): Likewise.
28613         (vect_convert_mask_for_vectype): Likewise.
28614         (vect_add_conversion_to_patterm): Likewise.
28615         (vect_try_gather_scatter_pattern): Likewise.
28616         (vect_recog_gather_scatter_pattern): New pattern recognizer.
28617         (vect_vect_recog_func_ptrs): Add it.
28618         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28619         internal_fn_mask_index and internal_gather_scatter_fn_p.
28620         (check_load_store_masking): Take the gather_scatter_info as an
28621         argument and handle gather loads.
28622         (vect_get_gather_scatter_ops): New function.
28623         (vectorizable_call): Check internal_load_fn_p.
28624         (vectorizable_load): Likewise.  Handle gather load internal
28625         functions.
28626         (vectorizable_store): Update call to check_load_store_masking.
28627         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
28628         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
28629         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
28630         (aarch64_gather_scale_operand_d): New predicates.
28631         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
28632         (mask_gather_load<mode>): New insns.
28634 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28635             Alan Hayward  <alan.hayward@arm.com>
28636             David Sherwood  <david.sherwood@arm.com>
28638         * optabs.def (fold_left_plus_optab): New optab.
28639         * doc/md.texi (fold_left_plus_@var{m}): Document.
28640         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
28641         * internal-fn.c (fold_left_direct): Define.
28642         (expand_fold_left_optab_fn): Likewise.
28643         (direct_fold_left_optab_supported_p): Likewise.
28644         * fold-const-call.c (fold_const_fold_left): New function.
28645         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
28646         * tree-parloops.c (valid_reduction_p): New function.
28647         (gather_scalar_reductions): Use it.
28648         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
28649         (vect_finish_replace_stmt): Declare.
28650         * tree-vect-loop.c (fold_left_reduction_fn): New function.
28651         (needs_fold_left_reduction_p): New function, split out from...
28652         (vect_is_simple_reduction): ...here.  Accept reductions that
28653         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
28654         (vect_force_simple_reduction): Also store the reduction type in
28655         the assignment's STMT_VINFO_REDUC_TYPE.
28656         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
28657         (merge_with_identity): New function.
28658         (vect_expand_fold_left): Likewise.
28659         (vectorize_fold_left_reduction): Likewise.
28660         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
28661         scalar phi in place for it.  Check for target support and reject
28662         cases that would reassociate the operation.  Defer the transform
28663         phase to vectorize_fold_left_reduction.
28664         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
28665         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
28666         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
28668 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28670         * tree-if-conv.c (predicate_mem_writes): Remove redundant
28671         call to ifc_temp_var.
28673 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28674             Alan Hayward  <alan.hayward@arm.com>
28675             David Sherwood  <david.sherwood@arm.com>
28677         * target.def (legitimize_address_displacement): Take the original
28678         offset as a poly_int.
28679         * targhooks.h (default_legitimize_address_displacement): Update
28680         accordingly.
28681         * targhooks.c (default_legitimize_address_displacement): Likewise.
28682         * doc/tm.texi: Regenerate.
28683         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
28684         as an argument, moving assert of ad->disp == ad->disp_term to...
28685         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
28686         Try calling targetm.legitimize_address_displacement before expanding
28687         the address rather than afterwards, and adjust for the new interface.
28688         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
28689         Match the new hook interface.  Handle SVE addresses.
28690         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
28691         new hook interface.
28693 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28695         * Makefile.in (OBJS): Add early-remat.o.
28696         * target.def (select_early_remat_modes): New hook.
28697         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
28698         * doc/tm.texi: Regenerate.
28699         * targhooks.h (default_select_early_remat_modes): Declare.
28700         * targhooks.c (default_select_early_remat_modes): New function.
28701         * timevar.def (TV_EARLY_REMAT): New timevar.
28702         * passes.def (pass_early_remat): New pass.
28703         * tree-pass.h (make_pass_early_remat): Declare.
28704         * early-remat.c: New file.
28705         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
28706         function.
28707         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
28709 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28710             Alan Hayward  <alan.hayward@arm.com>
28711             David Sherwood  <david.sherwood@arm.com>
28713         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
28714         vfm1 with a bound_epilog parameter.
28715         (vect_do_peeling): Update calls accordingly, and move the prologue
28716         call earlier in the function.  Treat the base bound_epilog as 0 for
28717         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
28718         this base when peeling for gaps.
28719         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
28720         with fully-masked loops.
28721         (vect_estimate_min_profitable_iters): Handle the single peeled
28722         iteration in that case.
28724 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28725             Alan Hayward  <alan.hayward@arm.com>
28726             David Sherwood  <david.sherwood@arm.com>
28728         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
28729         single-element interleaving even if the size is not a power of 2.
28730         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
28731         accesses for single-element interleaving if the group size is
28732         not a power of 2.
28734 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28735             Alan Hayward  <alan.hayward@arm.com>
28736             David Sherwood  <david.sherwood@arm.com>
28738         * doc/md.texi (fold_extract_last_@var{m}): Document.
28739         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
28740         * optabs.def (fold_extract_last_optab): New optab.
28741         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
28742         * internal-fn.c (fold_extract_direct): New macro.
28743         (expand_fold_extract_optab_fn): Likewise.
28744         (direct_fold_extract_optab_supported_p): Likewise.
28745         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
28746         * tree-vect-loop.c (vect_model_reduction_cost): Handle
28747         EXTRACT_LAST_REDUCTION.
28748         (get_initial_def_for_reduction): Do not create an initial vector
28749         for EXTRACT_LAST_REDUCTION reductions.
28750         (vectorizable_reduction): Leave the scalar phi in place for
28751         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
28752         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
28753         epilogue code for EXTRACT_LAST_REDUCTION and defer the
28754         transform phase to vectorizable_condition.
28755         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
28756         split out from...
28757         (vect_finish_stmt_generation): ...here.
28758         (vect_finish_replace_stmt): New function.
28759         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
28760         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
28761         pattern.
28762         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
28764 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28765             Alan Hayward  <alan.hayward@arm.com>
28766             David Sherwood  <david.sherwood@arm.com>
28768         * doc/md.texi (extract_last_@var{m}): Document.
28769         * optabs.def (extract_last_optab): New optab.
28770         * internal-fn.def (EXTRACT_LAST): New internal function.
28771         * internal-fn.c (cond_unary_direct): New macro.
28772         (expand_cond_unary_optab_fn): Likewise.
28773         (direct_cond_unary_optab_supported_p): Likewise.
28774         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
28775         loops using EXTRACT_LAST.
28776         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
28777         (extract_last_<mode>): ...this optab.
28778         (vec_extract<mode><Vel>): Update accordingly.
28780 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28781             Alan Hayward  <alan.hayward@arm.com>
28782             David Sherwood  <david.sherwood@arm.com>
28784         * target.def (empty_mask_is_expensive): New hook.
28785         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
28786         * doc/tm.texi: Regenerate.
28787         * targhooks.h (default_empty_mask_is_expensive): Declare.
28788         * targhooks.c (default_empty_mask_is_expensive): New function.
28789         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
28790         if the target says that empty masks are expensive.
28791         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
28792         New function.
28793         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
28795 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28796             Alan Hayward  <alan.hayward@arm.com>
28797             David Sherwood  <david.sherwood@arm.com>
28799         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
28800         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
28801         (vect_use_loop_mask_for_alignment_p): New function.
28802         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
28803         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
28804         niters_skip argument.  Make sure that the first niters_skip elements
28805         of the first iteration are inactive.
28806         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
28807         Update call to vect_set_loop_masks_directly.
28808         (get_misalign_in_elems): New function, split out from...
28809         (vect_gen_prolog_loop_niters): ...here.
28810         (vect_update_init_of_dr): Take a code argument that specifies whether
28811         the adjustment should be added or subtracted.
28812         (vect_update_init_of_drs): Likewise.
28813         (vect_prepare_for_masked_peels): New function.
28814         (vect_do_peeling): Skip prologue peeling if we're using a mask
28815         instead.  Update call to vect_update_inits_of_drs.
28816         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28817         mask_skip_niters.
28818         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
28819         alignment.  Do not include the number of peeled iterations in
28820         the minimum threshold in that case.
28821         (vectorizable_induction): Adjust the start value down by
28822         LOOP_VINFO_MASK_SKIP_NITERS iterations.
28823         (vect_transform_loop): Call vect_prepare_for_masked_peels.
28824         Take the number of skipped iterations into account when calculating
28825         the loop bounds.
28826         * tree-vect-stmts.c (vect_gen_while_not): New function.
28828 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28829             Alan Hayward  <alan.hayward@arm.com>
28830             David Sherwood  <david.sherwood@arm.com>
28832         * doc/sourcebuild.texi (vect_fully_masked): Document.
28833         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
28834         default value to 0.
28835         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
28836         split out from...
28837         (vect_analyze_loop_2): ...here. Don't check the vectorization
28838         factor against the number of loop iterations if the loop is
28839         fully-masked.
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         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
28846         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
28847         (dump_groups): Update accordingly.
28848         (iv_use::mem_type): New member variable.
28849         (address_p): New function.
28850         (record_use): Add a mem_type argument and initialize the new
28851         mem_type field.
28852         (record_group_use): Add a mem_type argument.  Use address_p.
28853         Remove obsolete null checks of base_object.  Update call to record_use.
28854         (find_interesting_uses_op): Update call to record_group_use.
28855         (find_interesting_uses_cond): Likewise.
28856         (find_interesting_uses_address): Likewise.
28857         (get_mem_type_for_internal_fn): New function.
28858         (find_address_like_use): Likewise.
28859         (find_interesting_uses_stmt): Try find_address_like_use before
28860         calling find_interesting_uses_op.
28861         (addr_offset_valid_p): Use the iv mem_type field as the type
28862         of the addressed memory.
28863         (add_autoinc_candidates): Likewise.
28864         (get_address_cost): Likewise.
28865         (split_small_address_groups_p): Use address_p.
28866         (split_address_groups): Likewise.
28867         (add_iv_candidate_for_use): Likewise.
28868         (autoinc_possible_for_pair): Likewise.
28869         (rewrite_groups): Likewise.
28870         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
28871         (determine_group_iv_cost): Update after split of USE_ADDRESS.
28872         (get_alias_ptr_type_for_ptr_address): New function.
28873         (rewrite_use_address): Rewrite address uses in calls that were
28874         identified by find_address_like_use.
28876 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28877             Alan Hayward  <alan.hayward@arm.com>
28878             David Sherwood  <david.sherwood@arm.com>
28880         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
28881         TARGET_MEM_REFs.
28882         * gimple-expr.h (is_gimple_addressable: Likewise.
28883         * gimple-expr.c (is_gimple_address): Likewise.
28884         * internal-fn.c (expand_call_mem_ref): New function.
28885         (expand_mask_load_optab_fn): Use it.
28886         (expand_mask_store_optab_fn): Likewise.
28888 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28889             Alan Hayward  <alan.hayward@arm.com>
28890             David Sherwood  <david.sherwood@arm.com>
28892         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
28893         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
28894         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
28895         (cond_umax@var{mode}): Document.
28896         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
28897         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
28898         (cond_umin_optab, cond_umax_optab): New optabs.
28899         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
28900         (COND_IOR, COND_XOR): New internal functions.
28901         * internal-fn.h (get_conditional_internal_fn): Declare.
28902         * internal-fn.c (cond_binary_direct): New macro.
28903         (expand_cond_binary_optab_fn): Likewise.
28904         (direct_cond_binary_optab_supported_p): Likewise.
28905         (get_conditional_internal_fn): New function.
28906         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
28907         Cope with reduction statements that are vectorized as calls rather
28908         than assignments.
28909         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
28910         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
28911         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
28912         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
28913         (UNSPEC_COND_EOR): New unspecs.
28914         (optab): Add mappings for them.
28915         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
28916         (sve_int_op, sve_fp_op): New int attributes.
28918 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28919             Alan Hayward  <alan.hayward@arm.com>
28920             David Sherwood  <david.sherwood@arm.com>
28922         * optabs.def (while_ult_optab): New optab.
28923         * doc/md.texi (while_ult@var{m}@var{n}): Document.
28924         * internal-fn.def (WHILE_ULT): New internal function.
28925         * internal-fn.h (direct_internal_fn_supported_p): New override
28926         that takes two types as argument.
28927         * internal-fn.c (while_direct): New macro.
28928         (expand_while_optab_fn): New function.
28929         (convert_optab_supported_p): Likewise.
28930         (direct_while_optab_supported_p): New macro.
28931         * wide-int.h (wi::udiv_ceil): New function.
28932         * tree-vectorizer.h (rgroup_masks): New structure.
28933         (vec_loop_masks): New typedef.
28934         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
28935         and fully_masked_p.
28936         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
28937         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
28938         (vect_max_vf): New function.
28939         (slpeel_make_loop_iterate_ntimes): Delete.
28940         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
28941         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
28942         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
28943         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
28944         internal-fn.h, stor-layout.h and optabs-query.h.
28945         (vect_set_loop_mask): New function.
28946         (add_preheader_seq): Likewise.
28947         (add_header_seq): Likewise.
28948         (interleave_supported_p): Likewise.
28949         (vect_maybe_permute_loop_masks): Likewise.
28950         (vect_set_loop_masks_directly): Likewise.
28951         (vect_set_loop_condition_masked): Likewise.
28952         (vect_set_loop_condition_unmasked): New function, split out from
28953         slpeel_make_loop_iterate_ntimes.
28954         (slpeel_make_loop_iterate_ntimes): Rename to..
28955         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
28956         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
28957         (vect_do_peeling): Update call accordingly.
28958         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
28959         loops.
28960         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28961         mask_compare_type, can_fully_mask_p and fully_masked_p.
28962         (release_vec_loop_masks): New function.
28963         (_loop_vec_info): Use it to free the loop masks.
28964         (can_produce_all_loop_masks_p): New function.
28965         (vect_get_max_nscalars_per_iter): Likewise.
28966         (vect_verify_full_masking): Likewise.
28967         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
28968         retries, and free the mask rgroups before retrying.  Check loop-wide
28969         reasons for disallowing fully-masked loops.  Make the final decision
28970         about whether use a fully-masked loop or not.
28971         (vect_estimate_min_profitable_iters): Do not assume that peeling
28972         for the number of iterations will be needed for fully-masked loops.
28973         (vectorizable_reduction): Disable fully-masked loops.
28974         (vectorizable_live_operation): Likewise.
28975         (vect_halve_mask_nunits): New function.
28976         (vect_double_mask_nunits): Likewise.
28977         (vect_record_loop_mask): Likewise.
28978         (vect_get_loop_mask): Likewise.
28979         (vect_transform_loop): Handle the case in which the final loop
28980         iteration might handle a partial vector.  Call vect_set_loop_condition
28981         instead of slpeel_make_loop_iterate_ntimes.
28982         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
28983         (check_load_store_masking): New function.
28984         (prepare_load_store_mask): Likewise.
28985         (vectorizable_store): Handle fully-masked loops.
28986         (vectorizable_load): Likewise.
28987         (supportable_widening_operation): Use vect_halve_mask_nunits for
28988         booleans.
28989         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
28990         (vect_gen_while): New function.
28991         * config/aarch64/aarch64.md (umax<mode>3): New expander.
28992         (aarch64_uqdec<mode>): New insn.
28994 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28995             Alan Hayward  <alan.hayward@arm.com>
28996             David Sherwood  <david.sherwood@arm.com>
28998         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
28999         (reduc_xor_scal_optab): New optabs.
29000         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
29001         (reduc_xor_scal_@var{m}): Document.
29002         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
29003         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
29004         internal functions.
29005         * fold-const-call.c (fold_const_call): Handle them.
29006         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
29007         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
29008         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
29009         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
29010         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
29011         (UNSPEC_XORV): New unspecs.
29012         (optab): Add entries for them.
29013         (BITWISEV): New int iterator.
29014         (bit_reduc_op): New int attributes.
29016 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29017             Alan Hayward  <alan.hayward@arm.com>
29018             David Sherwood  <david.sherwood@arm.com>
29020         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
29021         * internal-fn.def (VEC_SHL_INSERT): New internal function.
29022         * optabs.def (vec_shl_insert_optab): New optab.
29023         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
29024         (duplicate_and_interleave): Likewise.
29025         * tree-vect-loop.c: Include internal-fn.h.
29026         (neutral_op_for_slp_reduction): New function, split out from
29027         get_initial_defs_for_reduction.
29028         (get_initial_def_for_reduction): Handle option 2 for variable-length
29029         vectors by loading the neutral value into a vector and then shifting
29030         the initial value into element 0.
29031         (get_initial_defs_for_reduction): Replace the code argument with
29032         the neutral value calculated by neutral_op_for_slp_reduction.
29033         Use gimple_build_vector for constant-length vectors.
29034         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
29035         but the first group_size elements have a neutral value.
29036         Use duplicate_and_interleave otherwise.
29037         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
29038         Update call to get_initial_defs_for_reduction.  Handle SLP
29039         reductions for variable-length vectors by creating one vector
29040         result for each scalar result, with the elements associated
29041         with other scalar results stubbed out with the neutral value.
29042         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
29043         Require IFN_VEC_SHL_INSERT for double reductions on
29044         variable-length vectors, or SLP reductions that have
29045         a neutral value.  Require can_duplicate_and_interleave_p
29046         support for variable-length unchained SLP reductions if there
29047         is no neutral value, such as for MIN/MAX reductions.  Also require
29048         the number of vector elements to be a multiple of the number of
29049         SLP statements when doing variable-length unchained SLP reductions.
29050         Update call to vect_create_epilog_for_reduction.
29051         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
29052         and remove initial values.
29053         (duplicate_and_interleave): Make public.
29054         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
29055         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
29057 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29058             Alan Hayward  <alan.hayward@arm.com>
29059             David Sherwood  <david.sherwood@arm.com>
29061         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
29062         (can_duplicate_and_interleave_p): New function.
29063         (vect_get_and_check_slp_defs): Take the vector of statements
29064         rather than just the current one.  Remove excess parentheses.
29065         Restriction rejectinon of vect_constant_def and vect_external_def
29066         for variable-length vectors to boolean types, or types for which
29067         can_duplicate_and_interleave_p is false.
29068         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
29069         (duplicate_and_interleave): New function.
29070         (vect_get_constant_vectors): Use gimple_build_vector for
29071         constant-length vectors and suitable variable-length constant
29072         vectors.  Use duplicate_and_interleave for other variable-length
29073         vectors.  Don't defer the update when inserting new statements.
29075 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29076             Alan Hayward  <alan.hayward@arm.com>
29077             David Sherwood  <david.sherwood@arm.com>
29079         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
29080         min_profitable_iters doesn't go negative.
29082 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29083             Alan Hayward  <alan.hayward@arm.com>
29084             David Sherwood  <david.sherwood@arm.com>
29086         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
29087         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
29088         * optabs.def (vec_mask_load_lanes_optab): New optab.
29089         (vec_mask_store_lanes_optab): Likewise.
29090         * internal-fn.def (MASK_LOAD_LANES): New internal function.
29091         (MASK_STORE_LANES): Likewise.
29092         * internal-fn.c (mask_load_lanes_direct): New macro.
29093         (mask_store_lanes_direct): Likewise.
29094         (expand_mask_load_optab_fn): Handle masked operations.
29095         (expand_mask_load_lanes_optab_fn): New macro.
29096         (expand_mask_store_optab_fn): Handle masked operations.
29097         (expand_mask_store_lanes_optab_fn): New macro.
29098         (direct_mask_load_lanes_optab_supported_p): Likewise.
29099         (direct_mask_store_lanes_optab_supported_p): Likewise.
29100         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
29101         parameter.
29102         (vect_load_lanes_supported): Likewise.
29103         * tree-vect-data-refs.c (strip_conversion): New function.
29104         (can_group_stmts_p): Likewise.
29105         (vect_analyze_data_ref_accesses): Use it instead of checking
29106         for a pair of assignments.
29107         (vect_store_lanes_supported): Take a masked_p parameter.
29108         (vect_load_lanes_supported): Likewise.
29109         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
29110         vect_store_lanes_supported and vect_load_lanes_supported.
29111         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
29112         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
29113         parameter.  Don't allow gaps for masked accesses.
29114         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
29115         and vect_load_lanes_supported.
29116         (get_load_store_type): Take a masked_p parameter and update
29117         call to get_group_load_store_type.
29118         (vectorizable_store): Update call to get_load_store_type.
29119         Handle IFN_MASK_STORE_LANES.
29120         (vectorizable_load): Update call to get_load_store_type.
29121         Handle IFN_MASK_LOAD_LANES.
29123 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29124             Alan Hayward  <alan.hayward@arm.com>
29125             David Sherwood  <david.sherwood@arm.com>
29127         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
29128         modes for SVE.
29129         * config/aarch64/aarch64-protos.h
29130         (aarch64_sve_struct_memory_operand_p): Declare.
29131         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
29132         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
29133         (VPRED, vpred): Handle SVE structure modes.
29134         * config/aarch64/constraints.md (Utx): New constraint.
29135         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
29136         (aarch64_sve_struct_nonimmediate_operand): New predicates.
29137         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
29138         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
29139         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
29140         structure modes.  Split into pieces after RA.
29141         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
29142         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
29143         New patterns.
29144         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
29145         SVE structure modes.
29146         (aarch64_classify_address): Likewise.
29147         (sizetochar): Move earlier in file.
29148         (aarch64_print_operand): Handle SVE register lists.
29149         (aarch64_array_mode): New function.
29150         (aarch64_sve_struct_memory_operand_p): Likewise.
29151         (TARGET_ARRAY_MODE): Redefine.
29153 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29154             Alan Hayward  <alan.hayward@arm.com>
29155             David Sherwood  <david.sherwood@arm.com>
29157         * target.def (array_mode): New target hook.
29158         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
29159         * doc/tm.texi: Regenerate.
29160         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
29161         * hooks.c (hook_optmode_mode_uhwi_none): New function.
29162         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
29163         targetm.array_mode.
29164         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
29165         type sizes.
29167 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29168             Alan Hayward  <alan.hayward@arm.com>
29169             David Sherwood  <david.sherwood@arm.com>
29171         * fold-const.c (fold_binary_loc): Check the argument types
29172         rather than the result type when testing for a vector operation.
29174 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29176         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
29177         * doc/tm.texi: Regenerate.
29179 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29180             Alan Hayward  <alan.hayward@arm.com>
29181             David Sherwood  <david.sherwood@arm.com>
29183         * doc/invoke.texi (-msve-vector-bits=): Document new option.
29184         (sve): Document new AArch64 extension.
29185         * doc/md.texi (w): Extend the description of the AArch64
29186         constraint to include SVE vectors.
29187         (Upl, Upa): Document new AArch64 predicate constraints.
29188         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
29189         enum.
29190         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
29191         (msve-vector-bits=): New option.
29192         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
29193         SVE when these are disabled.
29194         (sve): New extension.
29195         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
29196         modes.  Adjust their number of units based on aarch64_sve_vg.
29197         (MAX_BITSIZE_MODE_ANY_MODE): Define.
29198         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
29199         aarch64_addr_query_type.
29200         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
29201         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
29202         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
29203         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
29204         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
29205         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
29206         (aarch64_simd_imm_zero_p): Delete.
29207         (aarch64_check_zero_based_sve_index_immediate): Declare.
29208         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29209         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29210         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29211         (aarch64_sve_float_mul_immediate_p): Likewise.
29212         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29213         rather than an rtx.
29214         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
29215         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
29216         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
29217         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
29218         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
29219         (aarch64_regmode_natural_size): Likewise.
29220         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
29221         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
29222         left one place.
29223         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
29224         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
29225         for VG and the SVE predicate registers.
29226         (V_ALIASES): Add a "z"-prefixed alias.
29227         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
29228         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
29229         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
29230         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
29231         (REG_CLASS_NAMES): Add entries for them.
29232         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
29233         and the predicate registers.
29234         (aarch64_sve_vg): Declare.
29235         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
29236         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
29237         (REGMODE_NATURAL_SIZE): Define.
29238         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
29239         SVE macros.
29240         * config/aarch64/aarch64.c: Include cfgrtl.h.
29241         (simd_immediate_info): Add a constructor for series vectors,
29242         and an associated step field.
29243         (aarch64_sve_vg): New variable.
29244         (aarch64_dbx_register_number): Handle VG and the predicate registers.
29245         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
29246         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
29247         (VEC_ANY_DATA, VEC_STRUCT): New constants.
29248         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
29249         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
29250         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
29251         (aarch64_get_mask_mode): New functions.
29252         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
29253         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
29254         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
29255         predicate modes and predicate registers.  Explicitly restrict
29256         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
29257         to store a vector mode if it is recognized by
29258         aarch64_classify_vector_mode.
29259         (aarch64_regmode_natural_size): New function.
29260         (aarch64_hard_regno_caller_save_mode): Return the original mode
29261         for predicates.
29262         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
29263         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
29264         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
29265         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
29266         functions.
29267         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
29268         does not overlap dest if the function is frame-related.  Handle
29269         SVE constants.
29270         (aarch64_split_add_offset): New function.
29271         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
29272         them aarch64_add_offset.
29273         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
29274         and update call to aarch64_sub_sp.
29275         (aarch64_add_cfa_expression): New function.
29276         (aarch64_expand_prologue): Pass extra temporary registers to the
29277         functions above.  Handle the case in which we need to emit new
29278         DW_CFA_expressions for registers that were originally saved
29279         relative to the stack pointer, but now have to be expressed
29280         relative to the frame pointer.
29281         (aarch64_output_mi_thunk): Pass extra temporary registers to the
29282         functions above.
29283         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
29284         IP0 and IP1 values for SVE frames.
29285         (aarch64_expand_vec_series): New function.
29286         (aarch64_expand_sve_widened_duplicate): Likewise.
29287         (aarch64_expand_sve_const_vector): Likewise.
29288         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
29289         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
29290         into the register, rather than emitting a SET directly.
29291         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
29292         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
29293         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
29294         (offset_9bit_signed_scaled_p): New functions.
29295         (aarch64_replicate_bitmask_imm): New function.
29296         (aarch64_bitmask_imm): Use it.
29297         (aarch64_cannot_force_const_mem): Reject expressions involving
29298         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
29299         (aarch64_classify_index): Handle SVE indices, by requiring
29300         a plain register index with a scale that matches the element size.
29301         (aarch64_classify_address): Handle SVE addresses.  Assert that
29302         the mode of the address is VOIDmode or an integer mode.
29303         Update call to aarch64_classify_symbol.
29304         (aarch64_classify_symbolic_expression): Update call to
29305         aarch64_classify_symbol.
29306         (aarch64_const_vec_all_in_range_p): New function.
29307         (aarch64_print_vector_float_operand): Likewise.
29308         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
29309         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
29310         and the FP immediates 1.0 and 0.5.
29311         (aarch64_print_address_internal): Handle SVE addresses.
29312         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
29313         (aarch64_regno_regclass): Handle predicate registers.
29314         (aarch64_secondary_reload): Handle big-endian reloads of SVE
29315         data modes.
29316         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
29317         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
29318         (aarch64_convert_sve_vector_bits): New function.
29319         (aarch64_override_options): Use it to handle -msve-vector-bits=.
29320         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29321         rather than an rtx.
29322         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
29323         Handle SVE vector and predicate modes.  Accept VL-based constants
29324         that need only one temporary register, and VL offsets that require
29325         no temporary registers.
29326         (aarch64_conditional_register_usage): Mark the predicate registers
29327         as fixed if SVE isn't available.
29328         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
29329         Return true for SVE vector and predicate modes.
29330         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
29331         rather than an unsigned int.  Handle SVE modes.
29332         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
29333         SVE modes.
29334         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
29335         if SVE is enabled.
29336         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29337         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29338         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29339         (aarch64_sve_float_mul_immediate_p): New functions.
29340         (aarch64_sve_valid_immediate): New function.
29341         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
29342         Explicitly reject structure modes.  Check for INDEX constants.
29343         Handle PTRUE and PFALSE constants.
29344         (aarch64_check_zero_based_sve_index_immediate): New function.
29345         (aarch64_simd_imm_zero_p): Delete.
29346         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
29347         vector modes.  Accept constants in the range of CNT[BHWD].
29348         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
29349         ask for an Advanced SIMD mode.
29350         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
29351         (aarch64_simd_vector_alignment): Handle SVE predicates.
29352         (aarch64_vectorize_preferred_vector_alignment): New function.
29353         (aarch64_simd_vector_alignment_reachable): Use it instead of
29354         the vector size.
29355         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
29356         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
29357         functions.
29358         (MAX_VECT_LEN): Delete.
29359         (expand_vec_perm_d): Add a vec_flags field.
29360         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
29361         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
29362         (aarch64_evpc_ext): Don't apply a big-endian lane correction
29363         for SVE modes.
29364         (aarch64_evpc_rev): Rename to...
29365         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
29366         (aarch64_evpc_rev_global): New function.
29367         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
29368         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
29369         MAX_VECT_LEN.
29370         (aarch64_evpc_sve_tbl): New function.
29371         (aarch64_expand_vec_perm_const_1): Update after rename of
29372         aarch64_evpc_rev.  Handle SVE permutes too, trying
29373         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
29374         than aarch64_evpc_tbl.
29375         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
29376         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
29377         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
29378         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
29379         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
29380         (aarch64_expand_sve_vcond): New functions.
29381         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
29382         of aarch64_vector_mode_p.
29383         (aarch64_dwarf_poly_indeterminate_value): New function.
29384         (aarch64_compute_pressure_classes): Likewise.
29385         (aarch64_can_change_mode_class): Likewise.
29386         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
29387         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
29388         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
29389         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
29390         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
29391         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
29392         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
29393         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
29394         constraints.
29395         (Dn, Dl, Dr): Accept const as well as const_vector.
29396         (Dz): Likewise.  Compare against CONST0_RTX.
29397         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
29398         of "vector" where appropriate.
29399         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
29400         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
29401         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
29402         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
29403         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
29404         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
29405         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
29406         (v_int_equiv): Extend to SVE modes.
29407         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
29408         mode attributes.
29409         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
29410         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
29411         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
29412         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
29413         (SVE_COND_FP_CMP): New int iterators.
29414         (perm_hilo): Handle the new unpack unspecs.
29415         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
29416         attributes.
29417         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
29418         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
29419         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
29420         (aarch64_equality_operator, aarch64_constant_vector_operand)
29421         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
29422         (aarch64_sve_nonimmediate_operand): Likewise.
29423         (aarch64_sve_general_operand): Likewise.
29424         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
29425         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
29426         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
29427         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
29428         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
29429         (aarch64_sve_float_arith_immediate): Likewise.
29430         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
29431         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
29432         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
29433         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
29434         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
29435         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
29436         (aarch64_sve_float_arith_operand): Likewise.
29437         (aarch64_sve_float_arith_with_sub_operand): Likewise.
29438         (aarch64_sve_float_mul_operand): Likewise.
29439         (aarch64_sve_vec_perm_operand): Likewise.
29440         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
29441         (aarch64_mov_operand): Accept const_poly_int and const_vector.
29442         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
29443         as well as const_vector.
29444         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
29445         in file.  Use CONST0_RTX and CONSTM1_RTX.
29446         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
29447         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
29448         Use aarch64_simd_imm_zero.
29449         * config/aarch64/aarch64-sve.md: New file.
29450         * config/aarch64/aarch64.md: Include it.
29451         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
29452         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
29453         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
29454         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
29455         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
29456         (sve): New attribute.
29457         (enabled): Disable instructions with the sve attribute unless
29458         TARGET_SVE.
29459         (movqi, movhi): Pass CONST_POLY_INT operaneds through
29460         aarch64_expand_mov_immediate.
29461         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
29462         CNT[BHSD] immediates.
29463         (movti): Split CONST_POLY_INT moves into two halves.
29464         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
29465         Split additions that need a temporary here if the destination
29466         is the stack pointer.
29467         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
29468         (*add<mode>3_poly_1): New instruction.
29469         (set_clobber_cc): New expander.
29471 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29473         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
29474         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
29475         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
29476         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
29477         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
29478         Change innermode from fixed_mode_size to machine_mode.
29479         (simplify_subreg): Update call accordingly.  Handle a constant-sized
29480         subreg of a variable-length CONST_VECTOR.
29482 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29483             Alan Hayward  <alan.hayward@arm.com>
29484             David Sherwood  <david.sherwood@arm.com>
29486         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
29487         (add_offset_to_base): New function, split out from...
29488         (create_mem_ref): ...here.  When handling a scale other than 1,
29489         check first whether the address is valid without the offset.
29490         Add it into the base if so, leaving the index and scale as-is.
29492 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29494         PR c++/83778
29495         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
29496         fold_for_warn before checking if arg2 is INTEGER_CST.
29498 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
29500         * config/rs6000/predicates.md (load_multiple_operation): Delete.
29501         (store_multiple_operation): Delete.
29502         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
29503         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
29504         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
29505         guarded by TARGET_STRING.
29506         (rs6000_output_load_multiple): Delete.
29507         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29508         OPTION_MASK_STRING / TARGET_STRING handling.
29509         (print_operand) <'N', 'O'>: Add comment that these are unused now.
29510         (const rs6000_opt_masks) <"string">: Change mask to 0.
29511         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
29512         (MASK_STRING): Delete.
29513         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
29514         parts.  Simplify.
29515         (load_multiple): Delete.
29516         (*ldmsi8): Delete.
29517         (*ldmsi7): Delete.
29518         (*ldmsi6): Delete.
29519         (*ldmsi5): Delete.
29520         (*ldmsi4): Delete.
29521         (*ldmsi3): Delete.
29522         (store_multiple): Delete.
29523         (*stmsi8): Delete.
29524         (*stmsi7): Delete.
29525         (*stmsi6): Delete.
29526         (*stmsi5): Delete.
29527         (*stmsi4): Delete.
29528         (*stmsi3): Delete.
29529         (movmemsi_8reg): Delete.
29530         (corresponding unnamed define_insn): Delete.
29531         (movmemsi_6reg): Delete.
29532         (corresponding unnamed define_insn): Delete.
29533         (movmemsi_4reg): Delete.
29534         (corresponding unnamed define_insn): Delete.
29535         (movmemsi_2reg): Delete.
29536         (corresponding unnamed define_insn): Delete.
29537         (movmemsi_1reg): Delete.
29538         (corresponding unnamed define_insn): Delete.
29539         * config/rs6000/rs6000.opt (mno-string): New.
29540         (mstring): Replace by deprecation warning stub.
29541         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
29543 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29545         * regrename.c (regrename_do_replace): If replacing the same
29546         reg multiple times, try to reuse last created gen_raw_REG.
29548         PR debug/81155
29549         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
29550         main to workaround a bug in GDB.
29552 2018-01-12  Tom de Vries  <tom@codesourcery.com>
29554         PR target/83737
29555         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
29557 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
29559         PR rtl-optimization/80481
29560         * ira-color.c (get_cap_member): New function.
29561         (allocnos_conflict_by_live_ranges_p): Use it.
29562         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
29563         (setup_slot_coalesced_allocno_live_ranges): Ditto.
29565 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
29567         PR target/83628
29568         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
29569         (*saddl_se_1): Ditto.
29570         (*ssubsi_1): Ditto.
29571         (*ssubl_se_1): Ditto.
29573 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29575         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
29576         rather than wi::to_widest for DR_INITs.
29577         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
29578         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
29579         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
29580         INTEGER_CSTs.
29581         (vect_analyze_group_access_1): Note that here.
29583 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29585         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
29586         polynomial type sizes.
29588 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29590         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
29591         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
29592         (gimple_add_tmp_var): Likewise.
29594 2018-01-12  Martin Liska  <mliska@suse.cz>
29596         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
29597         (gimple_alloc_sizes): Likewise.
29598         (dump_gimple_statistics): Use PRIu64 in printf format.
29599         * gimple.h: Change uint64_t to int.
29601 2018-01-12  Martin Liska  <mliska@suse.cz>
29603         * tree-core.h: Use uint64_t instead of int.
29604         * tree.c (tree_node_counts): Likewise.
29605         (tree_node_sizes): Likewise.
29606         (dump_tree_statistics): Use PRIu64 in printf format.
29608 2018-01-12  Martin Liska  <mliska@suse.cz>
29610         * Makefile.in: As qsort_chk is implemented in vec.c, add
29611         vec.o to linkage of gencfn-macros.
29612         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
29613         passing the info to record_node_allocation_statistics.
29614         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
29615         and pass the info.
29616         * ggc-common.c (struct ggc_usage): Add operator== and use
29617         it in operator< and compare function.
29618         * mem-stats.h (struct mem_usage): Likewise.
29619         * vec.c (struct vec_usage): Remove operator< and compare
29620         function. Can be simply inherited.
29622 2018-01-12  Martin Jambor  <mjambor@suse.cz>
29624         PR target/81616
29625         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
29626         * tree-ssa-math-opts.c: Include domwalk.h.
29627         (convert_mult_to_fma_1): New function.
29628         (fma_transformation_info): New type.
29629         (fma_deferring_state): Likewise.
29630         (cancel_fma_deferring): New function.
29631         (result_of_phi): Likewise.
29632         (last_fma_candidate_feeds_initial_phi): Likewise.
29633         (convert_mult_to_fma): Added deferring logic, split actual
29634         transformation to convert_mult_to_fma_1.
29635         (math_opts_dom_walker): New type.
29636         (math_opts_dom_walker::after_dom_children): New method, body moved
29637         here from pass_optimize_widening_mul::execute, added deferring logic
29638         bits.
29639         (pass_optimize_widening_mul::execute): Moved most of code to
29640         math_opts_dom_walker::after_dom_children.
29641         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
29642         * config/i386/i386.c (ix86_option_override_internal): Added
29643         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
29645 2018-01-12  Richard Biener  <rguenther@suse.de>
29647         PR debug/83157
29648         * dwarf2out.c (gen_variable_die): Do not reset old_die for
29649         inline instance vars.
29651 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
29653         PR target/81819
29654         * config/rx/rx.c (rx_is_restricted_memory_address):
29655         Handle SUBREG case.
29657 2018-01-12  Richard Biener  <rguenther@suse.de>
29659         PR tree-optimization/80846
29660         * target.def (split_reduction): New target hook.
29661         * targhooks.c (default_split_reduction): New function.
29662         * targhooks.h (default_split_reduction): Declare.
29663         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
29664         target requests first reduce vectors by combining low and high
29665         parts.
29666         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
29667         (get_vectype_for_scalar_type_and_size): Export.
29668         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
29669         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
29670         * doc/tm.texi: Regenerate.
29671         * config/i386/i386.c (ix86_split_reduction): Implement
29672         TARGET_VECTORIZE_SPLIT_REDUCTION.
29674 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29676         PR target/83368
29677         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
29678         in PIC mode except for TARGET_VXWORKS_RTP.
29679         * config/sparc/sparc.c: Include cfgrtl.h.
29680         (TARGET_INIT_PIC_REG): Define.
29681         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
29682         (sparc_pic_register_p): New predicate.
29683         (sparc_legitimate_address_p): Use it.
29684         (sparc_legitimize_pic_address): Likewise.
29685         (sparc_delegitimize_address): Likewise.
29686         (sparc_mode_dependent_address_p): Likewise.
29687         (gen_load_pcrel_sym): Remove 4th parameter.
29688         (load_got_register): Adjust call to above.  Remove obsolete stuff.
29689         (sparc_expand_prologue): Do not call load_got_register here.
29690         (sparc_flat_expand_prologue): Likewise.
29691         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
29692         (sparc_use_pseudo_pic_reg): New function.
29693         (sparc_init_pic_reg): Likewise.
29694         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
29695         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
29697 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
29699         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
29700         Add item for branch_cost.
29702 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29704         PR rtl-optimization/83565
29705         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
29706         not extend the result to a larger mode for rotate operations.
29707         (num_sign_bit_copies1): Likewise.
29709 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29711         PR target/40411
29712         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
29713         -symbolic.
29714         Use values-Xc.o for -pedantic.
29715         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
29717 2018-01-12  Martin Liska  <mliska@suse.cz>
29719         PR ipa/83054
29720         * ipa-devirt.c (final_warning_record::grow_type_warnings):
29721         New function.
29722         (possible_polymorphic_call_targets): Use it.
29723         (ipa_devirt): Likewise.
29725 2018-01-12  Martin Liska  <mliska@suse.cz>
29727         * profile-count.h (enum profile_quality): Use 0 as invalid
29728         enum value of profile_quality.
29730 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
29732         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
29733         -mext-string options.
29735 2018-01-12  Richard Biener  <rguenther@suse.de>
29737         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
29738         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
29739         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
29740         Likewise.
29741         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
29743 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
29745         * configure.ac (--with-long-double-format): Add support for the
29746         configuration option to change the default long double format on
29747         PowerPC systems.
29748         * config.gcc (powerpc*-linux*-*): Likewise.
29749         * configure: Regenerate.
29750         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
29751         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
29752         used without modification.
29754 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29756         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
29757         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
29758         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
29759         MISC_BUILTIN_SPEC_BARRIER.
29760         (rs6000_init_builtins): Likewise.
29761         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
29762         enum value.
29763         (speculation_barrier): New define_insn.
29764         * doc/extend.texi: Document __builtin_speculation_barrier.
29766 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29768         PR target/83203
29769         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
29770         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
29771         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
29772         iterators.
29773         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
29774         integral modes instead of "ss" and "sd".
29775         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
29776         vectors with 32-bit and 64-bit elements.
29777         (vecdupssescalarmodesuffix): New mode attribute.
29778         (vec_dup<mode>): Use it.
29780 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
29782         PR target/83330
29783         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
29784         frame if argument is passed on stack.
29786 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29788         PR target/82682
29789         * ree.c (combine_reaching_defs): Optimize also
29790         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
29791         reg2=any_extend(exp); reg1=reg2;, formatting fix.
29793 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29795         PR middle-end/83189
29796         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
29798 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29800         PR middle-end/83718
29801         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
29802         after they are computed.
29804 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
29806         PR tree-optimization/83695
29807         * gimple-loop-linterchange.cc
29808         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
29809         reset cached scev information after interchange.
29810         (pass_linterchange::execute): Remove call to scev_reset_htab.
29812 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29814         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
29815         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
29816         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
29817         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
29818         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
29819         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
29820         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
29821         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
29822         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
29823         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
29824         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
29825         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
29826         (V_lane_reg): Likewise.
29827         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
29828         New define_expand.
29829         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
29830         (vfmal_lane_low<mode>_intrinsic,
29831         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
29832         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
29833         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
29834         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
29835         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
29836         vfmsl_lane_high<mode>_intrinsic): New define_insns.
29838 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29840         * config/arm/arm-cpus.in (fp16fml): New feature.
29841         (ALL_SIMD): Add fp16fml.
29842         (armv8.2-a): Add fp16fml as an option.
29843         (armv8.3-a): Likewise.
29844         (armv8.4-a): Add fp16fml as part of fp16.
29845         * config/arm/arm.h (TARGET_FP16FML): Define.
29846         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
29847         when appropriate.
29848         * config/arm/arm-modes.def (V2HF): Define.
29849         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
29850         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
29851         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
29852         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
29853         vfmsl_low, vfmsl_high): New set of builtins.
29854         * config/arm/iterators.md (PLUSMINUS): New code iterator.
29855         (vfml_op): New code attribute.
29856         (VFMLHALVES): New int iterator.
29857         (VFML, VFMLSEL): New mode attributes.
29858         (V_reg): Define mapping for V2HF.
29859         (V_hi, V_lo): New mode attributes.
29860         (VF_constraint): Likewise.
29861         (vfml_half, vfml_half_selector): New int attributes.
29862         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
29863         define_expand.
29864         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
29865         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
29866         New define_insn.
29867         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
29868         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
29869         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
29870         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
29871         documentation.
29872         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
29873         Document new effective target and option set.
29875 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29877         * config/arm/arm-cpus.in (armv8_4): New feature.
29878         (ARMv8_4a): New fgroup.
29879         (armv8.4-a): New arch.
29880         * config/arm/arm-tables.opt: Regenerate.
29881         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
29882         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
29883         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
29884         Add matching rules for -march=armv8.4-a and extensions.
29885         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
29887 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
29889         PR target/81821
29890         * config/rx/rx.md (BW): New mode attribute.
29891         (sync_lock_test_and_setsi): Add mode suffix to insn output.
29893 2018-01-11  Richard Biener  <rguenther@suse.de>
29895         PR tree-optimization/83435
29896         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
29897         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
29898         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
29900 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29901             Alan Hayward  <alan.hayward@arm.com>
29902             David Sherwood  <david.sherwood@arm.com>
29904         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
29905         field.
29906         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
29907         (aarch64_print_address_internal): Use it to check for a zero offset.
29909 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29910             Alan Hayward  <alan.hayward@arm.com>
29911             David Sherwood  <david.sherwood@arm.com>
29913         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
29914         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
29915         Return a poly_int64 rather than a HOST_WIDE_INT.
29916         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
29917         rather than a HOST_WIDE_INT.
29918         * config/aarch64/aarch64.h (aarch64_frame): Protect with
29919         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
29920         hard_fp_offset, frame_size, initial_adjust, callee_offset and
29921         final_offset from HOST_WIDE_INT to poly_int64.
29922         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
29923         to_constant when getting the number of units in an Advanced SIMD
29924         mode.
29925         (aarch64_builtin_vectorized_function): Check for a constant number
29926         of units.
29927         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
29928         GET_MODE_SIZE.
29929         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
29930         attribute instead of GET_MODE_NUNITS.
29931         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
29932         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
29933         GET_MODE_SIZE for fixed-size registers.
29934         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
29935         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
29936         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
29937         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
29938         (aarch64_print_operand, aarch64_print_address_internal)
29939         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
29940         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
29941         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
29942         Handle polynomial GET_MODE_SIZE.
29943         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
29944         wider than SImode without modification.
29945         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
29946         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
29947         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
29948         passing and returning SVE modes.
29949         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
29950         rather than GEN_INT.
29951         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
29952         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
29953         (aarch64_allocate_and_probe_stack_space): Likewise.
29954         (aarch64_layout_frame): Cope with polynomial offsets.
29955         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
29956         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
29957         polynomial offsets.
29958         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
29959         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
29960         poly_int64 rather than a HOST_WIDE_INT.
29961         (aarch64_get_separate_components, aarch64_process_components)
29962         (aarch64_expand_prologue, aarch64_expand_epilogue)
29963         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
29964         (aarch64_anchor_offset): New function, split out from...
29965         (aarch64_legitimize_address): ...here.
29966         (aarch64_builtin_vectorization_cost): Handle polynomial
29967         TYPE_VECTOR_SUBPARTS.
29968         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
29969         GET_MODE_NUNITS.
29970         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
29971         number of elements from the PARALLEL rather than the mode.
29972         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
29973         rather than GET_MODE_BITSIZE.
29974         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
29975         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
29976         (aarch64_expand_vec_perm_const_1): Handle polynomial
29977         d->perm.length () and d->perm elements.
29978         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
29979         Apply to_constant to d->perm elements.
29980         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
29981         polynomial CONST_VECTOR_NUNITS.
29982         (aarch64_move_pointer): Take amount as a poly_int64 rather
29983         than an int.
29984         (aarch64_progress_pointer): Avoid temporary variable.
29985         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
29986         the mode attribute instead of GET_MODE.
29988 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29989             Alan Hayward  <alan.hayward@arm.com>
29990             David Sherwood  <david.sherwood@arm.com>
29992         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
29993         x exists before using it.
29994         (aarch64_add_constant_internal): Rename to...
29995         (aarch64_add_offset_1): ...this.  Replace regnum with separate
29996         src and dest rtxes.  Handle the case in which they're different,
29997         including when the offset is zero.  Replace scratchreg with an rtx.
29998         Use 2 additions if there is no spare register into which we can
29999         move a 16-bit constant.
30000         (aarch64_add_constant): Delete.
30001         (aarch64_add_offset): Replace reg with separate src and dest
30002         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
30003         Use aarch64_add_offset_1.
30004         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
30005         an rtx rather than an int.  Take the delta as a poly_int64
30006         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
30007         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
30008         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
30009         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
30010         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
30011         and aarch64_add_sp.
30012         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
30013         aarch64_add_constant.
30015 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
30017         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
30018         Use scalar_float_mode.
30020 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
30022         * config/aarch64/aarch64-simd.md
30023         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
30024         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
30025         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
30026         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
30027         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
30028         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
30029         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
30030         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
30031         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
30032         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
30034 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
30036         PR target/83514
30037         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
30038         targ_options->x_arm_arch_string is non NULL.
30040 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
30042         * config/aarch64/aarch64.h
30043         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
30045 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
30047         PR target/82096
30048         * expmed.c (emit_store_flag_force): Swap if const op0
30049         and change VOIDmode to mode of op0.
30051 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
30053         PR rtl-optimization/83761
30054         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
30055         than bytes to mode_for_size.
30057 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30059         PR middle-end/83189
30060         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
30061         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
30062         profile.
30064 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30066         PR middle-end/83575
30067         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
30068         when in layout mode.
30069         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
30070         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
30071         partition fixup.
30073 2018-01-10  Michael Collison  <michael.collison@arm.com>
30075         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
30076         * config/aarch64/aarch64-option-extension.def: Add
30077         AARCH64_OPT_EXTENSION of 'fp16fml'.
30078         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30079         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
30080         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
30081         * config/aarch64/constraints.md (Ui7): New constraint.
30082         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
30083         (VFMLA_SEL_W): Ditto.
30084         (f16quad): Ditto.
30085         (f16mac1): Ditto.
30086         (VFMLA16_LOW): New int iterator.
30087         (VFMLA16_HIGH): Ditto.
30088         (UNSPEC_FMLAL): New unspec.
30089         (UNSPEC_FMLSL): Ditto.
30090         (UNSPEC_FMLAL2): Ditto.
30091         (UNSPEC_FMLSL2): Ditto.
30092         (f16mac): New code attribute.
30093         * config/aarch64/aarch64-simd-builtins.def
30094         (aarch64_fmlal_lowv2sf): Ditto.
30095         (aarch64_fmlsl_lowv2sf): Ditto.
30096         (aarch64_fmlalq_lowv4sf): Ditto.
30097         (aarch64_fmlslq_lowv4sf): Ditto.
30098         (aarch64_fmlal_highv2sf): Ditto.
30099         (aarch64_fmlsl_highv2sf): Ditto.
30100         (aarch64_fmlalq_highv4sf): Ditto.
30101         (aarch64_fmlslq_highv4sf): Ditto.
30102         (aarch64_fmlal_lane_lowv2sf): Ditto.
30103         (aarch64_fmlsl_lane_lowv2sf): Ditto.
30104         (aarch64_fmlal_laneq_lowv2sf): Ditto.
30105         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
30106         (aarch64_fmlalq_lane_lowv4sf): Ditto.
30107         (aarch64_fmlsl_lane_lowv4sf): Ditto.
30108         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
30109         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
30110         (aarch64_fmlal_lane_highv2sf): Ditto.
30111         (aarch64_fmlsl_lane_highv2sf): Ditto.
30112         (aarch64_fmlal_laneq_highv2sf): Ditto.
30113         (aarch64_fmlsl_laneq_highv2sf): Ditto.
30114         (aarch64_fmlalq_lane_highv4sf): Ditto.
30115         (aarch64_fmlsl_lane_highv4sf): Ditto.
30116         (aarch64_fmlalq_laneq_highv4sf): Ditto.
30117         (aarch64_fmlsl_laneq_highv4sf): Ditto.
30118         * config/aarch64/aarch64-simd.md:
30119         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
30120         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
30121         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
30122         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
30123         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
30124         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
30125         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
30126         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
30127         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
30128         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
30129         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
30130         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
30131         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
30132         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
30133         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
30134         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
30135         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
30136         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
30137         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
30138         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
30139         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
30140         (vfmlsl_low_u32): Ditto.
30141         (vfmlalq_low_u32): Ditto.
30142         (vfmlslq_low_u32): Ditto.
30143         (vfmlal_high_u32): Ditto.
30144         (vfmlsl_high_u32): Ditto.
30145         (vfmlalq_high_u32): Ditto.
30146         (vfmlslq_high_u32): Ditto.
30147         (vfmlal_lane_low_u32): Ditto.
30148         (vfmlsl_lane_low_u32): Ditto.
30149         (vfmlal_laneq_low_u32): Ditto.
30150         (vfmlsl_laneq_low_u32): Ditto.
30151         (vfmlalq_lane_low_u32): Ditto.
30152         (vfmlslq_lane_low_u32): Ditto.
30153         (vfmlalq_laneq_low_u32): Ditto.
30154         (vfmlslq_laneq_low_u32): Ditto.
30155         (vfmlal_lane_high_u32): Ditto.
30156         (vfmlsl_lane_high_u32): Ditto.
30157         (vfmlal_laneq_high_u32): Ditto.
30158         (vfmlsl_laneq_high_u32): Ditto.
30159         (vfmlalq_lane_high_u32): Ditto.
30160         (vfmlslq_lane_high_u32): Ditto.
30161         (vfmlalq_laneq_high_u32): Ditto.
30162         (vfmlslq_laneq_high_u32): Ditto.
30163         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
30164         (AARCH64_FL_FOR_ARCH8_4): New.
30165         (AARCH64_ISA_F16FML): New ISA flag.
30166         (TARGET_F16FML): New feature flag for fp16fml.
30167         (doc/invoke.texi): Document new fp16fml option.
30169 2018-01-10  Michael Collison  <michael.collison@arm.com>
30171         * config/aarch64/aarch64-builtins.c:
30172         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
30173         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30174         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
30175         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
30176         (AARCH64_ISA_SHA3): New ISA flag.
30177         (TARGET_SHA3): New feature flag for sha3.
30178         * config/aarch64/iterators.md (sha512_op): New int attribute.
30179         (CRYPTO_SHA512): New int iterator.
30180         (UNSPEC_SHA512H): New unspec.
30181         (UNSPEC_SHA512H2): Ditto.
30182         (UNSPEC_SHA512SU0): Ditto.
30183         (UNSPEC_SHA512SU1): Ditto.
30184         * config/aarch64/aarch64-simd-builtins.def
30185         (aarch64_crypto_sha512hqv2di): New builtin.
30186         (aarch64_crypto_sha512h2qv2di): Ditto.
30187         (aarch64_crypto_sha512su0qv2di): Ditto.
30188         (aarch64_crypto_sha512su1qv2di): Ditto.
30189         (aarch64_eor3qv8hi): Ditto.
30190         (aarch64_rax1qv2di): Ditto.
30191         (aarch64_xarqv2di): Ditto.
30192         (aarch64_bcaxqv8hi): Ditto.
30193         * config/aarch64/aarch64-simd.md:
30194         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
30195         (aarch64_crypto_sha512su0qv2di): Ditto.
30196         (aarch64_crypto_sha512su1qv2di): Ditto.
30197         (aarch64_eor3qv8hi): Ditto.
30198         (aarch64_rax1qv2di): Ditto.
30199         (aarch64_xarqv2di): Ditto.
30200         (aarch64_bcaxqv8hi): Ditto.
30201         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
30202         (vsha512h2q_u64): Ditto.
30203         (vsha512su0q_u64): Ditto.
30204         (vsha512su1q_u64): Ditto.
30205         (veor3q_u16): Ditto.
30206         (vrax1q_u64): Ditto.
30207         (vxarq_u64): Ditto.
30208         (vbcaxq_u16): Ditto.
30209         * config/arm/types.md (crypto_sha512): New type attribute.
30210         (crypto_sha3): Ditto.
30211         (doc/invoke.texi): Document new sha3 option.
30213 2018-01-10  Michael Collison  <michael.collison@arm.com>
30215         * config/aarch64/aarch64-builtins.c:
30216         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
30217         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30218         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
30219         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
30220         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
30221         (AARCH64_ISA_SM4): New ISA flag.
30222         (TARGET_SM4): New feature flag for sm4.
30223         * config/aarch64/aarch64-simd-builtins.def
30224         (aarch64_sm3ss1qv4si): Ditto.
30225         (aarch64_sm3tt1aq4si): Ditto.
30226         (aarch64_sm3tt1bq4si): Ditto.
30227         (aarch64_sm3tt2aq4si): Ditto.
30228         (aarch64_sm3tt2bq4si): Ditto.
30229         (aarch64_sm3partw1qv4si): Ditto.
30230         (aarch64_sm3partw2qv4si): Ditto.
30231         (aarch64_sm4eqv4si): Ditto.
30232         (aarch64_sm4ekeyqv4si): Ditto.
30233         * config/aarch64/aarch64-simd.md:
30234         (aarch64_sm3ss1qv4si): Ditto.
30235         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
30236         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
30237         (aarch64_sm4eqv4si): Ditto.
30238         (aarch64_sm4ekeyqv4si): Ditto.
30239         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
30240         (sm3part_op): Ditto.
30241         (CRYPTO_SM3TT): Ditto.
30242         (CRYPTO_SM3PART): Ditto.
30243         (UNSPEC_SM3SS1): New unspec.
30244         (UNSPEC_SM3TT1A): Ditto.
30245         (UNSPEC_SM3TT1B): Ditto.
30246         (UNSPEC_SM3TT2A): Ditto.
30247         (UNSPEC_SM3TT2B): Ditto.
30248         (UNSPEC_SM3PARTW1): Ditto.
30249         (UNSPEC_SM3PARTW2): Ditto.
30250         (UNSPEC_SM4E): Ditto.
30251         (UNSPEC_SM4EKEY): Ditto.
30252         * config/aarch64/constraints.md (Ui2): New constraint.
30253         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
30254         * config/arm/types.md (crypto_sm3): New type attribute.
30255         (crypto_sm4): Ditto.
30256         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
30257         (vsm3tt1aq_u32): Ditto.
30258         (vsm3tt1bq_u32): Ditto.
30259         (vsm3tt2aq_u32): Ditto.
30260         (vsm3tt2bq_u32): Ditto.
30261         (vsm3partw1q_u32): Ditto.
30262         (vsm3partw2q_u32): Ditto.
30263         (vsm4eq_u32): Ditto.
30264         (vsm4ekeyq_u32): Ditto.
30265         (doc/invoke.texi): Document new sm4 option.
30267 2018-01-10  Michael Collison  <michael.collison@arm.com>
30269         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
30270         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
30271         (AARCH64_FL_FOR_ARCH8_4): New.
30272         (AARCH64_FL_V8_4): New flag.
30273         (doc/invoke.texi): Document new armv8.4-a option.
30275 2018-01-10  Michael Collison  <michael.collison@arm.com>
30277         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30278         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
30279         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
30280         * config/aarch64/aarch64-option-extension.def: Add
30281         AARCH64_OPT_EXTENSION of 'sha2'.
30282         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
30283         (crypto): Disable sha2 and aes if crypto disabled.
30284         (crypto): Enable aes and sha2 if enabled.
30285         (simd): Disable sha2 and aes if simd disabled.
30286         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
30287         New flags.
30288         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
30289         (TARGET_SHA2): New feature flag for sha2.
30290         (TARGET_AES): New feature flag for aes.
30291         * config/aarch64/aarch64-simd.md:
30292         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
30293         conditional on TARGET_AES.
30294         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
30295         (aarch64_crypto_sha1hsi): Make pattern conditional
30296         on TARGET_SHA2.
30297         (aarch64_crypto_sha1hv4si): Ditto.
30298         (aarch64_be_crypto_sha1hv4si): Ditto.
30299         (aarch64_crypto_sha1su1v4si): Ditto.
30300         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
30301         (aarch64_crypto_sha1su0v4si): Ditto.
30302         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
30303         (aarch64_crypto_sha256su0v4si): Ditto.
30304         (aarch64_crypto_sha256su1v4si): Ditto.
30305         (doc/invoke.texi): Document new aes and sha2 options.
30307 2018-01-10  Martin Sebor  <msebor@redhat.com>
30309         PR tree-optimization/83781
30310         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
30311         as string arrays.
30313 2018-01-11  Martin Sebor  <msebor@gmail.com>
30314             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
30316         PR tree-optimization/83501
30317         PR tree-optimization/81703
30319         * tree-ssa-strlen.c (get_string_cst): Rename...
30320         (get_string_len): ...to this.  Handle global constants.
30321         (handle_char_store): Adjust.
30323 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
30324             Jim Wilson  <jimw@sifive.com>
30326         * config/riscv/riscv-protos.h (riscv_output_return): New.
30327         * config/riscv/riscv.c (struct machine_function): New naked_p field.
30328         (riscv_attribute_table, riscv_output_return),
30329         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
30330         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
30331         (riscv_compute_frame_info): Only compute frame->mask if not a naked
30332         function.
30333         (riscv_expand_prologue): Add early return for naked function.
30334         (riscv_expand_epilogue): Likewise.
30335         (riscv_function_ok_for_sibcall): Return false for naked function.
30336         (riscv_set_current_function): New.
30337         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
30338         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
30339         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
30340         * doc/extend.texi (RISC-V Function Attributes): New.
30342 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
30344         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
30345         check for 128-bit long double before checking TCmode.
30346         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
30347         128-bit long doubles before checking TFmode or TCmode.
30348         (FLOAT128_IBM_P): Likewise.
30350 2018-01-10  Martin Sebor  <msebor@redhat.com>
30352         PR tree-optimization/83671
30353         * builtins.c (c_strlen): Unconditionally return zero for the empty
30354         string.
30355         Use -Warray-bounds for warnings.
30356         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
30357         for non-constant array indices with COMPONENT_REF, arrays of
30358         arrays, and pointers to arrays.
30359         (gimple_fold_builtin_strlen): Determine and set length range for
30360         non-constant character arrays.
30362 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
30364         PR middle-end/81897
30365         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
30366         empty blocks.
30368 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
30370         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
30372 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
30374         PR target/83399
30375         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
30376         VECTOR_MEM_ALTIVEC_OR_VSX_P.
30377         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
30378         indexed_or_indirect_operand predicate.
30379         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
30380         (*vsx_le_perm_load_v8hi): Likewise.
30381         (*vsx_le_perm_load_v16qi): Likewise.
30382         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
30383         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
30384         (*vsx_le_perm_store_v8hi): Likewise.
30385         (*vsx_le_perm_store_v16qi): Likewise.
30386         (eight unnamed splitters): Likewise.
30388 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
30390         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
30391         * config/rs6000/emmintrin.h: Likewise.
30392         * config/rs6000/mmintrin.h: Likewise.
30393         * config/rs6000/xmmintrin.h: Likewise.
30395 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
30397         PR c++/43486
30398         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
30399         "public_flag".
30400         * tree.c (tree_nop_conversion): Return true for location wrapper
30401         nodes.
30402         (maybe_wrap_with_location): New function.
30403         (selftest::check_strip_nops): New function.
30404         (selftest::test_location_wrappers): New function.
30405         (selftest::tree_c_tests): Call it.
30406         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
30407         (maybe_wrap_with_location): New decl.
30408         (EXPR_LOCATION_WRAPPER_P): New macro.
30409         (location_wrapper_p): New inline function.
30410         (tree_strip_any_location_wrapper): New inline function.
30412 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
30414         PR target/83735
30415         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
30416         stack_realign_offset for the largest alignment of stack slot
30417         actually used.
30418         (ix86_find_max_used_stack_alignment): New function.
30419         (ix86_finalize_stack_frame_flags): Use it.  Set
30420         max_used_stack_alignment if we don't realign stack.
30421         * config/i386/i386.h (machine_function): Add
30422         max_used_stack_alignment.
30424 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
30426         * config/arm/arm.opt (-mbranch-cost): New option.
30427         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
30428         account.
30430 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
30432         PR target/83629
30433         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
30434         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
30436 2018-01-10  Richard Biener  <rguenther@suse.de>
30438         PR debug/83765
30439         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
30440         early out so it also covers the case where we have a non-NULL
30441         origin.
30443 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
30445         PR tree-optimization/83753
30446         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
30447         for non-strided grouped accesses if the number of elements is 1.
30449 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30451         PR target/81616
30452         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
30453         * i386.h (TARGET_USE_GATHER): Define.
30454         * x86-tune.def (X86_TUNE_USE_GATHER): New.
30456 2018-01-10  Martin Liska  <mliska@suse.cz>
30458         PR bootstrap/82831
30459         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
30460         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
30461         partitioning.
30462         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
30463         CLEANUP_NO_PARTITIONING is not set.
30465 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
30467         * doc/rtl.texi: Remove documentation of (const ...) wrappers
30468         for vectors, as a partial revert of r254296.
30469         * rtl.h (const_vec_p): Delete.
30470         (const_vec_duplicate_p): Don't test for vector CONSTs.
30471         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
30472         * expmed.c (make_tree): Likewise.
30474         Revert:
30475         * common.md (E, F): Use CONSTANT_P instead of checking for
30476         CONST_VECTOR.
30477         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
30478         checking for CONST_VECTOR.
30480 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30482         PR middle-end/83575
30483         * predict.c (force_edge_cold): Handle in more sane way edges
30484         with no prediction.
30486 2018-01-09  Carl Love  <cel@us.ibm.com>
30488         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
30489         V4SI, V4SF types.
30490         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
30491         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
30492         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
30493         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
30494         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
30495         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
30496         * config/rs6000/rs6000-protos.h: Add extern defition for
30497         rs6000_generate_float2_double_code.
30498         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
30499         function.
30500         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
30501         (float2_v2df): Add define_expand.
30503 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
30505         PR target/83628
30506         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
30507         op_mode in the force_to_mode call.
30509 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30511         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
30512         instead of checking each element individually.
30513         (aarch64_evpc_uzp): Likewise.
30514         (aarch64_evpc_zip): Likewise.
30515         (aarch64_evpc_ext): Likewise.
30516         (aarch64_evpc_rev): Likewise.
30517         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
30518         instead of checking each element individually.  Return true without
30519         generating rtl if
30520         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
30521         whether all selected elements come from the same input, instead of
30522         checking each element individually.  Remove calls to gen_rtx_REG,
30523         start_sequence and end_sequence and instead assert that no rtl is
30524         generated.
30526 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30528         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
30529         order of HIGH and CONST checks.
30531 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30533         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
30534         if the destination isn't an SSA_NAME.
30536 2018-01-09  Richard Biener  <rguenther@suse.de>
30538         PR tree-optimization/83668
30539         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
30540         move prologue...
30541         (canonicalize_loop_form): ... here, renamed from ...
30542         (canonicalize_loop_closed_ssa_form): ... this and amended to
30543         swap successor edges for loop exit blocks to make us use
30544         the RPO order we need for initial schedule generation.
30546 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
30548         PR tree-optimization/64811
30549         * match.pd: When optimizing comparisons with Inf, avoid
30550         introducing or losing exceptions from comparisons with NaN.
30552 2018-01-09  Martin Liska  <mliska@suse.cz>
30554         PR sanitizer/82517
30555         * asan.c (shadow_mem_size): Add gcc_assert.
30557 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
30559         Don't save registers in main().
30561         PR target/83738
30562         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
30563         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
30564         * config/avr/avr.c (avr_set_current_function): Don't error if
30565         naked, OS_task or OS_main are specified at the same time.
30566         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
30567         OS_main.
30568         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
30569         attribute.
30570         * common/config/avr/avr-common.c (avr_option_optimization_table):
30571         Switch on -mmain-is-OS_task for optimizing compilations.
30573 2018-01-09  Richard Biener  <rguenther@suse.de>
30575         PR tree-optimization/83572
30576         * graphite.c: Include cfganal.h.
30577         (graphite_transform_loops): Connect infinite loops to exit
30578         and remove fake edges at the end.
30580 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30582         * ipa-inline.c (edge_badness): Revert accidental checkin.
30584 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30586         PR ipa/80763
30587         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
30588         symbols; not inline clones.
30590 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
30592         PR target/83507
30593         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
30594         hard registers.  Formatting fixes.
30596         PR preprocessor/83722
30597         * gcc.c (try_generate_repro): Pass
30598         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
30599         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
30600         do_report_bug.
30602 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
30603             Kito Cheng  <kito.cheng@gmail.com>
30605         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
30606         (riscv_leaf_function_p): Delete.
30607         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
30609 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30611         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
30612         function.
30613         (do_ifelse): New function.
30614         (do_isel): New function.
30615         (do_sub3): New function.
30616         (do_add3): New function.
30617         (do_load_mask_compare): New function.
30618         (do_overlap_load_compare): New function.
30619         (expand_compare_loop): New function.
30620         (expand_block_compare): Call expand_compare_loop() when appropriate.
30621         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
30622         option description.
30623         (-mblock-compare-inline-loop-limit): New option.
30625 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30627         PR target/83677
30628         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
30629         Reverse order of second and third operands in first alternative.
30630         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
30631         of first and second elements in UNSPEC_VPERMR vector.
30632         (altivec_expand_vec_perm_le): Likewise.
30634 2018-01-08  Jeff Law  <law@redhat.com>
30636         PR rtl-optimizatin/81308
30637         * tree-switch-conversion.c (cfg_altered): New file scoped static.
30638         (process_switch): If group_case_labels makes a change, then set
30639         cfg_altered.
30640         (pass_convert_switch::execute): If a switch is converted, then
30641         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
30643         PR rtl-optimization/81308
30644         * recog.c (split_all_insns): Conditionally cleanup the CFG after
30645         splitting insns.
30647 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
30649         PR target/83663 - Revert r255946
30650         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
30651         generation for cases where splatting a value is not useful.
30652         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
30653         across a vec_duplicate and a paradoxical subreg forming a vector
30654         mode to a vec_concat.
30656 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30658         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
30659         -march=armv8.3-a variants.
30660         * config/arm/t-multilib: Likewise.
30661         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
30663 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30665         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
30666         to generate rtl.
30667         (cceq_ior_compare_complement): Give it a name so I can use it, and
30668         change boolean_or_operator predicate to boolean_operator so it can
30669         be used to generate a crand.
30670         (eqne): New code iterator.
30671         (bd/bd_neg): New code_attrs.
30672         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
30673         a single define_insn.
30674         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
30675         decrement (bdnzt/bdnzf/bdzt/bdzf).
30676         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
30677         with the new names of the branch decrement patterns, and added the
30678         names of the branch decrement conditional patterns.
30680 2018-01-08  Richard Biener  <rguenther@suse.de>
30682         PR tree-optimization/83563
30683         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
30684         cache.
30686 2018-01-08  Richard Biener  <rguenther@suse.de>
30688         PR middle-end/83713
30689         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
30691 2018-01-08  Richard Biener  <rguenther@suse.de>
30693         PR tree-optimization/83685
30694         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
30695         references to abnormals.
30697 2018-01-08  Richard Biener  <rguenther@suse.de>
30699         PR lto/83719
30700         * dwarf2out.c (output_indirect_strings): Handle empty
30701         skeleton_debug_str_hash.
30702         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
30704 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30706         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
30707         (emit_store_direct): Likewise.
30708         (arc_trampoline_adjust_address): Likewise.
30709         (arc_asm_trampoline_template): New function.
30710         (arc_initialize_trampoline): Use asm_trampoline_template.
30711         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
30712         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
30713         * config/arc/arc.md (flush_icache): Delete pattern.
30715 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30717         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
30718         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
30719         munaligned-access.
30721 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30723         PR target/83681
30724         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
30725         by not USED_FOR_TARGET.
30726         (make_pass_resolve_sw_modes): Likewise.
30728 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30730         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
30731         USED_FOR_TARGET.
30733 2018-01-08  Richard Biener  <rguenther@suse.de>
30735         PR middle-end/83580
30736         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
30738 2018-01-08  Richard Biener  <rguenther@suse.de>
30740         PR middle-end/83517
30741         * match.pd ((t * 2) / 2) -> t): Add missing :c.
30743 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
30745         PR middle-end/81897
30746         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
30747         basic blocks with a small number of successors.
30748         (convert_control_dep_chain_into_preds): Improve handling of
30749         forwarder blocks.
30750         (dump_predicates): Split apart into...
30751         (dump_pred_chain): ...here...
30752         (dump_pred_info): ...and here.
30753         (can_one_predicate_be_invalidated_p): Add debugging printfs.
30754         (can_chain_union_be_invalidated_p): Improve check for invalidation
30755         of paths.
30756         (uninit_uses_cannot_happen): Avoid unnecessary if
30757         convert_control_dep_chain_into_preds yielded nothing.
30759 2018-01-06  Martin Sebor  <msebor@redhat.com>
30761         PR tree-optimization/83640
30762         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
30763         subtracting negative offset from size.
30764         (builtin_access::overlap): Adjust offset bounds of the access to fall
30765         within the size of the object if possible.
30767 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
30769         PR rtl-optimization/83699
30770         * expmed.c (extract_bit_field_1): Restrict the vector usage of
30771         extract_bit_field_as_subreg to cases in which the extracted
30772         value is also a vector.
30774         * lra-constraints.c (process_alt_operands): Test for the equivalence
30775         substitutions when detecting a possible reload cycle.
30777 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
30779         PR debug/83480
30780         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
30781         by default if flag_selective_schedling{,2}.  Formatting fixes.
30783         PR rtl-optimization/83682
30784         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
30785         if it has non-VECTOR_MODE element mode.
30786         (vec_duplicate_p): Likewise.
30788         PR middle-end/83694
30789         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
30790         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
30792 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30794         PR target/83604
30795         * config/i386/i386-builtin.def
30796         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
30797         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
30798         Require also OPTION_MASK_ISA_AVX512F in addition to
30799         OPTION_MASK_ISA_GFNI.
30800         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
30801         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
30802         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
30803         to OPTION_MASK_ISA_GFNI.
30804         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
30805         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
30806         OPTION_MASK_ISA_AVX512BW.
30807         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
30808         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
30809         addition to OPTION_MASK_ISA_GFNI.
30810         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
30811         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
30812         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
30813         to OPTION_MASK_ISA_GFNI.
30814         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
30815         a requirement for all ISAs rather than any of them with a few
30816         exceptions.
30817         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
30818         processing.
30819         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
30820         bitmasks to be enabled with 3 exceptions, instead of requiring any
30821         enabled ISA with lots of exceptions.
30822         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
30823         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
30824         Change avx512bw in isa attribute to avx512f.
30825         * config/i386/sgxintrin.h: Add license boilerplate.
30826         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
30827         to __AVX512F__ and __AVX512VL to __AVX512VL__.
30828         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
30829         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
30830         defined.
30831         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
30832         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
30833         temporarily sse2 rather than sse if not enabled already.
30835         PR target/83604
30836         * config/i386/sse.md (VI248_VLBW): Rename to ...
30837         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
30838         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
30839         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
30840         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
30841         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
30842         mode iterator instead of VI248_VLBW.
30844 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
30846         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
30847         (record_modified): Skip clobbers; add debug output.
30848         (param_change_prob): Use sreal frequencies.
30850 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30852         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
30853         punt for user-aligned variables.
30855 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30857         * tree-chrec.c (chrec_contains_symbols): Return true for
30858         POLY_INT_CST.
30860 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
30862         PR target/82439
30863         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
30864         of (x|y) == x for BICS pattern.
30866 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30868         PR tree-optimization/83605
30869         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
30870         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
30871         can throw.
30873 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30875         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
30876         * config/epiphany/rtems.h: New file.
30878 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30879             Uros Bizjak  <ubizjak@gmail.com>
30881         PR target/83554
30882         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
30883         QIreg_operand instead of register_operand predicate.
30884         * config/i386/i386.c (ix86_rop_should_change_byte_p,
30885         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
30886         comments instead of -fmitigate[-_]rop.
30888 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30890         PR bootstrap/81926
30891         * cgraphunit.c (symbol_table::compile): Switch to text_section
30892         before calling assembly_start debug hook.
30893         * run-rtl-passes.c (run_rtl_passes): Likewise.
30894         Include output.h.
30896 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30898         * tree-vrp.c (extract_range_from_binary_expr_1): Check
30899         range_int_cst_p rather than !symbolic_range_p before calling
30900         extract_range_from_multiplicative_op_1.
30902 2018-01-04  Jeff Law  <law@redhat.com>
30904         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
30905         redundant test in assertion.
30907 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30909         * doc/rtl.texi: Document machine_mode wrapper classes.
30911 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30913         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
30914         using tree_to_uhwi.
30916 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30918         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
30919         the VEC_PERM_EXPR fold to fail.
30921 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30923         PR debug/83585
30924         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
30925         to switched_sections.
30927 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30929         PR target/83680
30930         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
30931         test for d.testing.
30933 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
30935         PR target/83387
30936         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
30937         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
30939 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30941         PR debug/83666
30942         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
30943         is BLKmode and bitpos not zero or mode change is needed.
30945 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30947         PR target/83675
30948         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
30949         TARGET_VIS2.
30951 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
30953         PR target/83628
30954         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
30955         instead of MULT rtx.  Update all corresponding splitters.
30956         (*saddl_se): Ditto.
30957         (*ssub<modesuffix>): Ditto.
30958         (*ssubl_se): Ditto.
30959         (*cmp_sadd_di): Update split patterns.
30960         (*cmp_sadd_si): Ditto.
30961         (*cmp_sadd_sidi): Ditto.
30962         (*cmp_ssub_di): Ditto.
30963         (*cmp_ssub_si): Ditto.
30964         (*cmp_ssub_sidi): Ditto.
30965         * config/alpha/predicates.md (const23_operand): New predicate.
30966         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
30967         Look for ASHIFT, not MULT inner operand.
30968         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
30970 2018-01-04  Martin Liska  <mliska@suse.cz>
30972         PR gcov-profile/83669
30973         * gcov.c (output_intermediate_file): Add version to intermediate
30974         gcov file.
30975         * doc/gcov.texi: Document new field 'version' in intermediate
30976         file format. Fix location of '-k' option of gcov command.
30978 2018-01-04  Martin Liska  <mliska@suse.cz>
30980         PR ipa/82352
30981         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
30983 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30985         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
30987 2018-01-03  Martin Sebor  <msebor@redhat.com>
30989         PR tree-optimization/83655
30990         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
30991         checking calls with invalid arguments.
30993 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30995         * tree-vect-stmts.c (vect_get_store_rhs): New function.
30996         (vectorizable_mask_load_store): Delete.
30997         (vectorizable_call): Return false for masked loads and stores.
30998         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
30999         instead of gimple_assign_rhs1.
31000         (vectorizable_load): Handle IFN_MASK_LOAD.
31001         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
31003 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31005         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
31006         split out from..,
31007         (vectorizable_mask_load_store): ...here.
31008         (vectorizable_load): ...and here.
31010 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31012         * tree-vect-stmts.c (vect_build_all_ones_mask)
31013         (vect_build_zero_merge_argument): New functions, split out from...
31014         (vectorizable_load): ...here.
31016 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31018         * tree-vect-stmts.c (vect_check_store_rhs): New function,
31019         split out from...
31020         (vectorizable_mask_load_store): ...here.
31021         (vectorizable_store): ...and here.
31023 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31025         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
31026         split out from...
31027         (vectorizable_mask_load_store): ...here.
31029 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31031         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
31032         (vect_model_store_cost): Take a vec_load_store_type instead of a
31033         vect_def_type.
31034         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
31035         (vect_model_store_cost): Take a vec_load_store_type instead of a
31036         vect_def_type.
31037         (vectorizable_mask_load_store): Update accordingly.
31038         (vectorizable_store): Likewise.
31039         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
31041 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31043         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
31044         IFN_MASK_LOAD calls here rather than...
31045         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
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         * expmed.c (extract_bit_field_1): For vector extracts,
31052         fall back to extract_bit_field_as_subreg if vec_extract
31053         isn't available.
31055 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31056             Alan Hayward  <alan.hayward@arm.com>
31057             David Sherwood  <david.sherwood@arm.com>
31059         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
31060         they are variable or constant sized.
31061         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
31062         slots for constant-sized data.
31064 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31065             Alan Hayward  <alan.hayward@arm.com>
31066             David Sherwood  <david.sherwood@arm.com>
31068         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
31069         handling COND_EXPRs with boolean comparisons, try to find a better
31070         basis for the mask type than the boolean itself.
31072 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31074         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
31075         is calculated and how it can be overridden.
31076         * genmodes.c (max_bitsize_mode_any_mode): New variable.
31077         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
31078         if defined.
31079         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
31080         if nonzero.
31082 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31083             Alan Hayward  <alan.hayward@arm.com>
31084             David Sherwood  <david.sherwood@arm.com>
31086         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
31087         Remove the mode argument.
31088         (aarch64_simd_valid_immediate): Remove the mode and inverse
31089         arguments.
31090         * config/aarch64/iterators.md (bitsize): New iterator.
31091         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
31092         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
31093         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
31094         aarch64_simd_valid_immediate.
31095         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
31096         (aarch64_reg_or_bic_imm): Likewise.
31097         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
31098         with an insn_type enum and msl with a modifier_type enum.
31099         Replace element_width with a scalar_mode.  Change the shift
31100         to unsigned int.  Add constructors for scalar_float_mode and
31101         scalar_int_mode elements.
31102         (aarch64_vect_float_const_representable_p): Delete.
31103         (aarch64_can_const_movi_rtx_p)
31104         (aarch64_simd_scalar_immediate_valid_for_move)
31105         (aarch64_simd_make_constant): Update call to
31106         aarch64_simd_valid_immediate.
31107         (aarch64_advsimd_valid_immediate_hs): New function.
31108         (aarch64_advsimd_valid_immediate): Likewise.
31109         (aarch64_simd_valid_immediate): Remove mode and inverse
31110         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
31111         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
31112         and aarch64_float_const_representable_p on the result.
31113         (aarch64_output_simd_mov_immediate): Remove mode argument.
31114         Update call to aarch64_simd_valid_immediate and use of
31115         simd_immediate_info.
31116         (aarch64_output_scalar_simd_mov_immediate): Update call
31117         accordingly.
31119 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31120             Alan Hayward  <alan.hayward@arm.com>
31121             David Sherwood  <david.sherwood@arm.com>
31123         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
31124         (mode_nunits): Likewise CONST_MODE_NUNITS.
31125         * machmode.def (ADJUST_NUNITS): Document.
31126         * genmodes.c (mode_data::need_nunits_adj): New field.
31127         (blank_mode): Update accordingly.
31128         (adj_nunits): New variable.
31129         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
31130         parameter.
31131         (emit_mode_size_inline): Set need_bytesize_adj for all modes
31132         listed in adj_nunits.
31133         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
31134         listed in adj_nunits.  Don't emit case statements for such modes.
31135         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
31136         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
31137         nothing if adj_nunits is nonnull.
31138         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
31139         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
31140         (emit_mode_fbit): Update use of print_maybe_const_decl.
31141         (emit_move_size): Likewise.  Treat the array as non-const
31142         if adj_nunits.
31143         (emit_mode_adjustments): Handle adj_nunits.
31145 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31147         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
31148         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
31149         (VECTOR_MODES): Use it.
31150         (make_vector_modes): Take the prefix as an argument.
31152 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31153             Alan Hayward  <alan.hayward@arm.com>
31154             David Sherwood  <david.sherwood@arm.com>
31156         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
31157         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
31158         for MODE_VECTOR_BOOL.
31159         * machmode.def (VECTOR_BOOL_MODE): Document.
31160         * genmodes.c (VECTOR_BOOL_MODE): New macro.
31161         (make_vector_bool_mode): New function.
31162         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
31163         MODE_VECTOR_BOOL.
31164         * lto-streamer-in.c (lto_input_mode_table): Likewise.
31165         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
31166         Likewise.
31167         * stor-layout.c (int_mode_for_mode): Likewise.
31168         * tree.c (build_vector_type_for_mode): Likewise.
31169         * varasm.c (output_constant_pool_2): Likewise.
31170         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
31171         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
31172         for MODE_VECTOR_BOOL.
31173         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
31174         of mode class checks.
31175         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
31176         instead of a list of mode class checks.
31177         (expand_vector_scalar_condition): Likewise.
31178         (type_for_widest_vector_mode): Handle BImode as an inner mode.
31180 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31181             Alan Hayward  <alan.hayward@arm.com>
31182             David Sherwood  <david.sherwood@arm.com>
31184         * machmode.h (mode_size): Change from unsigned short to
31185         poly_uint16_pod.
31186         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
31187         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
31188         or if measurement_type is not polynomial.
31189         (fixed_size_mode::includes_p): Check for constant-sized modes.
31190         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
31191         return a poly_uint16 rather than an unsigned short.
31192         (emit_mode_size): Change the type of mode_size from unsigned short
31193         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
31194         (emit_mode_adjustments): Cope with polynomial vector sizes.
31195         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31196         for GET_MODE_SIZE.
31197         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31198         for GET_MODE_SIZE.
31199         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
31200         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
31201         * caller-save.c (setup_save_areas): Likewise.
31202         (replace_reg_with_saved_mem): Likewise.
31203         * calls.c (emit_library_call_value_1): Likewise.
31204         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
31205         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
31206         (gen_lowpart_for_combine): Likewise.
31207         * convert.c (convert_to_integer_1): Likewise.
31208         * cse.c (equiv_constant, cse_insn): Likewise.
31209         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
31210         (cselib_subst_to_values): Likewise.
31211         * dce.c (word_dce_process_block): Likewise.
31212         * df-problems.c (df_word_lr_mark_ref): Likewise.
31213         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
31214         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
31215         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
31216         (rtl_for_decl_location): Likewise.
31217         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
31218         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
31219         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
31220         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
31221         (expand_expr_real_1): Likewise.
31222         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
31223         (pad_below): Likewise.
31224         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31225         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
31226         * ira.c (get_subreg_tracking_sizes): Likewise.
31227         * ira-build.c (ira_create_allocno_objects): Likewise.
31228         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
31229         (ira_sort_regnos_for_alter_reg): Likewise.
31230         * ira-costs.c (record_operand_costs): Likewise.
31231         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
31232         (resolve_simple_move): Likewise.
31233         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
31234         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
31235         (lra_constraints): Likewise.
31236         (CONST_POOL_OK_P): Reject variable-sized modes.
31237         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
31238         (add_pseudo_to_slot, lra_spill): Likewise.
31239         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31240         * optabs-query.c (get_best_extraction_insn): Likewise.
31241         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31242         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
31243         (expand_mult_highpart, valid_multiword_target_p): Likewise.
31244         * recog.c (offsettable_address_addr_space_p): Likewise.
31245         * regcprop.c (maybe_mode_change): Likewise.
31246         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
31247         * regrename.c (build_def_use): Likewise.
31248         * regstat.c (dump_reg_info): Likewise.
31249         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
31250         (find_reloads, find_reloads_subreg_address): Likewise.
31251         * reload1.c (eliminate_regs_1): Likewise.
31252         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
31253         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
31254         (simplify_binary_operation_1, simplify_subreg): Likewise.
31255         * targhooks.c (default_function_arg_padding): Likewise.
31256         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
31257         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
31258         (verify_gimple_assign_ternary): Likewise.
31259         * tree-inline.c (estimate_move_cost): Likewise.
31260         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31261         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
31262         (get_address_cost_ainc): Likewise.
31263         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
31264         (vect_supportable_dr_alignment): Likewise.
31265         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31266         (vectorizable_reduction): Likewise.
31267         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
31268         (vectorizable_operation, vectorizable_load): Likewise.
31269         * tree.c (build_same_sized_truth_vector_type): Likewise.
31270         * valtrack.c (cleanup_auto_inc_dec): Likewise.
31271         * var-tracking.c (emit_note_insn_var_location): Likewise.
31272         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
31273         (ADDR_VEC_ALIGN): Likewise.
31275 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31276             Alan Hayward  <alan.hayward@arm.com>
31277             David Sherwood  <david.sherwood@arm.com>
31279         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
31280         unsigned short.
31281         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
31282         or if measurement_type is polynomial.
31283         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
31284         * combine.c (make_extraction): Likewise.
31285         * dse.c (find_shift_sequence): Likewise.
31286         * dwarf2out.c (mem_loc_descriptor): Likewise.
31287         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
31288         (extract_bit_field, extract_low_bits): Likewise.
31289         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
31290         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
31291         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
31292         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
31293         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31294         * reload.c (find_reloads): Likewise.
31295         * reload1.c (alter_reg): Likewise.
31296         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
31297         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
31298         * tree-if-conv.c (predicate_mem_writes): Likewise.
31299         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
31300         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
31301         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
31302         * valtrack.c (dead_debug_insert_temp): Likewise.
31303         * varasm.c (mergeable_constant_section): Likewise.
31304         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
31306 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31307             Alan Hayward  <alan.hayward@arm.com>
31308             David Sherwood  <david.sherwood@arm.com>
31310         * expr.c (expand_assignment): Cope with polynomial mode sizes
31311         when assigning to a CONCAT.
31313 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31314             Alan Hayward  <alan.hayward@arm.com>
31315             David Sherwood  <david.sherwood@arm.com>
31317         * machmode.h (mode_precision): Change from unsigned short to
31318         poly_uint16_pod.
31319         (mode_to_precision): Return a poly_uint16 rather than an unsigned
31320         short.
31321         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
31322         or if measurement_type is not polynomial.
31323         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
31324         in which the mode is already known to be a scalar_int_mode.
31325         * genmodes.c (emit_mode_precision): Change the type of mode_precision
31326         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
31327         initializer.
31328         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31329         for GET_MODE_PRECISION.
31330         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31331         for GET_MODE_PRECISION.
31332         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
31333         as polynomial.
31334         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
31335         (expand_field_assignment, make_extraction): Likewise.
31336         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
31337         (get_last_value): Likewise.
31338         * convert.c (convert_to_integer_1): Likewise.
31339         * cse.c (cse_insn): Likewise.
31340         * expr.c (expand_expr_real_1): Likewise.
31341         * lra-constraints.c (simplify_operand_subreg): Likewise.
31342         * optabs-query.c (can_atomic_load_p): Likewise.
31343         * optabs.c (expand_atomic_load): Likewise.
31344         (expand_atomic_store): Likewise.
31345         * ree.c (combine_reaching_defs): Likewise.
31346         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
31347         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
31348         * tree.h (type_has_mode_precision_p): Likewise.
31349         * ubsan.c (instrument_si_overflow): Likewise.
31351 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31352             Alan Hayward  <alan.hayward@arm.com>
31353             David Sherwood  <david.sherwood@arm.com>
31355         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
31356         polynomial numbers of units.
31357         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
31358         (valid_vector_subparts_p): New function.
31359         (build_vector_type): Remove temporary shim and take the number
31360         of units as a poly_uint64 rather than an int.
31361         (build_opaque_vector_type): Take the number of units as a
31362         poly_uint64 rather than an int.
31363         * tree.c (build_vector_from_ctor): Handle polynomial
31364         TYPE_VECTOR_SUBPARTS.
31365         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
31366         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
31367         (build_vector_from_val): If the number of units is variable,
31368         use build_vec_duplicate_cst for constant operands and
31369         VEC_DUPLICATE_EXPR otherwise.
31370         (make_vector_type): Remove temporary is_constant ().
31371         (build_vector_type, build_opaque_vector_type): Take the number of
31372         units as a poly_uint64 rather than an int.
31373         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
31374         VECTOR_CST_NELTS.
31375         * cfgexpand.c (expand_debug_expr): Likewise.
31376         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
31377         (store_constructor, expand_expr_real_1): Likewise.
31378         (const_scalar_mask_from_tree): Likewise.
31379         * fold-const-call.c (fold_const_reduction): Likewise.
31380         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
31381         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
31382         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
31383         (fold_relational_const): Likewise.
31384         (native_interpret_vector): Likewise.  Change the size from an
31385         int to an unsigned int.
31386         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
31387         TYPE_VECTOR_SUBPARTS.
31388         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
31389         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
31390         duplicating a non-constant operand into a variable-length vector.
31391         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
31392         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
31393         * ipa-icf.c (sem_variable::equals): Likewise.
31394         * match.pd: Likewise.
31395         * omp-simd-clone.c (simd_clone_subparts): Likewise.
31396         * print-tree.c (print_node): Likewise.
31397         * stor-layout.c (layout_type): Likewise.
31398         * targhooks.c (default_builtin_vectorization_cost): Likewise.
31399         * tree-cfg.c (verify_gimple_comparison): Likewise.
31400         (verify_gimple_assign_binary): Likewise.
31401         (verify_gimple_assign_ternary): Likewise.
31402         (verify_gimple_assign_single): Likewise.
31403         * tree-pretty-print.c (dump_generic_node): Likewise.
31404         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31405         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
31406         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
31407         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
31408         (vect_shift_permute_load_chain): Likewise.
31409         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
31410         (expand_vector_condition, optimize_vector_constructor): Likewise.
31411         (lower_vec_perm, get_compute_type): Likewise.
31412         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31413         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
31414         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
31415         (vect_recog_mask_conversion_pattern): Likewise.
31416         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
31417         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
31418         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31419         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
31420         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
31421         (vectorizable_shift, vectorizable_operation, vectorizable_store)
31422         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
31423         (supportable_widening_operation): Likewise.
31424         (supportable_narrowing_operation): Likewise.
31425         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
31426         Likewise.
31427         * varasm.c (output_constant): Likewise.
31429 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31430             Alan Hayward  <alan.hayward@arm.com>
31431             David Sherwood  <david.sherwood@arm.com>
31433         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
31434         so that both the length == 3 and length != 3 cases set up their
31435         own permute vectors.  Add comments explaining why we know the
31436         number of elements is constant.
31437         (vect_permute_load_chain): Likewise.
31439 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31440             Alan Hayward  <alan.hayward@arm.com>
31441             David Sherwood  <david.sherwood@arm.com>
31443         * machmode.h (mode_nunits): Change from unsigned char to
31444         poly_uint16_pod.
31445         (ONLY_FIXED_SIZE_MODES): New macro.
31446         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
31447         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
31448         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
31449         New typedefs.
31450         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
31451         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
31452         or if measurement_type is not polynomial.
31453         * genmodes.c (ZERO_COEFFS): New macro.
31454         (emit_mode_nunits_inline): Make mode_nunits_inline return a
31455         poly_uint16.
31456         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
31457         Use ZERO_COEFFS when emitting initializers.
31458         * data-streamer.h (bp_pack_poly_value): New function.
31459         (bp_unpack_poly_value): Likewise.
31460         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31461         for GET_MODE_NUNITS.
31462         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31463         for GET_MODE_NUNITS.
31464         * tree.c (make_vector_type): Remove temporary shim and make
31465         the real function take the number of units as a poly_uint64
31466         rather than an int.
31467         (build_vector_type_for_mode): Handle polynomial nunits.
31468         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
31469         * emit-rtl.c (const_vec_series_p_1): Likewise.
31470         (gen_rtx_CONST_VECTOR): Likewise.
31471         * fold-const.c (test_vec_duplicate_folding): Likewise.
31472         * genrecog.c (validate_pattern): Likewise.
31473         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
31474         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31475         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
31476         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
31477         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
31478         * rtlanal.c (subreg_get_info): Likewise.
31479         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31480         (vect_grouped_load_supported): Likewise.
31481         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
31482         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
31483         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
31484         (simplify_const_unary_operation, simplify_binary_operation_1)
31485         (simplify_const_binary_operation, simplify_ternary_operation)
31486         (test_vector_ops_duplicate, test_vector_ops): Likewise.
31487         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
31488         instead of CONST_VECTOR_NUNITS.
31489         * varasm.c (output_constant_pool_2): Likewise.
31490         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
31491         explicit-encoded elements in the XVEC for variable-length vectors.
31493 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31495         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
31497 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31498             Alan Hayward  <alan.hayward@arm.com>
31499             David Sherwood  <david.sherwood@arm.com>
31501         * coretypes.h (fixed_size_mode): Declare.
31502         (fixed_size_mode_pod): New typedef.
31503         * builtins.h (target_builtins::x_apply_args_mode)
31504         (target_builtins::x_apply_result_mode): Change type to
31505         fixed_size_mode_pod.
31506         * builtins.c (apply_args_size, apply_result_size, result_vector)
31507         (expand_builtin_apply_args_1, expand_builtin_apply)
31508         (expand_builtin_return): Update accordingly.
31510 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31512         * cse.c (hash_rtx_cb): Hash only the encoded elements.
31513         * cselib.c (cselib_hash_rtx): Likewise.
31514         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
31515         CONST_VECTOR encoding.
31517 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31518             Jeff Law  <law@redhat.com>
31520         PR target/83641
31521         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
31522         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
31523         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
31524         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
31526         PR target/83641
31527         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
31528         explicitly probe *sp in a noreturn function if there were any callee
31529         register saves or frame pointer is needed.
31531 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31533         PR debug/83621
31534         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
31535         BLKmode for ternary, binary or unary expressions.
31537         PR debug/83645
31538         * var-tracking.c (delete_vta_debug_insn): New inline function.
31539         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
31540         insns from get_insns () to NULL instead of each bb separately.
31541         Use delete_vta_debug_insn.  No longer static.
31542         (vt_debug_insns_local, variable_tracking_main_1): Adjust
31543         delete_vta_debug_insns callers.
31544         * rtl.h (delete_vta_debug_insns): Declare.
31545         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
31546         instead of variable_tracking_main.
31548 2018-01-03  Martin Sebor  <msebor@redhat.com>
31550         PR tree-optimization/83603
31551         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
31552         arguments past the endof the argument list in functions declared
31553         without a prototype.
31554         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
31555         Avoid checking when arguments are null.
31557 2018-01-03  Martin Sebor  <msebor@redhat.com>
31559         PR c/83559
31560         * doc/extend.texi (attribute const): Fix a typo.
31561         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
31562         issuing -Wsuggest-attribute for void functions.
31564 2018-01-03  Martin Sebor  <msebor@redhat.com>
31566         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
31567         offset_int::from instead of wide_int::to_shwi.
31568         (maybe_diag_overlap): Remove assertion.
31569         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
31570         * gimple-ssa-sprintf.c (format_directive): Same.
31571         (parse_directive): Same.
31572         (sprintf_dom_walker::compute_format_length): Same.
31573         (try_substitute_return_value): Same.
31575 2018-01-03  Jeff Law  <law@redhat.com>
31577         PR middle-end/83654
31578         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
31579         non-constant residual for zero at runtime and avoid probing in
31580         that case.  Reorganize code for trailing problem to mirror handling
31581         of the residual.
31583 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
31585         PR tree-optimization/83501
31586         * tree-ssa-strlen.c (get_string_cst): New.
31587         (handle_char_store): Call get_string_cst.
31589 2018-01-03  Martin Liska  <mliska@suse.cz>
31591         PR tree-optimization/83593
31592         * tree-ssa-strlen.c: Include tree-cfg.h.
31593         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
31594         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
31595         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
31596         to false.
31597         (strlen_dom_walker::before_dom_children): Call
31598         gimple_purge_dead_eh_edges. Dump tranformation with details
31599         dump flags.
31600         (strlen_dom_walker::before_dom_children): Update call by adding
31601         new argument cleanup_eh.
31602         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
31604 2018-01-03  Martin Liska  <mliska@suse.cz>
31606         PR ipa/83549
31607         * cif-code.def (VARIADIC_THUNK): New enum value.
31608         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
31609         thunks.
31611 2018-01-03  Jan Beulich  <jbeulich@suse.com>
31613         * sse.md (mov<mode>_internal): Tighten condition for when to use
31614         vmovdqu<ssescalarsize> for TI and OI modes.
31616 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31618         Update copyright years.
31620 2018-01-03  Martin Liska  <mliska@suse.cz>
31622         PR ipa/83594
31623         * ipa-visibility.c (function_and_variable_visibility): Skip
31624         functions with noipa attribure.
31626 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31628         * gcc.c (process_command): Update copyright notice dates.
31629         * gcov-dump.c (print_version): Ditto.
31630         * gcov.c (print_version): Ditto.
31631         * gcov-tool.c (print_version): Ditto.
31632         * gengtype.c (create_file): Ditto.
31633         * doc/cpp.texi: Bump @copying's copyright year.
31634         * doc/cppinternals.texi: Ditto.
31635         * doc/gcc.texi: Ditto.
31636         * doc/gccint.texi: Ditto.
31637         * doc/gcov.texi: Ditto.
31638         * doc/install.texi: Ditto.
31639         * doc/invoke.texi: Ditto.
31641 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31643         * vector-builder.h (vector_builder::m_full_nelts): Change from
31644         unsigned int to poly_uint64.
31645         (vector_builder::full_nelts): Update prototype accordingly.
31646         (vector_builder::new_vector): Likewise.
31647         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
31648         (vector_builder::operator ==): Likewise.
31649         (vector_builder::finalize): Likewise.
31650         * int-vector-builder.h (int_vector_builder::int_vector_builder):
31651         Take the number of elements as a poly_uint64 rather than an
31652         unsigned int.
31653         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
31654         from unsigned int to poly_uint64.
31655         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
31656         (vec_perm_indices::new_vector): Likewise.
31657         (vec_perm_indices::length): Likewise.
31658         (vec_perm_indices::nelts_per_input): Likewise.
31659         (vec_perm_indices::input_nelts): Likewise.
31660         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
31661         number of elements per input as a poly_uint64 rather than an
31662         unsigned int.  Use the original encoding for variable-length
31663         vectors, rather than clamping each individual element.
31664         For the second and subsequent elements in each pattern,
31665         clamp the step and base before clamping their sum.
31666         (vec_perm_indices::series_p): Handle polynomial element counts.
31667         (vec_perm_indices::all_in_range_p): Likewise.
31668         (vec_perm_indices_to_tree): Likewise.
31669         (vec_perm_indices_to_rtx): Likewise.
31670         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
31671         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
31672         (tree_vector_builder::new_binary_operation): Handle polynomial
31673         element counts.  Return false if we need to know the number
31674         of elements at compile time.
31675         * fold-const.c (fold_vec_perm): Punt if the number of elements
31676         isn't known at compile time.
31678 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31680         * vec-perm-indices.h (vec_perm_builder): Change element type
31681         from HOST_WIDE_INT to poly_int64.
31682         (vec_perm_indices::element_type): Update accordingly.
31683         (vec_perm_indices::clamp): Handle polynomial element_types.
31684         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31685         (vec_perm_indices::all_in_range_p): Likewise.
31686         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
31687         than shwi trees.
31688         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
31689         polynomial vec_perm_indices element types.
31690         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
31691         * fold-const.c (fold_vec_perm): Likewise.
31692         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
31693         * tree-vect-generic.c (lower_vec_perm): Likewise.
31694         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31695         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
31696         element type to HOST_WIDE_INT.
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         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
31703         rather than an int.  Use plus_constant.
31704         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
31705         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
31707 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31708             Alan Hayward  <alan.hayward@arm.com>
31709             David Sherwood  <david.sherwood@arm.com>
31711         * calls.c (emit_call_1, expand_call): Change struct_value_size from
31712         a HOST_WIDE_INT to a poly_int64.
31714 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31715             Alan Hayward  <alan.hayward@arm.com>
31716             David Sherwood  <david.sherwood@arm.com>
31718         * calls.c (load_register_parameters): Cope with polynomial
31719         mode sizes.  Require a constant size for BLKmode parameters
31720         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
31721         forces a parameter to be padded at the lsb end in order to
31722         fill a complete number of words, require the parameter size
31723         to be ordered wrt UNITS_PER_WORD.
31725 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31726             Alan Hayward  <alan.hayward@arm.com>
31727             David Sherwood  <david.sherwood@arm.com>
31729         * reload1.c (spill_stack_slot_width): Change element type
31730         from unsigned int to poly_uint64_pod.
31731         (alter_reg): Treat mode sizes as polynomial.
31733 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31734             Alan Hayward  <alan.hayward@arm.com>
31735             David Sherwood  <david.sherwood@arm.com>
31737         * reload.c (complex_word_subreg_p): New function.
31738         (reload_inner_reg_of_subreg, push_reload): Use it.
31740 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31741             Alan Hayward  <alan.hayward@arm.com>
31742             David Sherwood  <david.sherwood@arm.com>
31744         * lra-constraints.c (process_alt_operands): Reject matched
31745         operands whose sizes aren't ordered.
31746         (match_reload): Refer to this check here.
31748 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31749             Alan Hayward  <alan.hayward@arm.com>
31750             David Sherwood  <david.sherwood@arm.com>
31752         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
31753         that the mode size is in the set {1, 2, 4, 8, 16}.
31755 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31756             Alan Hayward  <alan.hayward@arm.com>
31757             David Sherwood  <david.sherwood@arm.com>
31759         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
31760         Use plus_constant instead of gen_rtx_PLUS.
31762 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31763             Alan Hayward  <alan.hayward@arm.com>
31764             David Sherwood  <david.sherwood@arm.com>
31766         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
31767         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
31768         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
31769         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
31770         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
31771         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
31772         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
31773         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
31774         * config/i386/i386.c (ix86_push_rounding): ...this new function.
31775         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
31776         a poly_int64.
31777         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
31778         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
31779         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
31780         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
31781         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
31782         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
31783         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
31784         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
31785         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
31786         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
31787         function.
31788         * expr.c (emit_move_resolve_push): Treat the input and result
31789         of PUSH_ROUNDING as a poly_int64.
31790         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
31791         (emit_push_insn): Likewise.
31792         * lra-eliminations.c (mark_not_eliminable): Likewise.
31793         * recog.c (push_operand): Likewise.
31794         * reload1.c (elimination_effects): Likewise.
31795         * rtlanal.c (nonzero_bits1): Likewise.
31796         * calls.c (store_one_arg): Likewise.  Require the padding to be
31797         known at compile time.
31799 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31800             Alan Hayward  <alan.hayward@arm.com>
31801             David Sherwood  <david.sherwood@arm.com>
31803         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
31804         Use plus_constant instead of gen_rtx_PLUS.
31806 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31807             Alan Hayward  <alan.hayward@arm.com>
31808             David Sherwood  <david.sherwood@arm.com>
31810         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
31811         rather than an int.
31813 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31814             Alan Hayward  <alan.hayward@arm.com>
31815             David Sherwood  <david.sherwood@arm.com>
31817         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
31818         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
31819         via stack temporaries.  Treat the mode size as polynomial too.
31821 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31822             Alan Hayward  <alan.hayward@arm.com>
31823             David Sherwood  <david.sherwood@arm.com>
31825         * expr.c (expand_expr_real_2): When handling conversions involving
31826         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
31827         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
31828         as a poly_uint64 too.
31830 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31831             Alan Hayward  <alan.hayward@arm.com>
31832             David Sherwood  <david.sherwood@arm.com>
31834         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
31836 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31837             Alan Hayward  <alan.hayward@arm.com>
31838             David Sherwood  <david.sherwood@arm.com>
31840         * combine.c (can_change_dest_mode): Handle polynomial
31841         REGMODE_NATURAL_SIZE.
31842         * expmed.c (store_bit_field_1): Likewise.
31843         * expr.c (store_constructor): Likewise.
31844         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
31845         and polynomial REGMODE_NATURAL_SIZE.
31846         (gen_lowpart_common): Likewise.
31847         * reginfo.c (record_subregs_of_mode): Likewise.
31848         * rtlanal.c (read_modify_subreg_p): Likewise.
31850 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31851             Alan Hayward  <alan.hayward@arm.com>
31852             David Sherwood  <david.sherwood@arm.com>
31854         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
31855         numbers of elements.
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         * match.pd: Cope with polynomial numbers of vector elements.
31863 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31864             Alan Hayward  <alan.hayward@arm.com>
31865             David Sherwood  <david.sherwood@arm.com>
31867         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
31868         in a POINTER_PLUS_EXPR.
31870 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31871             Alan Hayward  <alan.hayward@arm.com>
31872             David Sherwood  <david.sherwood@arm.com>
31874         * omp-simd-clone.c (simd_clone_subparts): New function.
31875         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
31876         (ipa_simd_modify_function_body): Likewise.
31878 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31879             Alan Hayward  <alan.hayward@arm.com>
31880             David Sherwood  <david.sherwood@arm.com>
31882         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
31883         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
31884         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
31885         (expand_vector_condition, vector_element): Likewise.
31886         (subparts_gt): New function.
31887         (get_compute_type): Use subparts_gt.
31888         (count_type_subparts): Delete.
31889         (expand_vector_operations_1): Use subparts_gt instead of
31890         count_type_subparts.
31892 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31893             Alan Hayward  <alan.hayward@arm.com>
31894             David Sherwood  <david.sherwood@arm.com>
31896         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
31897         (vect_compile_time_alias): ...this new function.  Do the calculation
31898         on poly_ints rather than trees.
31899         (vect_prune_runtime_alias_test_list): Update call accordingly.
31901 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31902             Alan Hayward  <alan.hayward@arm.com>
31903             David Sherwood  <david.sherwood@arm.com>
31905         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
31906         numbers of units.
31907         (vect_schedule_slp_instance): Likewise.
31909 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31910             Alan Hayward  <alan.hayward@arm.com>
31911             David Sherwood  <david.sherwood@arm.com>
31913         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
31914         constant and extern definitions for variable-length vectors.
31915         (vect_get_constant_vectors): Note that the number of units
31916         is known to be constant.
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-stmts.c (vectorizable_conversion): Treat the number
31923         of units as polynomial.  Choose between WIDE and NARROW based
31924         on multiple_p.
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-stmts.c (simd_clone_subparts): New function.
31931         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
31933 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31934             Alan Hayward  <alan.hayward@arm.com>
31935             David Sherwood  <david.sherwood@arm.com>
31937         * tree-vect-stmts.c (vectorizable_call): Treat the number of
31938         vectors as polynomial.  Use build_index_vector for
31939         IFN_GOMP_SIMD_LANE.
31941 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31942             Alan Hayward  <alan.hayward@arm.com>
31943             David Sherwood  <david.sherwood@arm.com>
31945         * tree-vect-stmts.c (get_load_store_type): Treat the number of
31946         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
31947         for variable-length vectors.
31948         (vectorizable_mask_load_store): Treat the number of units as
31949         polynomial, asserting that it is constant if the condition has
31950         already been enforced.
31951         (vectorizable_store, vectorizable_load): Likewise.
31953 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31954             Alan Hayward  <alan.hayward@arm.com>
31955             David Sherwood  <david.sherwood@arm.com>
31957         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
31958         of units as polynomial.  Punt if we can't tell at compile time
31959         which vector contains the final result.
31961 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31962             Alan Hayward  <alan.hayward@arm.com>
31963             David Sherwood  <david.sherwood@arm.com>
31965         * tree-vect-loop.c (vectorizable_induction): Treat the number
31966         of units as polynomial.  Punt on SLP inductions.  Use an integer
31967         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
31968         cast of such a series for variable-length floating-point
31969         reductions.
31971 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31972             Alan Hayward  <alan.hayward@arm.com>
31973             David Sherwood  <david.sherwood@arm.com>
31975         * tree.h (build_index_vector): Declare.
31976         * tree.c (build_index_vector): New function.
31977         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
31978         of units as polynomial, forcibly converting it to a constant if
31979         vectorizable_reduction has already enforced the condition.
31980         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
31981         to create a {1,2,3,...} vector.
31982         (vectorizable_reduction): Treat the number of units as polynomial.
31983         Choose vectype_in based on the largest scalar element size rather
31984         than the smallest number of units.  Enforce the restrictions
31985         relied on above.
31987 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31988             Alan Hayward  <alan.hayward@arm.com>
31989             David Sherwood  <david.sherwood@arm.com>
31991         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
31992         number of units as polynomial.
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.h (vector_sizes, auto_vector_sizes): New typedefs.
31999         * target.def (autovectorize_vector_sizes): Return the vector sizes
32000         by pointer, using vector_sizes rather than a bitmask.
32001         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
32002         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
32003         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
32004         Likewise.
32005         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
32006         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
32007         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
32008         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
32009         * omp-general.c (omp_max_vf): Likewise.
32010         * omp-low.c (omp_clause_aligned_alignment): Likewise.
32011         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
32012         * tree-vect-loop.c (vect_analyze_loop): Likewise.
32013         * tree-vect-slp.c (vect_slp_bb): Likewise.
32014         * doc/tm.texi: Regenerate.
32015         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
32016         to a poly_uint64.
32017         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
32018         the vector size as a poly_uint64 rather than an unsigned int.
32019         (current_vector_size): Change from an unsigned int to a poly_uint64.
32020         (get_vectype_for_scalar_type): Update accordingly.
32021         * tree.h (build_truth_vector_type): Take the size and number of
32022         units as a poly_uint64 rather than an unsigned int.
32023         (build_vector_type): Add a temporary overload that takes
32024         the number of units as a poly_uint64 rather than an unsigned int.
32025         * tree.c (make_vector_type): Likewise.
32026         (build_truth_vector_type): Take the number of units as a poly_uint64
32027         rather than an unsigned int.
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         * target.def (get_mask_mode): Take the number of units and length
32034         as poly_uint64s rather than unsigned ints.
32035         * targhooks.h (default_get_mask_mode): Update accordingly.
32036         * targhooks.c (default_get_mask_mode): Likewise.
32037         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
32038         * doc/tm.texi: Regenerate.
32040 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32041             Alan Hayward  <alan.hayward@arm.com>
32042             David Sherwood  <david.sherwood@arm.com>
32044         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
32045         * omp-general.c (omp_max_vf): Likewise.
32046         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
32047         (expand_omp_simd): Handle polynomial safelen.
32048         * omp-low.c (omplow_simd_context): Add a default constructor.
32049         (omplow_simd_context::max_vf): Change from int to poly_uint64.
32050         (lower_rec_simd_input_clauses): Update accordingly.
32051         (lower_rec_input_clauses): Likewise.
32053 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32054             Alan Hayward  <alan.hayward@arm.com>
32055             David Sherwood  <david.sherwood@arm.com>
32057         * tree-vectorizer.h (vect_nunits_for_cost): New function.
32058         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
32059         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
32060         (vect_analyze_slp_cost): Likewise.
32061         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
32062         (vect_model_load_cost): Likewise.
32064 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32065             Alan Hayward  <alan.hayward@arm.com>
32066             David Sherwood  <david.sherwood@arm.com>
32068         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
32069         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
32070         from an unsigned int * to a poly_uint64_pod *.
32071         (calculate_unrolling_factor): New function.
32072         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
32074 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32075             Alan Hayward  <alan.hayward@arm.com>
32076             David Sherwood  <david.sherwood@arm.com>
32078         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
32079         from an unsigned int to a poly_uint64.
32080         (_loop_vec_info::slp_unrolling_factor): Likewise.
32081         (_loop_vec_info::vectorization_factor): Change from an int
32082         to a poly_uint64.
32083         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
32084         (vect_get_num_vectors): New function.
32085         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
32086         (vect_get_num_copies): Use vect_get_num_vectors.
32087         (vect_analyze_data_ref_dependences): Change max_vf from an int *
32088         to an unsigned int *.
32089         (vect_analyze_data_refs): Change min_vf from an int * to a
32090         poly_uint64 *.
32091         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
32092         than an unsigned HOST_WIDE_INT.
32093         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
32094         (vect_analyze_data_ref_dependence): Change max_vf from an int *
32095         to an unsigned int *.
32096         (vect_analyze_data_ref_dependences): Likewise.
32097         (vect_compute_data_ref_alignment): Handle polynomial vf.
32098         (vect_enhance_data_refs_alignment): Likewise.
32099         (vect_prune_runtime_alias_test_list): Likewise.
32100         (vect_shift_permute_load_chain): Likewise.
32101         (vect_supportable_dr_alignment): Likewise.
32102         (dependence_distance_ge_vf): Take the vectorization factor as a
32103         poly_uint64 rather than an unsigned HOST_WIDE_INT.
32104         (vect_analyze_data_refs): Change min_vf from an int * to a
32105         poly_uint64 *.
32106         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
32107         vfm1 as a poly_uint64 rather than an int.  Make the same change
32108         for the returned bound_scalar.
32109         (vect_gen_vector_loop_niters): Handle polynomial vf.
32110         (vect_do_peeling): Likewise.  Update call to
32111         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
32112         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
32113         be constant.
32114         * tree-vect-loop.c (vect_determine_vectorization_factor)
32115         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
32116         (vect_get_known_peeling_cost): Likewise.
32117         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
32118         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
32119         (vect_transform_loop): Likewise.  Use the lowest possible VF when
32120         updating the upper bounds of the loop.
32121         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
32122         rather than an int.
32123         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
32124         polynomial unroll factors.
32125         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
32126         (vect_make_slp_decision): Likewise.
32127         (vect_supported_load_permutation_p): Likewise, and polynomial
32128         vf too.
32129         (vect_analyze_slp_cost): Handle polynomial vf.
32130         (vect_slp_analyze_node_operations): Likewise.
32131         (vect_slp_analyze_bb_1): Likewise.
32132         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
32133         than an unsigned HOST_WIDE_INT.
32134         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
32135         (vectorizable_load): Handle polynomial vf.
32136         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
32137         a poly_uint64.
32138         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
32140 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32141             Alan Hayward  <alan.hayward@arm.com>
32142             David Sherwood  <david.sherwood@arm.com>
32144         * match.pd: Handle bit operations involving three constants
32145         and try to fold one pair.
32147 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32149         * tree-vect-loop-manip.c: Include gimple-fold.h.
32150         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
32151         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
32152         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
32153         Add a path that uses a step of VF instead of 1, but disable it
32154         for now.
32155         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
32156         and niters_no_overflow parameters.  Update calls to
32157         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
32158         Create a new SSA name if the latter choses to use a ste other
32159         than zero, and return it via niters_vector_mult_vf_var.
32160         * tree-vect-loop.c (vect_transform_loop): Update calls to
32161         vect_do_peeling, vect_gen_vector_loop_niters and
32162         slpeel_make_loop_iterate_ntimes.
32163         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
32164         (vect_gen_vector_loop_niters): Update declarations after above changes.
32166 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
32168         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
32169         128-bit round to integer instructions.
32170         (ceil<mode>2): Likewise.
32171         (btrunc<mode>2): Likewise.
32172         (round<mode>2): Likewise.
32174 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32176         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
32177         unaligned VSX load/store on P8/P9.
32178         (expand_block_clear): Allow the use of unaligned VSX
32179         load/store on P8/P9.
32181 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32183         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
32184         New function.
32185         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
32186         swap associated with both a load and a store.
32188 2018-01-02  Andrew Waterman  <andrew@sifive.com>
32190         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
32191         * config/riscv/riscv.md (clear_cache): Use it.
32193 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
32195         * web.c: Remove out-of-date comment.
32197 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32199         * expr.c (fixup_args_size_notes): Check that any existing
32200         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
32201         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
32202         (emit_single_push_insn): ...here.
32204 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32206         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
32207         (const_vector_encoded_nelts): New function.
32208         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
32209         (const_vector_int_elt, const_vector_elt): Declare.
32210         * emit-rtl.c (const_vector_int_elt_1): New function.
32211         (const_vector_elt): Likewise.
32212         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
32213         of CONST_VECTOR_ELT.
32215 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32217         * expr.c: Include rtx-vector-builder.h.
32218         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
32219         directly on the tree encoding.
32220         (const_vector_from_tree): Likewise.
32221         * optabs.c: Include rtx-vector-builder.h.
32222         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
32223         sequence of "u" values.
32224         * vec-perm-indices.c: Include rtx-vector-builder.h.
32225         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
32226         directly on the vec_perm_indices encoding.
32228 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32230         * doc/rtl.texi (const_vector): Describe new encoding scheme.
32231         * Makefile.in (OBJS): Add rtx-vector-builder.o.
32232         * rtx-vector-builder.h: New file.
32233         * rtx-vector-builder.c: Likewise.
32234         * rtl.h (rtx_def::u2): Add a const_vector field.
32235         (CONST_VECTOR_NPATTERNS): New macro.
32236         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
32237         (CONST_VECTOR_DUPLICATE_P): Likewise.
32238         (CONST_VECTOR_STEPPED_P): Likewise.
32239         (CONST_VECTOR_ENCODED_ELT): Likewise.
32240         (const_vec_duplicate_p): Check for a duplicated vector encoding.
32241         (unwrap_const_vec_duplicate): Likewise.
32242         (const_vec_series_p): Check for a non-duplicated vector encoding.
32243         Say that the function only returns true for integer vectors.
32244         * emit-rtl.c: Include rtx-vector-builder.h.
32245         (gen_const_vec_duplicate_1): Delete.
32246         (gen_const_vector): Call gen_const_vec_duplicate instead of
32247         gen_const_vec_duplicate_1.
32248         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
32249         (gen_const_vec_duplicate): Use rtx_vector_builder.
32250         (gen_const_vec_series): Likewise.
32251         (gen_rtx_CONST_VECTOR): Likewise.
32252         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
32253         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32254         Build a new vector rather than modifying a CONST_VECTOR in-place.
32255         (handle_special_swappables): Update call accordingly.
32256         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
32257         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32258         Build a new vector rather than modifying a CONST_VECTOR in-place.
32259         (handle_special_swappables): Update call accordingly.
32261 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32263         * simplify-rtx.c (simplify_const_binary_operation): Use
32264         CONST_VECTOR_ELT instead of XVECEXP.
32266 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32268         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
32269         the selector elements to be different from the data elements
32270         if the selector is a VECTOR_CST.
32271         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
32272         ssizetype for the selector.
32274 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32276         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
32277         before testing each element individually.
32278         * tree-vect-generic.c (lower_vec_perm): Likewise.
32280 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32282         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
32283         * selftest-run-tests.c (selftest::run_tests): Call it.
32284         * vector-builder.h (vector_builder::operator ==): New function.
32285         (vector_builder::operator !=): Likewise.
32286         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
32287         (vec_perm_indices::all_from_input_p): New function.
32288         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
32289         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
32290         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
32291         instead of reading the VECTOR_CST directly.  Detect whether both
32292         vector inputs are the same before constructing the vec_perm_indices,
32293         and update the number of inputs argument accordingly.  Use the
32294         utility functions added above.  Only construct sel2 if we need to.
32296 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32298         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
32299         the broadcast of the low byte.
32300         (expand_mult_highpart): Use an explicit encoding for the permutes.
32301         * optabs-query.c (can_mult_highpart_p): Likewise.
32302         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
32303         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32304         (vectorizable_bswap): Likewise.
32305         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
32306         explicit encoding for the power-of-2 permutes.
32307         (vect_permute_store_chain): Likewise.
32308         (vect_grouped_load_supported): Likewise.
32309         (vect_permute_load_chain): Likewise.
32311 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32313         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
32314         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
32315         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
32316         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
32317         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
32318         (vect_gen_perm_mask_any): Likewise.
32320 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32322         * int-vector-builder.h: New file.
32323         * vec-perm-indices.h: Include int-vector-builder.h.
32324         (vec_perm_indices): Redefine as an int_vector_builder.
32325         (auto_vec_perm_indices): Delete.
32326         (vec_perm_builder): Redefine as a stand-alone class.
32327         (vec_perm_indices::vec_perm_indices): New function.
32328         (vec_perm_indices::clamp): Likewise.
32329         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
32330         (vec_perm_indices::new_vector): New function.
32331         (vec_perm_indices::new_expanded_vector): Update for new
32332         vec_perm_indices class.
32333         (vec_perm_indices::rotate_inputs): New function.
32334         (vec_perm_indices::all_in_range_p): Operate directly on the
32335         encoded form, without computing elided elements.
32336         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
32337         encoding.  Update for new vec_perm_indices class.
32338         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
32339         the given vec_perm_builder.
32340         (expand_vec_perm_var): Update vec_perm_builder constructor.
32341         (expand_mult_highpart): Use vec_perm_builder instead of
32342         auto_vec_perm_indices.
32343         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
32344         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
32345         or double series encoding as appropriate.
32346         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
32347         vec_perm_indices instead of auto_vec_perm_indices.
32348         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32349         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32350         (vect_permute_store_chain): Likewise.
32351         (vect_grouped_load_supported): Likewise.
32352         (vect_permute_load_chain): Likewise.
32353         (vect_shift_permute_load_chain): Likewise.
32354         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32355         (vect_transform_slp_perm_load): Likewise.
32356         (vect_schedule_slp_instance): Likewise.
32357         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32358         (vectorizable_mask_load_store): Likewise.
32359         (vectorizable_bswap): Likewise.
32360         (vectorizable_store): Likewise.
32361         (vectorizable_load): Likewise.
32362         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
32363         vec_perm_indices instead of auto_vec_perm_indices.  Use
32364         tree_to_vec_perm_builder to read the vector from a tree.
32365         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
32366         vec_perm_builder instead of a vec_perm_indices.
32367         (have_whole_vector_shift): Use vec_perm_builder and
32368         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
32369         truncation to calc_vec_perm_mask_for_shift.
32370         (vect_create_epilog_for_reduction): Likewise.
32371         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
32372         from auto_vec_perm_indices to vec_perm_indices.
32373         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32374         instead of changing individual elements.
32375         (aarch64_vectorize_vec_perm_const): Use new_vector to install
32376         the vector in d.perm.
32377         * config/arm/arm.c (expand_vec_perm_d::perm): Change
32378         from auto_vec_perm_indices to vec_perm_indices.
32379         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32380         instead of changing individual elements.
32381         (arm_vectorize_vec_perm_const): Use new_vector to install
32382         the vector in d.perm.
32383         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
32384         Update vec_perm_builder constructor.
32385         (rs6000_expand_interleave): Likewise.
32386         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
32387         (rs6000_expand_interleave): Likewise.
32389 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32391         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
32392         to qimode could truncate the indices.
32393         * optabs.c (expand_vec_perm_var): Likewise.
32395 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32397         * Makefile.in (OBJS): Add vec-perm-indices.o.
32398         * vec-perm-indices.h: New file.
32399         * vec-perm-indices.c: Likewise.
32400         * target.h (vec_perm_indices): Replace with a forward class
32401         declaration.
32402         (auto_vec_perm_indices): Move to vec-perm-indices.h.
32403         * optabs.h: Include vec-perm-indices.h.
32404         (expand_vec_perm): Delete.
32405         (selector_fits_mode_p, expand_vec_perm_var): Declare.
32406         (expand_vec_perm_const): Declare.
32407         * target.def (vec_perm_const_ok): Replace with...
32408         (vec_perm_const): ...this new hook.
32409         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
32410         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
32411         * doc/tm.texi: Regenerate.
32412         * optabs.def (vec_perm_const): Delete.
32413         * doc/md.texi (vec_perm_const): Likewise.
32414         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
32415         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
32416         expand_vec_perm for constant permutation vectors.  Assert that
32417         the mode of variable permutation vectors is the integer equivalent
32418         of the mode that is being permuted.
32419         * optabs-query.h (selector_fits_mode_p): Declare.
32420         * optabs-query.c: Include vec-perm-indices.h.
32421         (selector_fits_mode_p): New function.
32422         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
32423         is defined, instead of checking whether the vec_perm_const_optab
32424         exists.  Use targetm.vectorize.vec_perm_const instead of
32425         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
32426         fit in the vector mode before using a variable permute.
32427         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
32428         vec_perm_indices instead of an rtx.
32429         (expand_vec_perm): Replace with...
32430         (expand_vec_perm_const): ...this new function.  Take the selector
32431         as a vec_perm_indices rather than an rtx.  Also take the mode of
32432         the selector.  Update call to shift_amt_for_vec_perm_mask.
32433         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
32434         Use vec_perm_indices::new_expanded_vector to expand the original
32435         selector into bytes.  Check whether the indices fit in the vector
32436         mode before using a variable permute.
32437         (expand_vec_perm_var): Make global.
32438         (expand_mult_highpart): Use expand_vec_perm_const.
32439         * fold-const.c: Includes vec-perm-indices.h.
32440         * tree-ssa-forwprop.c: Likewise.
32441         * tree-vect-data-refs.c: Likewise.
32442         * tree-vect-generic.c: Likewise.
32443         * tree-vect-loop.c: Likewise.
32444         * tree-vect-slp.c: Likewise.
32445         * tree-vect-stmts.c: Likewise.
32446         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
32447         Delete.
32448         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
32449         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
32450         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
32451         (aarch64_vectorize_vec_perm_const): ...this new function.
32452         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32453         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32454         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
32455         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
32456         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32457         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32458         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
32459         into...
32460         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
32461         check for NEON modes.
32462         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
32463         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
32464         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
32465         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
32466         into...
32467         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
32468         the old VEC_PERM_CONST conditions.
32469         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
32470         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
32471         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
32472         (ia64_vectorize_vec_perm_const_ok): Merge into...
32473         (ia64_vectorize_vec_perm_const): ...this new function.
32474         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
32475         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
32476         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
32477         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
32478         * config/mips/mips.c (mips_expand_vec_perm_const)
32479         (mips_vectorize_vec_perm_const_ok): Merge into...
32480         (mips_vectorize_vec_perm_const): ...this new function.
32481         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
32482         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
32483         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
32484         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
32485         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
32486         (rs6000_expand_vec_perm_const): Delete.
32487         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
32488         Delete.
32489         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32490         (altivec_expand_vec_perm_const_le): Take each operand individually.
32491         Operate on constant selectors rather than rtxes.
32492         (altivec_expand_vec_perm_const): Likewise.  Update call to
32493         altivec_expand_vec_perm_const_le.
32494         (rs6000_expand_vec_perm_const): Delete.
32495         (rs6000_vectorize_vec_perm_const_ok): Delete.
32496         (rs6000_vectorize_vec_perm_const): New function.
32497         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32498         an element count and rtx array.
32499         (rs6000_expand_extract_even): Update call accordingly.
32500         (rs6000_expand_interleave): Likewise.
32501         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
32502         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
32503         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
32504         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
32505         (rs6000_expand_vec_perm_const): Delete.
32506         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32507         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32508         (altivec_expand_vec_perm_const_le): Take each operand individually.
32509         Operate on constant selectors rather than rtxes.
32510         (altivec_expand_vec_perm_const): Likewise.  Update call to
32511         altivec_expand_vec_perm_const_le.
32512         (rs6000_expand_vec_perm_const): Delete.
32513         (rs6000_vectorize_vec_perm_const_ok): Delete.
32514         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
32515         reference to the SPE evmerge intructions.
32516         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32517         an element count and rtx array.
32518         (rs6000_expand_extract_even): Update call accordingly.
32519         (rs6000_expand_interleave): Likewise.
32520         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
32521         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
32522         new function.
32523         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32525 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32527         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
32528         vector mode and that that mode matches the mode of the data
32529         being permuted.
32530         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
32531         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
32532         directly using expand_vec_perm_1 when forcing selectors into
32533         registers.
32534         (expand_vec_perm_var): New function, split out from expand_vec_perm.
32536 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32538         * optabs-query.h (can_vec_perm_p): Delete.
32539         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
32540         * optabs-query.c (can_vec_perm_p): Split into...
32541         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
32542         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
32543         particular selector is valid.
32544         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32545         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32546         (vect_grouped_load_supported): Likewise.
32547         (vect_shift_permute_load_chain): Likewise.
32548         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32549         (vect_transform_slp_perm_load): Likewise.
32550         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32551         (vectorizable_bswap): Likewise.
32552         (vect_gen_perm_mask_checked): Likewise.
32553         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
32554         implementations of variable permutation vectors into account
32555         when deciding which selector to use.
32556         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
32557         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
32558         with a false third argument.
32559         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
32560         to test whether the constant selector is valid and can_vec_perm_var_p
32561         to test whether a variable selector is valid.
32563 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32565         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
32566         * optabs-query.c (can_vec_perm_p): Likewise.
32567         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
32568         instead of vec_perm_indices.
32569         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
32570         (vect_gen_perm_mask_checked): Likewise,
32571         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
32572         (vect_gen_perm_mask_checked): Likewise,
32574 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32576         * optabs-query.h (qimode_for_vec_perm): Declare.
32577         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
32578         (qimode_for_vec_perm): ...this new function.
32579         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
32581 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32583         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
32584         does not have a conditional at the top.
32586 2018-01-02  Richard Biener  <rguenther@suse.de>
32588         * ipa-inline.c (big_speedup_p): Fix expression.
32590 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32592         PR target/81616
32593         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
32594         for generic 4->6.
32596 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32598         PR target/81616
32599         Generic tuning.
32600         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
32601         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
32602         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
32603         cond_taken_branch_cost 3->4.
32605 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
32607         PR tree-optimization/83581
32608         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
32609         TODO_cleanup_cfg if any changes have been made.
32611         PR middle-end/83608
32612         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
32613         convert_modes if target mode has the right side, but different mode
32614         class.
32616         PR middle-end/83609
32617         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
32618         last argument when extracting from CONCAT.  If either from_real or
32619         from_imag is NULL, use expansion through memory.  If result is not
32620         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
32621         the parts directly to inner mode, if even that fails, use expansion
32622         through memory.
32624         PR middle-end/83623
32625         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
32626         check for bswap in mode rather than HImode and use that in expand_unop
32627         too.
32629 Copyright (C) 2018 Free Software Foundation, Inc.
32631 Copying and distribution of this file, with or without modification,
32632 are permitted in any medium without royalty provided the copyright
32633 notice and this notice are preserved.