PR tree-optimization/87546
[official-gcc.git] / gcc / ChangeLog
blob3392f8f498a0e7b50dd98c63c004a5ea4563c873
1 2018-11-17  Jakub Jelinek  <jakub@redhat.com>
3         PR tree-optimization/87546
4         * tree-vect-patterns.c (vect_look_through_possible_promotion): Add
5         min_precision variable, initially set it to orig_precision, only does
6         something if op_type's precision is <= min_precision and update
7         min_precision whenever calling set_op.
9 2018-11-16  Jan Hubicka  <hubicka@ucw.cz>
11         PR ipa/87957
12         * ipa-devirt.c (warn_odr): Look for main variant to get TYPE_DECL.
14 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
16         * doc/invoke.texi (Option Summary): Fix whitespace and line
17         breaks in @gccoptlist environments.
18         (Warning Options): Likewise.
19         (Optimize Options): Likewise.
20         (PowerPC SPE Options): Likewise.
21         (RS/6000 and PowerPC Options): Likewise.
23 2018-11-16  Jeff Law  <law@redhat.com>
25         * config/mn10300/mn10300.md (adddi3_degenerate): Remove bogus
26         gcc_assert.
28 2018-11-16  Sandra Loosemore  <sandra@codesourcery.com>
30         PR middle-end/23197
31         * doc/invoke.texi (Optimize Options): Update options enabled by
32         fprofile-generate, -fprofile-use, and -fauto-profile.
34 2018-11-16  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
36         PR target/87927
37         * target-def.h: Initialize TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP.
38         Add them to the TARGET_ASM_{,UN}ALIGNED_INT_OP structs.
39         * target.def: Enumerate TARGET_ASM_{,UN}ALIGNED_P{S,D,T}I_OP in
40         the byte_op hook.
41         * target.h: Add psi, pdi, pti to struct asm_int_op definition.
42         * targhooks.c (default_print_patchable_function_entry): Assert
43         asm_int_op does not return a NULL string.
44         * varasm.c (integer_asm_op): Return the op for a partial int type
45         when the requested size does not correspond to an integer type.
46         * config/msp430/msp430.c: Initialize TARGET_ASM_{,UN}ALIGNED_PSI_OP.
47         * doc/tm.texi: Regenerate.
49 2018-11-12  Jason Merrill  <jason@redhat.com>
51         * gimplify.c (gimplify_case_label_expr): Handle hot/cold attributes.
53 2018-11-16  Michael Meissner  <meissner@linux.ibm.com>
55         * config/rs6000/constraints.md (wF constraint): Remove power9
56         fusion documentation.  Just document wF's use for power8 fusion.
57         * config/rs6000/predicates.md (p9_fusion_reg_operand): Delete.
58         (fusion_gpr_addis): Delete power9 fusion support.  Change power8
59         fusion support to require the upper 12 bits to be all 0's or all
60         1's.
61         (fusion_gpr_mem_load): Add comment.
62         (fusion_addis_mem_combo_load): Remove power9 fusion support.
63         (fusion_addis_mem_combo_store): Delete.
64         (fusion_offsettable_mem_operand): Delete.
65         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Do not set
66         power8 fusion here.
67         (ISA_3_0_MASKS_SERVER): Delete power9 fusion.
68         (POWERPC_MASKS): Delete power9 fusion.
69         * config/rs6000/rs6000-protos.h (emit_fusion_load_store): Delete.
70         (fusion_p9_p): Delete.
71         (expand_fusion_p9_load): Delete.
72         (expand_fusion_p9_store): Delete.
73         (emit_fusion_p9_load): Delete.
74         (emit_fusion_p9_store): Delete.
75         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Delete power9
76         fusion support.
77         (rs6000_option_override_internal): Set power8 fusion based on
78         whether we are tuning for power8.  Delete power9 fusion support.
79         (rs6000_opt_masks): Delete -mpower9-fusion switch.
80         (emit_fusion_load): Rename emit_fusion_load_store to
81         emit_fusion_load, and drop fusion store support.  Update callers.
82         (emit_fusion_load_store): Likewise.
83         (emit_fusion_gpr_load): Likewise.
84         (fusion_p9_p): Delete.
85         (expand_fusion_p9_load): Delete.
86         (expand_fusion_p9_store): Delete.
87         (emit_fusion_p9_load): Delete.
88         (emit_fusion_p9_store): Delete.
89         * config/rs6000/rs6000.md (UNSPEC_FUSION_P9): Delete.
90         (GPR_FUSION): Delete.
91         (FPR_FUSION): Delete.
92         (power9 fusion peephole2s): Delete.
93         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Delete.
94         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Delete.
95         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_load): Delete.
96         (fusion_vsx_<P:mode>_<FPR_FUSION:mode>_store): Delete.
97         (fusion_p9_<mode>_constant): Delete.
98         * config/rs6000/rs6000.opt (-mpower9-fusion): Delete undocumented
99         power9 fusion switch.
100         * doc/md.texi (PowerPC constraints): Update wF constraint
101         documentation for power8 fusion only.
103 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
105         PR rtl-optimization/87475
106         * cfgrtl.c (patch_jump_insn): Allow redirection failure for
107         CROSSING_JUMP_P insns.
108         (cfg_layout_redirect_edge_and_branch): Don't ICE if ret is NULL.
110 2018-11-16  Uros Bizjak  <ubizjak@gmail.com>
112         PR target/88051
113         * config/i386/i386.md (floatunsdidf2): Allow only 64bit AVX512F targets.
114         * config/i386/sse.md (UNSPEC_MOVDI_TO_SSE): New UNSPEC.
115         (movdi_to_sse): Rewrite using UNSPEC_MOVDI_TO_SSE unspec.
117 2018-11-16  Jakub Jelinek  <jakub@redhat.com>
119         PR middle-end/88032
120         * optabs.c (expand_binop): For op0_mode use GET_MODE (op0), unless it
121         is VOIDmode, in which case use int_mode.  Similarly for op1_mode.
123 2018-11-16  Eric Botcazou  <ebotcazou@adacore.com>
125         * tree-switch-conversion.h (switch_decision_tree::emit_case_nodes): Add
126         location_t parameter.
127         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
128         (switch_decision_tree::do_jump_if_equal): Likewise.
129         * tree-switch-conversion.c (switch_decision_tree::emit): Pass location
130         of switch statement to emit_case_nodes.
131         (switch_decision_tree::emit_cmp_and_jump_insns): Add LOC parameter and
132         set it on the newly built GIMPLE comparison statement.
133         (switch_decision_tree::do_jump_if_equal): Likewise.
134         (switch_decision_tree::emit_case_nodes): Add LOC parameter and pass it
135         in calls to do_jump_if_equal as well as recursive calls.
137 2018-11-16  Andi Kleen  <ak@linux.intel.com>
139         * config/i386/i386.md: Allow memory operands to ptwrite.
141 2018-11-16  Michael Matz  <matz@suse.de>
143         * system.h (PRsa): New macro.
144         (SIZE_AMOUNT): Cast number to uint64_t.
145         * alloc-pool.h (pool_usage::dump): Don't use %zu but PRsa.
146         (pool_usage::dump_footer): Likewise and also use PRIu64.
147         * bitmap.h (bitmap_usage::dump): Likewise.
148         * ggc-common.c (ggc_usage::dump): Likewise.
149         * ggc-page.c (ggc_print_statistics): Likewise.
150         * input.c (dump_line_table_statistics): Likewise.
151         * mem-stats.h (mem_usage::dump): Likewise.
152         (mem_usage::dump_footer): Likewise.
153         * rtl.c (dump_rtx_statistics): Likewise.
154         * tree-cfg.c (dump_cfg_stats): Likewise.
155         * tree-dfa.c (dump_dfa_stats): Likewise.
156         * tree-phinodes.c (phinodes_print_statistics): Likewise.
157         * tree-ssanames (ssanames_print_statistics): Likewise.
158         * vec.c (vec_usage::dump): Likewise.
159         (vec_usage::dump_footer): Likewise.
161 2018-11-16  Richard Biener  <rguenther@suse.de>
163         PR tree-optimization/88011
164         * tree-vrp.c (extract_range_from_binary_expr): Fix error in
165         replacing set_value_range_to_undefined and
166         set_value_range_to_varying with method calls.
168 2018-11-16  Ilya Leoshkevich  <iii@linux.ibm.com>
170         * config/s390/s390.md
171         (*r<noxa>sbg_<mode>_srl_bitmask): Do not delegate arithmetic to
172         assembler.
173         (*r<noxa>sbg_<mode>_sll): Likewise.
174         (*r<noxa>sbg_<mode>_srl): Likewise.
175         (*r<noxa>sbg_sidi_srl): New pattern.
177 2018-11-16  Jerome Lambourg  <lambourg@adacore.com>
179         * config/vxworks.h (TARGET_VXWORKS_HAVE_CTORS_DTORS): New macro.
180         Default to TARGET_VXWORKS_RTP.
181         (SUPPORTS_INIT_PRIORITY): Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
182         of TARGET_VXWORKS_RTP.
183         * config/vxworksae.h: Also define TARGET_VXWORKS_HAVE_CTORS_DTORS.
184         * config/vxworks.c: Use TARGET_VXWORKS_HAVE_CTORS_DTORS instead
185         of TARGET_VXWORKS_RTP to set targetm.have_ctors_dtors.
187 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
189         PR c++/25759
190         * doc/extend.texi (Common Type Attributes): Make it explicit
191         that attribute "packed" can apply to C++ classes.
193 2018-11-15  Martin Sebor  <msebor@redhat.com>
195         PR c++/87541
196         PR c++/87542
197         * tree.c (type_argument_type): New function.
198         * tree.h (type_argument_type): Declare it.
199         * gcc/doc/extend.texi (alloc_align): Update and clarify.
200         (alloc_size, nonnull, sentinel): Same.
202 2018-11-15  Andrew Stubbs  <ams@codesourcery.com>
203             Kwok Cheung Yeung  <kcy@codesourcery.com>
205         * tree-vect-stmts.c (vectorizable_store): Don't ICE when
206         int_mode_for_size fails.
207         (vectorizable_load): Likewise.
209 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
211         * doc/ux.texi (Group logically-related diagnostics): Move
212         discussion of auto_diagnostic_group into this new subsection.
213         Give an example of where this grouping is used.
215 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
217         PR other/19165
218         * Makefile.in (OBJS): Move json.o to...
219         (OBJS-libcommon): ...here and add diagnostic-format-json.o.
220         * common.opt (fdiagnostics-format=): New option.
221         (diagnostics_output_format): New enum.
222         * diagnostic-format-json.cc: New file.
223         * diagnostic.c (default_diagnostic_final_cb): New function, taken
224         from start of diagnostic_finish.
225         (diagnostic_initialize): Initialize final_cb to
226         default_diagnostic_final_cb.
227         (diagnostic_finish): Move "being treated as errors" messages to
228         default_diagnostic_final_cb.  Call any final_cb.
229         (default_diagnostic_finalizer): Add diagnostic_t param.
230         (diagnostic_report_diagnostic): Pass "orig_diag_kind" to
231         diagnostic_finalizer callback.
232         * diagnostic.h (enum diagnostics_output_format): New enum.
233         (diagnostic_finalizer_fn): Reimplement, adding diagnostic_t param.
234         (struct diagnostic_context): Add "final_cb".
235         (default_diagnostic_finalizer): Add diagnostic_t param.
236         (diagnostic_output_format_init): New decl.
237         * doc/invoke.texi (-fdiagnostics-format): New option.
238         * dwarf2out.c (gen_producer_string): Ignore
239         OPT_fdiagnostics_format_.
240         * gcc.c (driver_handle_option): Handle OPT_fdiagnostics_format_.
241         * lto-wrapper.c (append_diag_options): Ignore it.
242         * opts.c (common_handle_option): Handle it.
244 2018-11-15  David Malcolm  <dmalcolm@redhat.com>
246         PR tree-optimization/88015
247         * graphite-isl-ast-to-gimple.c
248         (translate_isl_ast_to_gimple::scop_to_isl_ast): Add missing check
249         for dump_enabled_p.
250         * graphite-sese-to-poly.c (build_poly_scop): Likewise.
252 2018-11-15  Richard Biener  <rguenther@suse.de>
254         PR middle-end/88029
255         * gimple.c (gimple_call_flags): Union flags from decl, type
256         and call fntype.
257         * trans-mem.c (is_tm_pure_call): Simplify.
259 2018-11-15  Richard Biener  <rguenther@suse.de>
261         PR tree-optimization/88031
262         * tree-vect-loop.c (vectorizable_reduction): Move check
263         for multiple types earlier so we get the expected dump.
264         Simplify calls to vectorizable_condition.
265         * tree-vect-stmts.h (vectorizable_condition): Update prototype.
266         * tree-vect-stmts.c (vectorizable_condition): Instead of
267         reduc_def and reduc_index take just a flag.  Simplify
268         code-generation now that we can rely on the defs being set up.
269         (vectorizable_comparison): Remove unused argument.
271 2018-11-15  Richard Biener  <rguenther@suse.de>
273         PR tree-optimization/88030
274         * tree-complex.c (need_eh_cleanup): New global.
275         (update_complex_assignment): Mark blocks that need EH update.
276         (expand_complex_comparison): Likewise.
277         (tree_lower_complex): Allocate and deallocate need_eh_cleanup,
278         perform EH cleanup and schedule CFG cleanup if that did anything.
280 2018-11-15  Jakub Jelinek  <jakub@redhat.com>
282         PR rtl-optimization/88018
283         * cfgrtl.c (fixup_abnormal_edges): Guard moving insns to fallthru edge
284         on the presence of fallthru edge, rather than if it is a USE or not.
286 2018-11-15  Richard Biener  <rguenther@suse.de>
288         PR middle-end/87917
289         * tree-data-ref.c (analyze_miv_subscript): Guard calls to
290         analyze_subscript_affine_affine properly.
292 2018-11-15  Xianmiao Qu  <xianmiao_qu@c-sky.com>
294         * config/csky/csky-linux-elf.h (LINUX_DYNAMIC_LINKER): Remove.
295         (GLIBC_DYNAMIC_LINKER): Define.
296         (LINUX_TARGET_LINK_SPEC): Update the dynamic linker's name.
298 2018-11-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
300         PR tree-optimization/84648
301         * tree-ssa-loop-niter.c (adjust_cond_for_loop_until_wrap): New.
302         (number_of_iterations_cond): Adjust exit cond for loop-until-wrap case
303         by calling adjust_cond_for_loop_until_wrap.
305 2018-11-15  Sandra Loosemore  <sandra@codesourcery.com>
307         PR other/56334
308         * doc/extend.texi (Common Function Attributes): Clarify linker
309         restrictions on "aligned" attribute.
310         (Common Variable Attributes): Likewise.  Mention that linker
311         restrictions don't apply to stack-allocated variables.
313 2018-11-15  Joern Rennecke  <joern.rennecke@riscy-ip.com>
315         * gensupport.c (add_predicate_code): Properly handle ZERO_EXTRACT
316         as an lvalue.
318 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
320         PR lto/55102
321         PR lto/56700
322         * doc/invoke.texi (Optimize Options): Remove bad example about
323         interaction between -flto and -O.  Replace it with a note that
324         you need to compile with -O and not just link.  Copy-edit -flto
325         discussion to reduce verbiage and improve flow.
327 2018-11-14  Sandra Loosemore  <sandra@codesourcery.com>
329         PR middle-end/59658
330         * doc/invoke.texi (Optimize Options): Clarify that -O0 and -Og
331         also suppress many optimizations.  Alphabetize option lists for
332         -O1, -O2, and -Os.  Add list of options disabled with -Og, and
333         correct documentation for those options to say that.
334         * opts.c (default_options_table): Sort table by level and option
335         name, to make it easier to correlate to the manual.
337 2018-11-14  Uros Bizjak  <ubizjak@gmail.com>
339         * config/i386/i386.c (ix86_print_operand_address_as): Simplify
340         printing of the names of segment registers.
342 2018-11-14  Nathan Sidwell  <nathan@acm.org>
344         PR debug/88006
345         PR debug/87462
346         * dwarf2out.c (dwarf2out_finish): Apply resolve_addr to comdat
347         type list.
349 2018-11-14  David Malcolm  <dmalcolm@redhat.com>
351         * Makefile.in (CFLAGS-optinfo-emit-json.o): Add $(ZLIBINC).
353 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
355         P1236R1 - Signed integers are two's complement
356         * doc/invoke.texi (Wshift-overflow): Adjust documentation for c++2a.
358         PR bootstrap/86739
359         * hash-map.h (hash_map::iterator::reference_pair): New class.
360         (hash_map::iterator::operator*): Return it rather than std::pair.
362 2018-11-14  Jeff Law  <law@redhat.com>
364         * optabs.c (expand_binop): Pass INT_MODE to operand_subword_force
365         iff the operand is a constant.
367 2018-11-14  Aldy Hernandez  <aldyh@redhat.com>
369         * gimple-ssa-evrp-analyze.c
370         (evrp_range_analyzer::record_ranges_from_incoming_edge): Rename
371         ignore_equivs_equal_p to equal_p.
372         * ipa-cp.c (meet_with_1): Use equal_p instead of
373         ignore_equivs_equal_p.
374         * ipa-prop.c (ipa_vr_ggc_hash_traits::equal): Same.
375         * tree-vrp.c (value_range::ignore_equivs_equal_p): Remove.
376         (value_range::operator==): Remove.
377         (value_range::operator!=): Remove.
378         (vrp_prop::visit_stmt): Use equal_p.
379         * tree-vrp.h (value_range): Remove operator==, operator!=,
380         ignore_equivs_equal_p.
381         * vr-values.c (update_value_range): Use equal_p.
383 2018-11-14  Michael Matz  <matz@suse.de>
385         PR middle-end/86575
386         * gimplify.c (collect_fallthrough_labels): Add new argument,
387         return location via that, don't modify statements.
388         (warn_implicit_fallthrough_r): Adjust call, don't use
389         statement location directly.
391 2018-11-14  Richard Biener  <rguenther@suse.de>
393         PR middle-end/87985
394         * tree-data-ref.c (split_constant_offset): Add wrapper
395         allocating a cache hash-map.
396         (split_constant_offset_1): Cache results of expanding
397         expressions from SSA def stmts.
399 2018-11-14  Richard Biener  <rguenther@suse.de>
401         PR middle-end/88021
402         * tree-data-ref.c (lambda_matrix_row_add): Change const1 argument
403         to lambda_int.
404         (lambda_vector_mult_const): Likewise.
405         (lambda_matrix_right_hermite): Use lambda_int temporaries.
407 2018-11-14  Wilco Dijkstra  <wdijkstr@arm.com>
408             Jackson Woodruff  <jackson.woodruff@arm.com>
410         PR 71026/tree-optimization
411         * match.pd: Simplify floating point comparisons.
413 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
415         PR rtl-optimization/87817
416         * config/i386/i386.c (ix86_fold_builtin): For _bzhi_u{32,64} if
417         last argument has low 8 bits clear, fold to 0.
419 2018-11-14  Iain Sandoe  <iain@sandoe.co.uk>
421         * tree-vect-data-refs.c (vect_can_force_dr_alignment_p): Cast
422         MAX_OFILE_ALIGNMENT as needed.
424 2018-11-14  Richard Biener  <rguenther@suse.de>
426         PR tree-optimization/88019
427         * graphite-isl-ast-to-gimple.c (ternary_op_to_tree): Rewrite
428         COND_EXPR arguments to non-trapping overflow.
430 2018-11-14  Claudiu Zissulescu  <claziss@synopsys.com>
432         * config/arc/arc.h (reg_class): Reorder registers classes, remove
433         unused register classes.
434         (REG_CLASS_NAMES): Likewise.
435         (REG_CLASS_CONTENTS): Likewise.
436         (FIXED_REGISTERS): Make lp_count fixed.
437         (BASE_REG_CLASS): Remove ACC16_BASE_REGS reference.
438         (PROGRAM_COUNTER_REGNO): Remove.
439         * config/arc/arc.c (arc_conditional_register_usage): Remove unused
440         register classes, use constants for register numbers, remove
441         reg_class_contents references.
442         (arc_process_double_reg_moves): Add asserts.
443         (arc_secondary_reload): Remove LPCOUNT_REG reference, use
444         lra_in_progress predicate.
445         (arc_init_reg_tables): Remove unused register classes.
446         (arc_register_move_cost): Likewise.
447         (arc_preferred_reload_class): Likewise.
448         (hwloop_optimize): Update rtx patterns involving lp_count register.
449         (arc_return_address_register): Rename ILINK1, INLINK2 regnums macros.
450         * config/arc/constraints.md ("c"): Choose between GENERAL_REGS and
451         CHEAP_CORE_REGS.  Former one will be used for LRA.
452         ("Rac"): Choose between GENERAL_REGS and ALL_CORE_REGS.  Former
453         one will be used for LRA.
454         ("w"): Choose between GENERAL_REGS and WRITABLE_CORE_REGS.  Former
455         one will be used for LRA.
456         ("W"): Choose between GENERAL_REGS and MPY_WRITABLE_CORE_REGS.
457         Former one will be used for LRA.
458         ("f"): Delete constraint.
459         ("k"): Likewise.
460         ("e"): Likewise.
461         ("l"): Change it from register constraint to constraint.
462         * config/arc/arc.md (movqi_insn): Remove unsed lp_count constraints.
463         (movhi_insn): Likewise.
464         (movsi_insn): Update pattern.
465         (arc_lp): Likewise.
466         (dbnz): Likewise.
467         (stack_tie): Remove 'b' constraint letter.
468         (R4_REG): Define.
469         (R9_REG, R15_REG, R16_REG, R25_REG): Likewise.
470         (R32_REG, R40_REG, R41_REG, R42_REG, R43_REG, R44_REG): Likewise.
471         (R57_REG, R59_REG, PCL_REG): Likewise.
472         (ILINK1_REGNUM): Renamed to ILINK1_REG.
473         (ILINK2_REGNUM): Renamed to ILINK2_REG.
474         (Rgp): Remove.
475         (SP_REGS): Likewise.
476         (Rcw): Remove unused reg classes.
477         * config/arc/predicates.md (dest_reg_operand): Just default on
478         register_operand predicate.
479         (mpy_dest_reg_operand): Likewise.
480         (move_dest_operand): Use macros instead of constants.
481         * config/arc/arc.opt (mlra): Switch to lra as default.
483 2018-11-14  Richard Biener  <rguenther@suse.de>
485         PR tree-optimization/87974
486         * tree-vect-loop.c (vectorizable_reduction): When computing
487         the vectorized reduction PHI vector type ignore constant
488         and external defs.
490 2018-11-14  Jakub Jelinek  <jakub@redhat.com>
492         PR tree-optimization/87977
493         * tree-ssa-math-opts.c (optimize_recip_sqrt): Don't reuse division
494         stmt, build a new one and replace the old one with it.  Formatting fix.
495         Call release_ssa_name (x) if !has_other_use and !delete_div.
496         (pass_cse_reciprocals::execute): Before calling optimize_recip_sqrt
497         verify lhs of stmt is still def.
499 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
501         PR rtl-optimization/87507
502         * lower-subreg.c (operand_for_swap_move_operator): New function.
503         (simple_move): Strip simple operators.
504         (find_pseudo_copy): Likewise.
505         (resolve_operand_for_swap_move_operator): New function.
506         (resolve_simple_move): Strip simple operators and swap operands.
508 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
510         PR tree-optimization/87898
511         * omp-simd-clone.c (ipa_simd_modify_stmt_ops): Formatting fix.
512         (ipa_simd_modify_function_body): Remove debug stmts where the first
513         argument was changed into a non-decl.
515 2018-11-13  Peter Bergner  <bergner@linux.ibm.com>
517         PR rtl-optimization/87899
518         * lra-lives.c (start_living): Update white space in comment.
519         (enum point_type): New.
520         (sparseset_contains_pseudos_p): New function.
521         (update_pseudo_point): Likewise.
522         (make_hard_regno_live): Use HARD_REGISTER_NUM_P macro.
523         (make_hard_regno_dead): Likewise.  Remove ignore_reg_for_conflicts
524         handling.  Move early exit after adding conflicts.
525         (mark_pseudo_live): Use HARD_REGISTER_NUM_P macro.  Add early exit
526         if regno is already live.  Remove all handling of program points.
527         (mark_pseudo_dead): Use HARD_REGISTER_NUM_P macro.  Add early exit
528         after adding conflicts.  Remove all handling of program points and
529         ignore_reg_for_conflicts.
530         (mark_regno_live): Use HARD_REGISTER_NUM_P macro.  Remove return value
531         and do not guard call to mark_pseudo_live.
532         (mark_regno_dead): Use HARD_REGISTER_NUM_P macro.  Remove return value
533         and do not guard call to mark_pseudo_dead.
534         (check_pseudos_live_through_calls): Use HARD_REGISTER_NUM_P macro.
535         (process_bb_lives): Use HARD_REGISTER_NUM_P and HARD_REGISTER_P macros.
536         Use new function update_pseudo_point.  Handle register copies by
537         removing the source register from the live set.  Handle INOUT operands.
538         Update to the next program point using the unused_set, dead_set and
539         start_dying sets.
540         (lra_create_live_ranges_1): Use HARD_REGISTER_NUM_P macro.
542 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
544         * builtins.c: Replace "source_location" with "location_t".
545         * diagnostic-show-locus.c: Likewise.
546         * diagnostic.c: Likewise.
547         * dumpfile.c: Likewise.
548         * gcc-rich-location.h: Likewise.
549         * genmatch.c: Likewise.
550         * gimple.h: Likewise.
551         * gimplify.c: Likewise.
552         * input.c: Likewise.
553         * input.h: Likewise.  Eliminate the typedef.
554         * omp-expand.c: Likewise.
555         * selftest.h: Likewise.
556         * substring-locations.h (get_source_location_for_substring):
557         Rename to...
558         (get_location_within_string): ...this.
559         * tree-cfg.c: Replace "source_location" with "location_t".
560         * tree-cfgcleanup.c: Likewise.
561         * tree-diagnostic.c: Likewise.
562         * tree-into-ssa.c: Likewise.
563         * tree-outof-ssa.c: Likewise.
564         * tree-parloops.c: Likewise.
565         * tree-phinodes.c: Likewise.
566         * tree-phinodes.h: Likewise.
567         * tree-ssa-loop-ivopts.c: Likewise.
568         * tree-ssa-loop-manip.c: Likewise.
569         * tree-ssa-phiopt.c: Likewise.
570         * tree-ssa-phiprop.c: Likewise.
571         * tree-ssa-threadupdate.c: Likewise.
572         * tree-ssa.c: Likewise.
573         * tree-ssa.h: Likewise.
574         * tree-vect-loop-manip.c: Likewise.
576 2018-11-13  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
578         * doc/extend.texi: Fix typo in the weakref description.
580 2018-11-13  Richard Biener  <rguenther@suse.de>
582         PR tree-optimization/86991
583         * tree-vect-loop.c (vect_is_slp_reduction): Delay reduction
584         group building until we have successfully detected the SLP
585         reduction.
586         (vect_is_simple_reduction): Remove fixup code here.
588 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
590         * dumpfile.c (VERIFY_DUMP_ENABLED_P): New macro.
591         (dump_gimple_stmt): Use it.
592         (dump_gimple_stmt_loc): Likewise.
593         (dump_gimple_expr): Likewise.
594         (dump_gimple_expr_loc): Likewise.
595         (dump_generic_expr): Likewise.
596         (dump_generic_expr_loc): Likewise.
597         (dump_printf): Likewise.
598         (dump_printf_loc): Likewise.
599         (dump_dec): Likewise.
600         (dump_dec): Likewise.
601         (dump_hex): Likewise.
602         (dump_symtab_node): Likewise.
603         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
604         Guard dump call with dump_enabled_p.
605         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl): Likewise.
606         * graphite-optimize-isl.c (optimize_isl): Likewise.
607         * graphite.c (graphite_transform_loops): Likewise.
608         * tree-loop-distribution.c (pass_loop_distribution::execute): Likewise.
609         * tree-parloops.c (parallelize_loops): Likewise.
610         * tree-ssa-loop-niter.c (number_of_iterations_exit): Likewise.
611         * tree-vect-data-refs.c (vect_analyze_group_access_1): Likewise.
612         (vect_prune_runtime_alias_test_list): Likewise.
613         * tree-vect-loop.c (vect_update_vf_for_slp): Likewise.
614         (vect_estimate_min_profitable_iters): Likewise.
615         * tree-vect-slp.c (vect_record_max_nunits): Likewise.
616         (vect_build_slp_tree_2): Likewise.
617         (vect_supported_load_permutation_p): Likewise.
618         (vect_slp_analyze_operations): Likewise.
619         (vect_slp_analyze_bb_1): Likewise.
620         (vect_slp_bb): Likewise.
621         * tree-vect-stmts.c (vect_analyze_stmt): Likewise.
622         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.
623         (pass_slp_vectorize::execute): Likewise.
624         (increase_alignment): Likewise.
626 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
628         PR ipa/87955
629         * ipa-inline.c (report_inline_failed_reason): Guard calls to
630         cl_target_option_print_diff and cl_optimization_print_diff with
631         if (dump_file).
633 2018-11-13  David Malcolm  <dmalcolm@redhat.com>
635         * doc/invoke.texi (-fsave-optimization-record): Note that the
636         output is compressed.
637         * optinfo-emit-json.cc: Include <zlib.h>.
638         (optrecord_json_writer::write): Compress the output.
640 2018-11-13  Aldy Hernandez  <aldyh@redhat.com>
642         * tree-vrp.c (value_range_base::dump): Dump type.
643         Do not use INF nomenclature for 1-bit types.
644         (dump_value_range): Group all variants to common dumping code.
645         (debug): New overloaded functions for value_ranges.
646         (value_range_base::dump): Remove no argument version.
647         (value_range::dump): Same.
649 2018-11-13  Richard Biener  <rguenther@suse.de>
651         PR tree-optimization/87931
652         * tree-vect-loop.c (vect_is_simple_reduction): Restrict
653         nested cycles we support to latch computations vectorizable_reduction
654         handles.
656 2018-11-13  Martin Liska  <mliska@suse.cz>
658         PR tree-optimization/87885
659         * cfghooks.c (account_profile_record): Rename
660         to ...
661         (profile_record_check_consistency): ... this.
662         Calculate missing num_mismatched_freq_in.
663         (profile_record_account_profile): New function
664         that calculates time and size of a function.
665         * cfghooks.h (struct profile_record): Remove
666         all tuples.
667         (struct cfg_hooks): Remove after_pass flag.
668         (account_profile_record): Rename to ...
669         (profile_record_check_consistency): ... this.
670         (profile_record_account_profile): New.
671         * cfgrtl.c (rtl_account_profile_record): Remove
672         after_pass flag.
673         * passes.c (check_profile_consistency): Do only
674         checking.
675         (account_profile): Calculate size and time of
676         function only.
677         (pass_manager::dump_profile_report): Reformat
678         output.
679         (execute_one_ipa_transform_pass): Call
680         consistency check before clean upand call account_profile
681         after a clean up is done.
682         (execute_one_pass): Call check_profile_consistency and
683         account_profile instead of using after_pass flag..
684         * tree-cfg.c (gimple_account_profile_record): Likewise.
686 2018-11-13  Martin Liska  <mliska@suse.cz>
688         PR sanitizer/87930
689         * config/i386/i386.c (ix86_option_override_internal): Error
690         about usage -mabi=ms and -fsanitize={,kernel-}address.
692 2018-11-13  Andre Vieira  <andre.simoesdiasvieira@arm.com>
694         * config/aarch64/aarch64.c
695         (aarch64_vectorize_preferred_vector_alignment): Change return type to
696         poly_uint64.
697         (aarch64_simd_vector_alignment_reachable): Adapt to preferred vector
698         alignment being a poly int.
699         * doc/tm.texi (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Change
700         return type to poly_uint64.
701         * target.def (default_preferred_vector_alignment): Likewise.
702         * targhooks.c (default_preferred_vector_alignment): Likewise.
703         * targhooks.h (default_preferred_vector_alignment): Likewise.
704         * tree-vect-data-refs.c (vect_calculate_target_alignment): Likewise.
705         (vect_compute_data_ref_alignment): Adapt to vector alignment being a
706         poly int.
707         (vect_update_misalignment_for_peel): Likewise.
708         (vect_enhance_data_refs_alignment): Likewise.
709         (vect_find_same_alignment_drs): Likewise.
710         (vect_duplicate_ssa_name_ptr_info): Likewise.
711         (vect_setup_realignment): Likewise.
712         (vect_can_force_dr_alignment_p): Change alignment parameter type to
713         poly_uint64.
714         * tree-vect-loop-manip.c (get_misalign_in_elems): Learn to construct a
715         mask with a compile time variable vector alignment.
716         (vect_gen_prolog_loop_niters): Adapt to vector alignment being a poly
717         int.
718         (vect_do_peeling): Exit early if vector alignment is not constant.
719         * tree-vect-stmts.c (ensure_base_align): Adapt to vector alignment being
720         a poly int.
721         (vectorizable_store): Likewise.
722         (vectorizable_load): Likweise.
723         * tree-vectorizer.h (struct dr_vec_info): Make target_alignment field a
724         poly_uint64.
725         (vect_known_alignment_in_bytes): Adapt to vector alignment being a
726         poly int.
727         (vect_can_force_dr_alignment_p): Change alignment parameter type to
728         poly_uint64.
730 2018-11-13  Richard Biener  <rguenther@suse.de>
732         PR tree-optimization/87962
733         * tree-vect-loop.c (vect_is_simple_reduction): More reliably
734         detect outer reduction for disqualifying in-loop uses.
736 2018-11-13  Richard Biener  <rguenther@suse.de>
738         PR tree-optimization/87967
739         * tree-vect-loop.c (vect_transform_loop): Also copy PHIs
740         for constants for the scalar loop.
742 2018-11-13  Alan Modra  <amodra@gmail.com>
744         * config/rs6000/linux64.h (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Exclude
745         integer constants when -mcmodel=medium.
747 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
749         * config/arc/arc.h (FUNCTION_PROFILER): Redefine to empty.
750         * config/arc/elf.h (PROFILE_HOOK): Define.
751         * config/arc/linux.h (PROFILE_HOOK): Likewise.
753 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
755         * config/arc/arc.c (hwloop_optimize): Bailout when detecting a
756         jump table data in the text section.
758 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
760         * config/arc/arc.c (arc_eh_return_address_location): Repurpose it
761         to fit the eh_return pattern.
762         * config/arc/arc.md (eh_return): Define.
763         (VUNSPEC_ARC_EH_RETURN): Likewise.
764         * config/arc/arc-protos.h (arc_eh_return_address_location): Match
765         new implementation.
766         * config/arc/arc.h (EH_RETURN_HANDLER_RTX): Remove it.
769 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
771         * common/config/arc/arc-common.c (arc_option_optimization_table):
772         Millicode optimization is default on for size optimizations.
773         * config/arc/arc-protos.h (arc_check_multi): New function.
774         * config/arc/arc.c (RTX_OK_FOR_OFFSET_P): Rearange.
775         (ENTER_LEAVE_START_REG): Define.
776         (ENTER_LEAVE_END_REG): Likewise.
777         (arc_override_options): Disable millicode when long calls option
778         is on.
779         (arc_frame_info): Change it from int to bool.
780         (arc_compute_frame_size): Clean up.
781         (arc_save_restore): Remove.
782         (frame_save_reg): New function.
783         (frame_restore_reg): Likewise.
784         (arc_enter_leave_p): Likewise.
785         (arc_save_callee_saves): Likewise.
786         (arc_restore_callee_saves): Likewise.
787         (arc_save_callee_enter): Likewise.
788         (arc_restore_callee_leave): Likewise.
789         (arc_save_callee_milli): Likewise.
790         (arc_restore_callee_milli): Likewise.
791         (arc_expand_prologue): Reimplement to emit enter/leave
792         instructions.
793         (arc_expand_epilogue): Likewise.
794         (arc_check_multi): New function.
795         * config/arc/arc.md (push_multi_fp): New pattern.
796         (push_multi_fp_blink): Likewise.
797         (pop_multi_fp): Likewise.
798         (pop_multi_fp_blink): Likewise.
799         (pop_multi_fp_ret): Likewise.
800         (pop_multi_fp_blink_ret): Likewise.
801         * config/arc/arc.opt (mmillicode): Update option.
802         (mcode-density-frame): New option.
803         * config/arc/predicates.md (push_multi_operand): New predicate.
804         (pop_multi_operand): Likewise.
805         * doc/invoke.texi (ARC): Update ARC options information.
808 2018-11-13  Claudiu Zissulescu  <claziss@synopsys.com>
810         * config/arc/arc-protos.h (gen_operands_ldd_std): Add.
811         * config/arc/arc.c (operands_ok_ldd_std): New function.
812         (mem_ok_for_ldd_std): Likewise.
813         (gen_operands_ldd_std): Likewise.
814         * config/arc/arc.md: Add peephole2 rules for std/ldd.
816 2018-11-13  Eric Botcazou  <ebotcazou@adacore.com>
818         * toplev.c (output_stack_usage): Turn test on flag_stack_usage into
819         test on stack_usage_file.
820         (lang_dependent_init): Do not open the .su file if generating LTO.
822 2018-11-13  Jakub Jelinek  <jakub@redhat.com>
824         PR rtl-optimization/87918
825         * simplify-rtx.c (simplify_merge_mask): For COMPARISON_P, use
826         simplify_gen_relational rather than simplify_gen_binary.
828 2018-11-13  Richard Biener  <rguenther@suse.de>
830         * tree-ssanames.h (set_range_info): Use value_range_base.
831         (get_range_info): Likewise.
832         * tree-ssanames.c (set_range_info): Likewise.
833         (get_range_info): Likewise.
834         * tree-vrp.c (value_range_base::union_helper): Split
835         out common parts of value_range[_base]::union_.
836         (value_range_base::union_): Update.
837         (value_range::union_): Likewise.
838         (determine_value_range_1): Use value_range_base.
839         (determine_value_range): Likewise.
840         * tree-vrp.h (value_range_base::union_helper): Move ...
841         (value_range::union_helper): ... from here.
843 2018-11-13  Alan Modra  <amodra@gmail.com>
845         * config/rs6000/rs6000.c (rs6000_secondary_reload_inner): Negate
846         offset for PRE_DEC.
847         (rs6000_secondary_reload_gpr): Don't call find_replacement.
849 2018-11-13  Sandra Loosemore  <sandra@codesourcery.com>
851         PR middle-end/59634
852         * doc/invoke.texi (Optimize Options): Clarify that the
853         l1-cache-line-size, l1-cache-size, and l2-cache-size parameters
854         apply to data cache size.
856 2018-11-13  Alan Modra  <amodra@gmail.com>
858         * config/rs6000/predicates.md (logical_const_operand),
859         (logical_operand): Correct comment.
860         * config/rs6000/rs6000.c (rs6000_emit_prologue): Comment fix.
862 2018-11-13  Alan Modra  <amodra@gmail.com>
864         * config/rs6000/rs6000.md (addsi3_high): Prefix with '*'.
866 2018-11-13  Alan Modra  <amodra@gmail.com>
868         * gcc/doc/md.texi (Machine Constraints): Correct rs6000 h constraint
869         description.
870         * config/rs6000/rs6000.md (movsi_internal1): Delete MT%0 case
871         covered by alternative.
872         (movcc_internal1): Ignore h for register preference.
873         (mov<mode>_hardfloat64): Likewise.
874         (mov<mode>_softfloat): Ignore c, l, h for register preference.
876 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
878         PR preprocessor/47823
879         * doc/cpp.texi (Alternatives to Wrapper #ifndef): Move #pragma once
880         documentation to...
881         (Pragmas): ...here.
882         * doc/extend.texi (Pragmas): Note additional pragmas documented
883         in the CPP manual.
885 2018-11-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
887         PR middle-end/86677
888         PR middle-end/87528
889         * tree-scalar-evolution.c (expression_expensive_p): Make
890         BUILTIN POPCOUNT as expensive when backend does not define it.
892 2018-11-12  Fredrik Noring  <noring@nocrew.org>
894         * config.gcc: Update with-llsc defaults for MIPS r5900.
896 2018-11-12  Martin Liska  <mliska@suse.cz>
898         PR gcov-profile/87442
899         * common.opt: Add -fprofile-filter-files and -fprofile-exclude-files
900         options.
901         * doc/invoke.texi: Document them.
902         * tree-profile.c (parse_profile_filter): New.
903         (parse_profile_file_filtering): Likewise.
904         (release_profile_file_filtering): Likewise.
905         (include_source_file_for_profile): Likewise.
906         (tree_profiling): Filter source files based on the
907         newly added options.
909 2018-11-12  Martin Liska  <mliska@suse.cz>
911         PR target/87903
912         * doc/extend.texi: Add missing values for __builtin_cpu_is and
913         __builtin_cpu_supports for x86 target.
915 2018-11-12  Sandra Loosemore  <sandra@codesourcery.com>
917         PR middle-end/21110
918         * doc/rtl.texi (Constants): Clarify that mode of "high" doesn't
919         have to be Pmode.
920         (Arithmetic): Likewise for "lo_sum".
922 2018-11-12  Renlin Li  <renlin.li@arm.com>
924         PR target/87815
925         * dse.c (get_stored_val): Add check for compile-time constantness
926         of gap.
928 2018-11-12  Sudakshina Das  <sudi.das@arm.com>
930         * config/arm/arm-cpus.in (armv8_5, sb, predres): New features.
931         (ARMv8_5a): New fgroup.
932         (armv8.5-a): New arch.
933         (armv8-a, armv8.1-a, armv8.2-a, armv8.3-a, armv8.4-a): New options
934         sb and predres.
935         * config/arm/arm-tables.opt: Regenerate.
936         * config/arm/t-aprofile: Add matching rules for -march=armv8.5-a
937         * config/arm/t-arm-elf (all_v8_archs): Add armv8.5-a.
938         * config/arm/t-multilib (v8_5_a_simd_variants): New variable.
939         Add matching rules for -march=armv8.5-a and extensions.
940         * doc/invoke.texi (ARM options): Document -march=armv8.5-a.
941         Add sb and predres to all armv8-a except armv8.5-a.
943 2018-11-12  Richard Biener  <rguenther@suse.de>
945         * tree-vrp.h (value_range[_base]::set): Make public.  Provide
946         overload for single value.
947         (value_range[_base]::set_nonnull): New.
948         (value_range[_base]::set_null): Likewise.
949         (value_range): Document bitmap copying behavior, mark
950         copy constructor and assignment operator deleted.
951         (value_range::move): New.
952         (value_range::set_and_canonicalize): Default bitmap to zero.
953         (set_value_range_to_nonnull): Remove.
954         (set_value_range_to_null): Likewise.
955         (set_value_range): Likewise.
956         (set_value_range_to_value): Likewise.
957         (extract_range_from_unary_expr): Work on value_range_base.
958         (extract_range_from_binary_expr_1): Likewise.  Rename to...
959         (extract_range_from_binary_expr): ... this.
960         * tree-vrp.c (value_range::update): Clear equiv bitmap
961         if required.
962         (value_range::move): New, move equiv bitmap.
963         (value_range_base::set_undefined): Avoid assignment.
964         (value_range::set_undefined): Likewise.
965         (value_range_base::set_varying): Likewise.
966         (value_range::set_varying): Likewise.
967         (set_value_range): Remove.
968         (value_range_base::set): New overload for value.
969         (value_range::set): Likewise.
970         (set_value_range_to_nonnull): Remove.
971         (value_range_base::set_nonnull): New.
972         (value_range::set_nonnull): Likewise.
973         (set_value_range_to_null): Remove.
974         (value_range_base::set_null): New.
975         (value_range::set_null): Likewise.
976         (range_is_null): Work on value_range_base.
977         (range_is_nonnull): Likewise.
978         (ranges_from_anti_range): Likewise.
979         (extract_range_into_wide_ints): Likewise.
980         (extract_range_from_multiplicative_op): Likewise.
981         (extract_range_from_binary_expr): Likewise.  Update for API changes.
982         (extract_range_from_unary_expr): Likewise.  Remove OBJ_TYPE_REF
983         handling.
984         (value_range::intersect_helper): Avoid copy and assignment.
985         (value_range::union_helper): Likewise.
986         (determine_value_range_1): Adjust.
987         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
988         Avoid assignment by using move.
989         (evrp_range_analyzer::record_ranges_from_stmt): Avoid assignment.
990         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
991         Likewise.
992         * tree-ssanames.c (get_range_info): Likewise.
993         * vr-values.h (vr_values::get_vr_for_comparison): Adjust API.
994         * vr-values.c (vr_values::get_value_range): Adjust.
995         (vr_values::update_value_range): Likewise.
996         (symbolic_range_based_on_p): Work on value_range_base.
997         (vr_values::extract_range_from_binary_expr): Use value_range_base.
998         (vr_values::extract_range_from_unary_expr): Likewise.
999         (vr_values::extract_range_from_cond_expr): Avoid assignment.
1000         (vr_values::extract_range_from_comparison): Adjust.
1001         (vr_values::check_for_binary_op_overflow): Use value_range_base.
1002         (vr_values::extract_range_basic): Adjust.
1003         (vr_values::adjust_range_with_scev): Likewise.
1004         (vr_values::vrp_visit_assignment_or_call): Likewise.
1005         (vr_values::get_vr_for_comparison): Change API to avoid
1006         assignment and copy construction.
1007         (vr_values::compare_name_with_value): Adjust accordingly.
1008         (vr_values::compare_names): Likewise.
1009         (vr_values::extract_range_from_phi_node): Avoid assignment and
1010         bogus in-place modify of equiv bitmap.
1011         (vr_values::simplify_bit_ops_using_ranges): Use value_range_base.
1012         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Adjust
1013         for extract_range_from_unary_expr API change.
1014         * ipa-cp.c (ipa_vr_operation_and_type_effects): Likewise.
1016 2018-11-12  Eric Botcazou  <ebotcazou@adacore.com>
1018         * config/mcore/mcore.h (WORD_REGISTER_OPERATIONS): Remove duplicate.
1020 2018-11-12  Richard Biener  <rguenther@suse.de>
1022         * tree-vrp.h (value_range_base::symbolic_p,
1023         value_range_base::constant_p, value_range_base::zero_p,
1024         value_range_base::singleton_p): Move from value_range.
1025         (value_range::dump): Add.
1026         * gimple-ssa-evrp-analyze.c
1027         (evrp_range_analyzer::record_ranges_from_phis): Use set_varying.
1028         * ipa-cp.c (ipcp_vr_lattice::print): Use dump_value_range.
1029         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
1030         Use set_varying.
1031         * tree-vrp.c (value_range::symbolic_p): Move to value_range_base.
1032         (value_range::constant_p): Likewise.
1033         (value_range::singleton_p): Likewise.
1034         (value_range_base::dump): Add.
1035         (set_value_range_to_undefined): Remove.
1036         (set_value_range_to_varying): Likewise.
1037         (range_int_cst_p): Take value_range_base argument.
1038         (range_int_cst_singleton_p): Likewise.
1039         (value_range_constant_singleton): Likewise.
1040         (vrp_set_zero_nonzero_bits): Likewise.
1041         (extract_range_from_multiplicative_op): Use set_varying.
1042         (extract_range_from_binary_expr_1): Likewise. Use set_undefined.
1043         (extract_range_from_unary_expr): Likewise.
1044         (dump_value_range_base): Change to overload of dump_value_range.
1045         (vrp_prop::vrp_initialize): Use set_varying and set_undefined.
1046         (vrp_prop::visit_stmt): Likewise.
1047         (value_range::intersect_helper): Likewise.
1048         (value_range::union_helper): Likewise.
1049         (determine_value_range_1): Likewise.
1051 2018-11-12  Richard Biener  <rguenther@suse.de>
1053         * tree-vrp.c (set_value_range_to_nonnull): Clear equiv.
1054         (set_value_range_to_null): Likewise.
1055         * vr-values.c (vr_values::extract_range_from_comparison):
1056         Clear equiv for constant singleton ranges.
1058 2018-11-12 Wei Xiao <wei3.xiao@intel.com>
1060         * config/i386/sse.md: Combine VFIXUPIMM* patterns
1061         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1062         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1063         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Remove.
1064         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
1065         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
1066         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Remove.
1068 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1070         PR c/69502
1071         * doc/extend.texi (Common Type Attributes): For the align type
1072         attribute, copy language about decreasing alignment from the
1073         corresponding variable attribute.
1075 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1077         * config/csky/csky.md (*fpuv2_nmulsf3_1, *fpuv2_nmuldf3_1): Handle
1078         -frounding-math.
1080 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1082         PR c++/43105
1083         * doc/invoke.texi (C++ Dialect Options): Add warning about mixing
1084         -frtti and -fno-rtti code.
1086 2018-11-11  Sandra Loosemore  <sandra@codesourcery.com>
1088         PR c/26366
1089         * doc/extend.texi (Other Builtins): Document probability associated
1090         with __builtin_expect.
1092 2018-11-11  Uros Bizjak  <ubizjak@gmail.com>
1094         PR target/87928
1095         * config/i386/i386.h (STACK_BOUNDARY): Use TARGET_64BIT_MS_ABI
1096         instead of (TARGET_64BIT && ix86_abi == MS_ABI).
1097         * config/i386/darwin.h (STACK_BOUNDARY): Ditto.
1098         * config/i386/cygming.h (STACK_BOUNDARY): Remove.
1100 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1102         * config/csky/csky-linux-elf.h (CC1_SPEC): Support -profile.
1104 2018-11-11  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1106         * config/csky/csky.h (ASM_PREFERRED_EH_DATA_FORMAT): Define.
1108 2018-11-11  Richard Biener  <rguenther@suse.de>
1110         * tree-vrp.h (class value_range_base): New base class for
1111         value_range containing all but the m_equiv member.
1112         (dump_value_range_base): Add.
1113         (range_includes_zero_p): Work on value_range_base.
1114         * tree-vrp.c (value_range_base::set): Split out base handling
1115         from...
1116         (value_range::set): this.
1117         (value_range::set_equiv): New.
1118         (value_range_base::value_range_base): New constructors.
1119         (value_range_base::check): Split out base handling from...
1120         (value_range::check): this.
1121         (value_range::equal_p): Refactor in terms of
1122         ignore_equivs_equal_p which is now member of the base.
1123         (value_range_base::set_undefined): New.
1124         (value_range_base::set_varying): Likewise.
1125         (value_range_base::dump):Split out base handling from...
1126         (value_range::dump): this.
1127         (value_range_base::set_and_canonicalize): Split out base handling
1128         from...
1129         (value_range::set_and_canonicalize): this.
1130         (value_range_base::union_): New.
1131         * ipa-prop.h (struct ipa_jump_func): Use value_range_base *
1132         for m_vr.
1133         * ipa-cp.c (class ipcp_vr_lattice): Use value_range_base
1134         instead of value_range everywhere.
1135         (ipcp_vr_lattice::print): Use dump_value_range_base.
1136         (ipcp_vr_lattice::meet_with): Adjust.
1137         (ipcp_vr_lattice::meet_with_1): Likewise.
1138         (ipa_vr_operation_and_type_effects): Likewise.
1139         (propagate_vr_across_jump_function): Likewise.
1140         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Likewise.
1141         (ipa_get_value_range): Likewise.
1142         (ipa_set_jfunc_vr): Likewise.
1143         (ipa_compute_jump_functions_for_edge): Likewise.
1145 2018-11-10  Sandra Loosemore  <sandra@codesourcery.com>
1147         PR middle-end/65703
1148         * doc/invoke.texi (Optimize Options): Add @opindex entries
1149         for the positive forms of -fno-xxx and -mno-xxx options
1150         that were lacking them.
1152 2018-11-10  Segher Boessenkool  <segher@kernel.crashing.org>
1154         * combine.c (make_more_copies): Only make an intermediate copy if the
1155         dest of a move is a pseudo.
1157 2018-11-09  Maya Rashish  <coypu@sdf.org>
1159         PR target/87221
1160         * config/netbsd-elf.h (NETBSD_STARTFILE_SPEC): Use crtbeginS.o for PIE.
1161         (NETBSD_ENDFILE_SPEC): Use crtendS.o for PIE.
1163 2018-11-09  Sandra Loosemore  <sandra@codesourcery.com>
1165         PR driver/41179
1166         PR middle-end/65703
1167         * doc/invoke.texi (Optimize Options): Clarify default behavior
1168         for -fno-toplevel-reorder, -fno-defer-pop, and -fno-branch-count-reg.
1170 2018-11-09  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1172         PR tree-optimization/87940
1173         * expr.c (string_constant): Don't strip NOPS in subexpressions.
1174         Fold PLUS_EXPR correctly.
1176 2018-11-09  Ilya Leoshkevich  <iii@linux.ibm.com>
1178         PR target/87762
1179         * config/s390/s390.c (s390_safe_relative_long_p): New function.
1180         (annotate_constant_pool_refs): Skip insns which support
1181         relative addressing.
1182         (annotate_constant_pool_refs_1): New helper function.
1183         (find_constant_pool_ref): Skip insns which support relative
1184         addression.
1185         (find_constant_pool_ref_1): New helper function.
1186         (replace_constant_pool_ref): Skip insns which support
1187         relative addressing.
1188         (replace_constant_pool_ref_1): New helper function.
1189         (s390_mainpool_start): Adapt to the new signature.
1190         (s390_mainpool_finish): Likewise.
1191         (s390_chunkify_start): Likewise.
1192         (s390_chunkify_finish): Likewise.
1193         (pass_s390_early_mach::execute): Likewise.
1194         (s390_prologue_plus_offset): Likewise.
1195         (s390_emit_prologue): Likewise.
1196         (s390_emit_epilogue): Likewise.
1198 2018-11-09  Jakub Jelinek  <jakub@redhat.com>
1200         * gimplify.c (gimplify_scan_omp_clauses): Call sorry_at for valid
1201         but unsupported lastprivate with conditional modifier.
1203 2018-11-09  Jeff Law  <law@redhat.com>
1205         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): Handle
1206         unused argument better.  Add gcc_unreachable to silence warning.
1208 2018-11-09  Martin Sebor  <msebor@redhat.com>
1210         PR middle-end/81824
1211         * attribs.c (has_attribute): New helper function.
1212         (decls_mismatched_attributes, maybe_diag_alias_attributes): Same.
1213         * attribs.h (decls_mismatched_attributes): Declare.
1214         * cgraphunit.c (handle_alias_pairs): Call maybe_diag_alias_attributes.
1215         (maybe_diag_incompatible_alias): Use OPT_Wattribute_alias_.
1216         * common.opt (-Wattribute-alias): Take an argument.
1217         (-Wno-attribute-alias): New option.
1218         * doc/extend.texi (Common Function Attributes): Document copy.
1219         (Common Variable Attributes): Same.
1220         * doc/invoke.texi (-Wmissing-attributes): Document enhancement.
1221         (-Wattribute-alias): Document new option argument.
1223 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
1225         * config/arm/parsecpu.awk (/alias/): Tighten invisible alias
1226         matching criteria.  Remove unused array initializer.
1228 2018-11-09  Bill Schmidt  <wschmidt@linux.ibm.com>
1229             Jinsong Ji  <jji@us.ibm.com>
1231         * config/rs6000/xmmintrin.h (_mm_cvtss_si32): Fix incorrect
1232         constraints by introducing a new temporary.
1233         (_mm_cvtss_si64): Likewise.
1235 2018-11-09  Martin Liska  <mliska@suse.cz>
1237         * common.opt: Add -fipa-stack-alignment flag.
1238         * doc/invoke.texi: Document it.
1239         * final.c (rest_of_clean_state): Guard stack
1240         shrinking with flag.
1242 2018-11-09  Martin Liska  <mliska@suse.cz>
1244         * cgraph.h (ipa_discover_readonly_nonaddressable_vars): Rename
1245         to ...
1246         (ipa_discover_variable_flags): ... this.
1247         * common.opt: Come up with new flag -fipa-reference-addressable.
1248         * doc/invoke.texi: Document it.
1249         * ipa-reference.c (propagate): Call the renamed fn.
1250         * ipa-visibility.c (whole_program_function_and_variable_visibility):
1251         Likewise.
1252         * ipa.c (ipa_discover_readonly_nonaddressable_vars): Renamed to
1253         ...
1254         (ipa_discover_variable_flags): ... this.  Discover
1255         non-addressable variables only with the newly added flag.
1256         * opts.c: Enable the newly added flag with -O1 and higher
1257         optimization level.
1259 2018-11-09  David Malcolm  <dmalcolm@redhat.com>
1261         * json.cc (selftest::test_writing_literals): Fix comment.
1263 2018-11-09  Martin Liska  <mliska@suse.cz>
1265         * config/aarch64/aarch64.c (aarch64_parse_arch): Do not copy
1266         string to a stack buffer.
1267         (aarch64_parse_cpu): Likewise.
1268         (aarch64_parse_tune): Likewise.
1270 2018-11-09  Richard Biener  <rguenther@suse.de>
1272         PR tree-optimization/87953
1273         * tree-vect-loop.c (vectorizable_reduction): For analysis
1274         always pass ops[0] to vectorizable_condition.
1276 2018-11-09  Stafford Horne  <shorne@gmail.com>
1277             Richard Henderson  <rth@twiddle.net>
1278             Joel Sherrill  <joel@rtems.org>
1280         * common/config/or1k/or1k-common.c: New file.
1281         * config/or1k/*: New.
1282         * config.gcc (or1k*-*-*): New.
1283         * configure.ac (or1k*-*-*): New test for openrisc tls.
1284         * configure: Regenerated.
1285         * doc/install.texi: Document OpenRISC triplets.
1286         * doc/invoke.texi: Document OpenRISC arguments.
1287         * doc/md.texi: Document OpenRISC.
1289 2018-11-09  Richard Earnshaw  <rearnsha@arm.com>
1291         * config/arm/arm-cpus.in (arm7tdmi): Add an alias for arm7tdmi-s.
1292         (arm7tdmi-s): Delete CPU.
1293         (arm710t): Add aliases for arm720t and arm740t.
1294         (arm720t, arm740t): Delete CPUs.
1295         (arm920t): Add aliases for arm920, arm922t and arm940t.
1296         (arm920, arm922t, arm940t): Delete CPUs.
1297         (arm10tdmi): Add alias for arm1020t.
1298         (arm1020t): Delete CPU.
1299         (arm9e): Add aliases for arm946e-s, arm966e-s and arm968e-s.
1300         (arm946e-s, arm966e-s, arm968e-s): Delete CPUs.
1301         (arm10e): Add aliases for arm1020e and arm1022e.
1302         (arm1020e, arm1022e): Delete CPU.
1303         * config/arm/arm.md (generic_sched): Remove entries that are now
1304         handled by aliases.
1305         (generic_vfp): Likewise.
1306         * config/arm/arm1020e.md: Simplify tuning selection based on alias
1307         changes.
1308         * config/arm/arm-tune.md: Regenerated.
1309         * config/arm/arm-tables.opt: Regenerated.
1311 2018-11-09  Richard Biener  <rguenther@suse.de>
1313         PR tree-optimization/87621
1314         * tree-vect-loop.c (vectorizable_reduction): Handle reduction
1315         op with only phi inputs.
1316         * tree-ssa-loop-ch.c: Include tree-ssa-sccvn.h.
1317         (ch_base::copy_headers): Run CSE on copied loop headers.
1318         (pass_ch_vect::process_loop_p): Simplify.
1320 2018-11-09  Alexandre Oliva <oliva@adacore.com>
1322         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Adjust
1323         for non-w64 x86_64 biarch.
1325 2018-11-09  Alexandre Oliva <aoliva@redhat.com>
1327         PR rtl-optimization/86438
1328         * compare-elim.c (try_eliminate_compare): Use SET_SRC instead
1329         of in_b for the compare if in_b is SET_DEST.
1331         PR target/87793
1332         * config/i386/i386.c (ix86_const_not_ok_for_debug_p): Reject
1333         non-toplevel UNSPEC.
1335 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1337         * tree-vrp.c (value_range::check): Do not access internals
1338         directly.
1339         (value_range::singleton_p): Same.
1340         (value_range::type): Same.
1341         (vrp_finalize): Use value_range API.
1343 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1345         * tree-vrp.c (may_contain_p): Do not access m_min/m_max directly.
1347 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1349         * vr-values.c (vr_values::get_value_range): Use value_range API
1350         instead of piecing together ranges.
1351         (vr_values::update_value_range): Same.
1353 2018-11-08  Aldy Hernandez  <aldyh@redhat.com>
1355         * gimple-fold.c (size_must_be_zero_p): Use value_range API instead
1356         of performing ad-hoc calculations.
1357         * tree-ssanames.c (set_range_info): New overloaded function
1358         accepting value_range &.
1359         (get_range_info): Same.
1360         * tree-ssanames.h (set_range_info_raw): Remove.
1361         (set_range_info): New prototype.
1362         (get_range_info): Same.
1363         * tree-vrp.h (value_range::null_p): Rename to zero_p.
1364         * tree-vrp.c (value_range::null_p): Same.
1366 2018-11-09  Jan Hubicka  <jh@suse.cz>
1368         * tree.c (fld_type_variant_equal_p): Test user align flag.
1369         (flt_type_variant): Copy user align flag.
1370         (fld_incomplete_type_of): Clear it.
1372 2018-11-09  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
1374         * config/arm/neon.md (div<mode>3): New pattern.
1376 2018-11-08  Andi Kleen  <ak@linux.intel.com>
1378         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PTWRITE_SET): New.
1379         (OPTION_MASK_ISA_PTWRITE_UNSET): New.
1380         (ix86_handle_option): Handle OPT_mptwrite.
1381         * config/i386/cpuid.h (bit_PTWRITE): Add.
1382         * config/i386/driver-i386.c (host_detect_local_cpu): Detect ptwrite.
1383         * config/i386/i386-builtin.def (BDESC): Add ptwrite32/64.
1384         * config/i386/i386-c.c (ix86_target_macros_internal): Define __PTWRITE__.
1385         * config/i386/i386.c (ix86_target_string): Handle ptwrite.
1386         (ix86_option_override_internal): Handle PTA_PTWRITE.
1387         (ix86_valid_target_attribute_inner_p): Define ptwrite.
1388         (def_builtin2): Force UINT64 to be 64bit only.
1389         * config/i386/i386.h (TARGET_PTWRITE): Add.
1390         (TARGET_PTWRITE_P): Add.
1391         (PTA_PTWRITE): Add.
1392         * config/i386/i386.md: Define ptwrite.
1393         * config/i386/i386.opt: Add -mptwrite.
1394         * config/i386/immintrin.h (_ptwrite64): Add.
1395         (_ptwrite32): Add
1396         * doc/extend.texi: Document __builtin_ia32_ptwrite*.
1397         * doc/invoke.texi: Document -mptwrite.
1399 2018-11-08  Peter Bergner  <bergner@linux.ibm.com>
1401         PR rtl-optimization/87600
1402         * cfgexpand.c (expand_asm_stmt): Catch illegal asm constraint usage.
1403         * lra-constraints.c (process_alt_operands): Skip illegal hard
1404         register usage.  Prefer reloading non hard register operands.
1406 2018-11-08  Sandra Loosemore  <sandra@codesourcery.com>
1408         PR other/36572
1409         * doc/invoke.texi (Optimize Options): Clarify default behavior
1410         for -fno-sched-interblock and -fno-sched-spec.
1412 2018-11-08  Roman Geissler  <roman.geissler@amadeus.com>
1414         * collect2.c (linker_select):  Add USE_LLD_LD.
1415         (ld_suffixes): Add ld.lld.
1416         (main): Handle -fuse-ld=lld.
1417         * common.opt (-fuse-ld=lld): New option.
1418         * doc/invoke.texi (-fuse-ld=lld): Document.
1419         * opts.c (common_handle_option): Handle OPT_fuse_ld_lld.
1421 2018-11-08  Paul Koning  <ni1d@arrl.net>
1423         * config/pdp11/constraints.md: Add "Z" series constraints for use
1424         with pre-dec and post-inc addressing.
1425         * config/pdp11/pdp11-protos.m (expand_block_move): Delete.
1426         (pdp11_expand_operands): Add int argument (word count).
1427         (pdp11_sp_frame_offset): Delete.
1428         (pdp11_cmp_length): New function.
1429         (pushpop_regeq): New function.
1430         * config/pdp11/pdp11.c (TARGET_STACK_PROTECT_RUNTIME_ENABLED_P):
1431         Add hook.
1432         (pdp11_expand_prologue, pdp11_expand_epilogue): Rewrite for new
1433         frame layout.
1434         (pdp11_initial_elimination_offset): Ditto.
1435         (pdp11_expand_operands): Add word count argument.  Bugfixes.
1436         (output_move_multiple): Change how pointer adjustment is done.
1437         (pdp11_gen_int_label): Correct format.
1438         (output_ascii): Ditto.
1439         (pdp11_asm_output_var): Add code for DEC assembler case.
1440         (pdp11_asm_print_operand): Bugfix for CONST_DOUBLE holding integer
1441         value.
1442         (legitimate_const_double_p): Ditto.
1443         (pdp11_register_move_cost): Adjust for new register classes.
1444         (pdp11_regno_reg_class): Ditto.
1445         (expand_block_move): Delete.
1446         (pushpop_regeq): New function.
1447         (pdp11_legitimate_address_p): Bugfix in check for constant
1448         offset.
1449         (pdp11_sp_frame_offset): Delete.
1450         (pdp11_reg_save_size): New helper function for new frame layout.
1451         (output_addr_const_pdp11): Remove CONST_DOUBLE case.
1452         (pdp11_expand_shift): Bugfix in check for constant shift count.
1453         (pdp11_shift_length): Ditto.
1454         (pdp11_assemble_shift): Copy input to pdp11_expand_operands.
1455         (pdp11_cmp_length): New function.
1456         * config/pdp11/pdp11.h (TARGET_CPU_CPP_BUILTINS): Add macros for
1457         some compile options.
1458         (FIXED_REGISTERS): Remove HARD_FRAME_POINTER_REGNUM.
1459         (CALL_USED_REGISTERS): Ditto.
1460         (ELIMINABLE_REGS): Ditto.
1461         (REGISTER_NAMES): Ditto.
1462         (reg_class): Add classes NOTR0_REG through NOTSP_REG for use by Z
1463         constraints.
1464         (REG_CLASS_NAMES): Ditto.
1465         (REG_CLASS_CONTENTS): Ditto.  Also remove
1466         HARD_FRAME_POINTER_REGNUM.
1467         (CPU_REG_CLASS): New macro.
1468         (CLASS_MAX_NREGS): Adjust for new register classes.
1469         (FUNCTION_PROFILER): Make no-op.
1470         (may_call_alloca): Remove unused declaration.
1471         (ASM_OUTPUT_ALIGN): Add workaround for PR87795.
1472         (ASM_OUTPUT_SKIP): Fix format.
1473         * config/pdp11/pdp11.md (unspecv): Add UNSPECV_MOVMEM.
1474         (HARD_FRAME_POINTER_REGNUM): Remove.
1475         (return): Delete.
1476         (*rts): Rename.  Remove epilogue related checks.
1477         (cmpsi, cmpdi): New insn.
1478         (cbranch<mode>4): Change to apply to SI and DI modes as well.
1479         (mov<mode>): Change constraints to enforce that push/pop
1480         destination cannot use the same register as source.
1481         (*mov<mode><cc_cc>): Ditto.
1482         (movmemhi, movmemhi1, movmemhi_nocc): Change to expand block move
1483         at assembly output rather than as RTL expander.
1484         (zero_extendqihi2): Bugfix in check for same registers.
1485         (adddi3_nocc): Bugfix in check for constant operand.
1486         (addsi3_nocc): Ditto.
1487         (subdi3_nocc): Ditto.
1488         (subsi3_nocc): Ditto.
1489         (negdi2_nocc): Copy input to pdp11_expand_operands.
1490         (negsi2_nocc): Ditto.
1491         (bswap2_nocc): Ditto.
1492         * config/pdp11/pdp11.opt (mlra): Fix documentation.
1493         * config/pdp11/t-pdp11: Use -Os.
1495 2018-11-08  Richard Earnshaw  <rearnsha@arm.com>
1497         * config/arm/parsecpu.awk (/alias/): New parsing rule.
1498         (/begin cpu/): Check that the cpu name hasn't been previously defined.
1499         (gen_comm_data): Print out CPU alias tables.
1500         (check_cpu): Match aliases when checking the CPU name.
1501         * config/arm/arm-protos.h (cpu_alias): New structure.
1502         (cpu_option): Add entry for aliases.
1503         * config/arm/arm-cpus.in (strongarm): Add aliases for strongarm110
1504         strongarm1100 and strongarm1110.
1505         (strongarm110, strongarm1100, strongarm1110): Delete CPU entries.
1506         (config/arm/arm-generic.md): Remove redundant references to
1507         strongarm110, strongarm1100 and strongarm1110.
1508         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
1509         Scan aliases for additional hints.
1510         (arm_parse_cpu_option_name): Also match a cpu name against the list
1511         of aliases.
1512         * config/arm/arm-tables.opt: Regenerated.
1513         * config/arm/arm-tune.md: Regenerated.
1515 2018-11-08  Jakub Jelinek  <jakub@redhat.com>
1517         * builtin-types.def (BT_FN_VOID_BOOL, BT_FN_VOID_SIZE_SIZE_PTR,
1518         BT_FN_UINT_UINT_PTR_PTR, BT_FN_UINT_OMPFN_PTR_UINT_UINT,
1519         BT_FN_BOOL_UINT_LONGPTR_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1520         BT_FN_BOOL_UINT_ULLPTR_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR,
1521         BT_FN_BOOL_LONG_LONG_LONG_LONG_LONG_LONGPTR_LONGPTR_PTR_PTR,
1522         BT_FN_BOOL_BOOL_ULL_ULL_ULL_LONG_ULL_ULLPTR_ULLPTR_PTR_PTR): New.
1523         * gengtype.c (open_base_files): Add omp-general.h.
1524         * gimple.c (gimple_build_omp_critical):
1525         (gimple_build_omp_taskgroup): Add CLAUSES argument.  Call
1526         gimple_omp_taskgroup_set_clauses.
1527         (gimple_build_omp_atomic_load): Add mo argument, call
1528         gimple_omp_atomic_set_memory_order.
1529         (gimple_build_omp_atomic_store): Likewise.
1530         (gimple_copy): Adjust handling of GIMPLE_OMP_TASKGROUP.
1531         * gimple.def (GIMPLE_OMP_TASKGROUP): Use GSS_OMP_SINGLE_LAYOUT
1532         instead of GSS_OMP.
1533         (GIMPLE_OMP_TEAMS): Use GSS_OMP_PARALLEL_LAYOUT instead
1534         of GSS_OMP_SINGLE_LAYOUT, adjust comments.
1535         * gimple.h (enum gf_mask): Add GF_OMP_TEAMS_HOST, GF_OMP_TASK_TASKWAIT
1536         and GF_OMP_ATOMIC_MEMORY_ORDER.  Remove GF_OMP_ATOMIC_SEQ_CST, use
1537         different value for GF_OMP_ATOMIC_NEED_VALUE.
1538         (struct gimple_statement_omp_taskreg): Add GIMPLE_OMP_TEAMS to
1539         comments.
1540         (struct gimple_statement_omp_single_layout): And remove here.
1541         (struct gomp_teams): Inherit from gimple_statement_omp_taskreg rather
1542         than gimple_statement_omp_single_layout.
1543         (is_a_helper <gimple_statement_omp_taskreg *>::test): Allow
1544         GIMPLE_OMP_TEAMS.
1545         (is_a_helper <const gimple_statement_omp_taskreg *>::test): Likewise.
1546         (gimple_omp_subcode): Formatting fix.
1547         (gimple_omp_teams_child_fn, gimple_omp_teams_child_fn_ptr,
1548         gimple_omp_teams_set_child_fn, gimple_omp_teams_data_arg,
1549         gimple_omp_teams_data_arg_ptr, gimple_omp_teams_set_data_arg,
1550         gimple_omp_teams_host, gimple_omp_teams_set_host,
1551         gimple_omp_task_taskwait_p, gimple_omp_task_set_taskwait_p,
1552         gimple_omp_taskgroup_clauses, gimple_omp_taskgroup_clauses_ptr,
1553         gimple_omp_taskgroup_set_clauses): New inline functions.
1554         (gimple_build_omp_atomic_load): Add enum omp_memory_order argument.
1555         (gimple_build_omp_atomic_store): Likewise.
1556         (gimple_omp_atomic_seq_cst_p): Remove.
1557         (gimple_omp_atomic_memory_order): New function.
1558         (gimple_omp_atomic_set_seq_cst): Remove.
1559         (gimple_omp_atomic_set_memory_order): New function.
1560         (gimple_build_omp_taskgroup): Add clauses argument.
1561         * gimple-pretty-print.c (dump_gimple_omp_taskgroup): New function.
1562         (dump_gimple_omp_task): Print taskwait with depend clauses.
1563         (dump_gimple_omp_atomic_load, dump_gimple_omp_atomic_store): Use
1564         dump_omp_atomic_memory_order.
1565         (pp_gimple_stmt_1): Handle GIMPLE_OMP_TASKGROUP.
1566         * gimplify.c (enum gimplify_omp_var_data): Add GOVD_MAP_ALLOC_ONLY,
1567         GOVD_MAP_FROM_ONLY and GOVD_NONTEMPORAL.
1568         (enum omp_region_type): Reserve bits 1 and 2 for auxiliary flags,
1569         renumber values of most of ORT_* enumerators, add ORT_HOST_TEAMS,
1570         ORT_COMBINED_HOST_TEAMS, ORT_TASKGROUP, ORT_TASKLOOP and
1571         ORT_UNTIED_TASKLOOP enumerators.
1572         (enum gimplify_defaultmap_kind): New.
1573         (struct gimplify_omp_ctx): Remove target_map_scalars_firstprivate and
1574         target_map_pointers_as_0len_arrays members, add defaultmap.
1575         (new_omp_context): Initialize defaultmap member.
1576         (gimple_add_tmp_var): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1577         (maybe_fold_stmt): Don't fold even in host teams regions.
1578         (omp_firstprivatize_variable): Handle ORT_TASKGROUP like
1579         ORT_WORKSHARE.  Test ctx->defaultmap[GDMK_SCALAR] instead of
1580         ctx->omp_firstprivatize_variable.
1581         (omp_add_variable): Don't add private/firstprivate for VLAs in
1582         ORT_TASKGROUP.
1583         (omp_default_clause): Print "taskloop" rather than "task" if
1584         ORT_*TASKLOOP.
1585         (omp_notice_variable): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1586         Handle new defaultmap clause kinds.
1587         (omp_is_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.  Allow simd
1588         iterator to be lastprivate or private.  Fix up diagnostics if linear
1589         is used on collapse>1 simd iterator.
1590         (omp_check_private): Handle ORT_TASKGROUP like ORT_WORKSHARE.
1591         (gimplify_omp_depend): New function.
1592         (gimplify_scan_omp_clauses): Add shared clause on parallel for
1593         combined parallel master taskloop{, simd} if taskloop has
1594         firstprivate, lastprivate or reduction clause.  Handle
1595         OMP_CLAUSE_REDUCTION_TASK diagnostics.  Adjust tests for
1596         ORT_COMBINED_TEAMS.  Gimplify depend clauses with iterators.  Handle
1597         cancel and simd OMP_CLAUSE_IF_MODIFIERs.  Handle
1598         OMP_CLAUSE_NONTEMPORAL.  Handle new defaultmap clause kinds.  Handle
1599         OMP_CLAUSE_{TASK,IN}_REDUCTION.  Diagnose invalid conditional
1600         lastprivate.
1601         (gimplify_adjust_omp_clauses_1): Ignore GOVD_NONTEMPORAL.  Handle
1602         GOVD_MAP_ALLOC_ONLY and GOVD_MAP_FROM_ONLY.  
1603         (gimplify_adjust_omp_clauses): Handle OMP_CLAUSE_NONTEMPORAL.  Handle
1604         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1605         (gimplify_omp_task): Handle taskwait with depend clauses.
1606         (gimplify_omp_for): Add shared clause on parallel for combined
1607         parallel master taskloop{, simd} if taskloop has firstprivate,
1608         lastprivate or reduction clause.  Use ORT_TASKLOOP or
1609         ORT_UNTIED_TASKLOOP instead of ORT_TASK or ORT_UNTIED_TASK.  Adjust
1610         tests for ORT_COMBINED_TEAMS.  Handle C++ range for loops with
1611         NULL TREE_PURPOSE in OMP_FOR_ORIG_DECLS.  Firstprivatize
1612         __for_end and __for_range temporaries on OMP_PARALLEL for
1613         distribute parallel for{, simd}.  Move OMP_CLAUSE_REDUCTION
1614         and OMP_CLAUSE_IN_REDUCTION from taskloop to the task construct
1615         sandwiched in between two taskloops.
1616         (computable_teams_clause): Test ctx->defaultmap[GDMK_SCALAR]
1617         instead of ctx->omp_firstprivatize_variable.
1618         (gimplify_omp_workshare): Set ort to ORT_HOST_TEAMS or
1619         ORT_COMBINED_HOST_TEAMS if not inside of target construct.  If
1620         host teams, use gimplify_and_return_first etc. for body like
1621         for target or target data constructs, and at the end call
1622         gimple_omp_teams_set_host on the GIMPLE_OMP_TEAMS object.
1623         (gimplify_omp_atomic): Use OMP_ATOMIC_MEMORY_ORDER instead
1624         of OMP_ATOMIC_SEQ_CST, pass it as new argument to
1625         gimple_build_omp_atomic_load and gimple_build_omp_atomic_store, remove
1626         gimple_omp_atomic_set_seq_cst calls.
1627         (gimplify_expr) <case OMP_TASKGROUP>: Move handling into a separate
1628         case, handle taskgroup clauses.
1629         * lto-streamer-out.c (hash_tree): Handle
1630         OMP_CLAUSE_{TASK,IN}_REDUCTION.
1631         * Makefile.in (GTFILES): Add omp-general.h.
1632         * omp-builtins.def (BUILT_IN_GOMP_TASKWAIT_DEPEND,
1633         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_START,
1634         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_START,
1635         BUILT_IN_GOMP_LOOP_START, BUILT_IN_GOMP_LOOP_ORDERED_START,
1636         BUILT_IN_GOMP_LOOP_DOACROSS_START,
1637         BUILT_IN_GOMP_LOOP_NONMONOTONIC_RUNTIME_NEXT,
1638         BUILT_IN_GOMP_LOOP_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1639         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_START,
1640         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_START,
1641         BUILT_IN_GOMP_LOOP_ULL_START, BUILT_IN_GOMP_LOOP_ULL_ORDERED_START,
1642         BUILT_IN_GOMP_LOOP_ULL_DOACROSS_START,
1643         BUILT_IN_GOMP_LOOP_ULL_NONMONOTONIC_RUNTIME_NEXT,
1644         BUILT_IN_GOMP_LOOP_ULL_MAYBE_NONMONOTONIC_RUNTIME_NEXT,
1645         BUILT_IN_GOMP_PARALLEL_LOOP_NONMONOTONIC_RUNTIME,
1646         BUILT_IN_GOMP_PARALLEL_LOOP_MAYBE_NONMONOTONIC_RUNTIME,
1647         BUILT_IN_GOMP_PARALLEL_REDUCTIONS, BUILT_IN_GOMP_SECTIONS2_START,
1648         BUILT_IN_GOMP_TEAMS_REG, BUILT_IN_GOMP_TASKGROUP_REDUCTION_REGISTER,
1649         BUILT_IN_GOMP_TASKGROUP_REDUCTION_UNREGISTER,
1650         BUILT_IN_GOMP_TASK_REDUCTION_REMAP,
1651         BUILT_IN_GOMP_WORKSHARE_TASK_REDUCTION_UNREGISTER): New builtins.
1652         * omp-expand.c (workshare_safe_to_combine_p): Return false for
1653         non-worksharing loops.
1654         (omp_adjust_chunk_size): Don't adjust anything if chunk_size is zero.
1655         (determine_parallel_type): Don't combine parallel with worksharing
1656         which has _reductemp_ clause.
1657         (expand_parallel_call): Emit the GOMP_*nonmonotonic_runtime* or
1658         GOMP_*maybe_nonmonotonic_runtime* builtins instead of GOMP_*runtime*
1659         if there is nonmonotonic modifier or if there is no modifier and no
1660         ordered clause.  For dynamic and guided schedule without monotonic
1661         and nonmonotonic modifier, default to nonmonotonic.
1662         (expand_omp_for): Likewise.  Adjust expand_omp_for_generic caller, use
1663         GOMP_loop{,_ull}{,_ordered,_doacross}_start builtins if there are
1664         task reductions.
1665         (expand_task_call): Add GOMP_TASK_FLAG_REDUCTION flag to flags if
1666         there are any reduction clauses.
1667         (expand_taskwait_call): New function.
1668         (expand_teams_call): New function.
1669         (expand_omp_taskreg): Allow GIMPLE_OMP_TEAMS and call
1670         expand_teams_call for it.  Formatting fix.  Handle taskwait with
1671         depend clauses.
1672         (expand_omp_for_generic): Add SCHED_ARG argument.  Handle expansion
1673         of worksharing loops with task reductions.
1674         (expand_omp_for_static_nochunk, expand_omp_for_static_chunk): Handle
1675         expansion of worksharing loops with task reductions.
1676         (expand_omp_sections): Handle expansion of sections with task
1677         reductions.
1678         (expand_omp_synch): For host teams call expand_omp_taskreg.
1679         (omp_memory_order_to_memmodel): New function.
1680         (expand_omp_atomic_load, expand_omp_atomic_store,
1681         expand_omp_atomic_fetch_op): Use it and gimple_omp_atomic_memory_order
1682         instead of gimple_omp_atomic_seq_cst_p.
1683         (build_omp_regions_1, omp_make_gimple_edges): Treat taskwait with
1684         depend clauses as a standalone directive.
1685         * omp-general.c (enum omp_requires): New variable.
1686         (omp_extract_for_data): Initialize have_reductemp member.  Allow
1687         NE_EXPR even in OpenMP loops, transform them into LT_EXPR or
1688         GT_EXPR loops depending on incr sign.  Formatting fixes.
1689         * omp-general.h (struct omp_for_data): Add have_reductemp member.
1690         (enum omp_requires): New enum.
1691         (omp_requires_mask): Declare.
1692         * omp-grid.c (grid_eliminate_combined_simd_part): Formatting fix.
1693         Fix comment typos.
1694         * omp-low.c (struct omp_context): Add task_reductions and
1695         task_reduction_map fields.
1696         (is_host_teams_ctx): New function.
1697         (is_taskreg_ctx): Return true also if is_host_teams_ctx.
1698         (use_pointer_for_field): Use is_global_var instead of
1699         TREE_STATIC || DECL_EXTERNAL, and apply only if not privatized
1700         in outer contexts.
1701         (build_outer_var_ref): Ignore taskgroup outer contexts.
1702         (delete_omp_context): Release task_reductions and task_reduction_map.
1703         (scan_sharing_clauses): Don't add any fields for reduction clause on
1704         taskloop.  Handle OMP_CLAUSE__REDUCTEMP_.  Handle
1705         OMP_CLAUSE_{IN,TASK}_REDUCTION and OMP_CLAUSE_REDUCTION with task
1706         modifier.  Don't ignore shared clauses in is_host_teams_ctx contexts.
1707         Handle OMP_CLAUSE_NONTEMPORAL.
1708         (add_taskreg_looptemp_clauses): Add OMP_CLAUSE__REDUCTEMP_ clause if
1709         needed.
1710         (scan_omp_parallel): Add _reductemp_ clause if there are any reduction
1711         clauses with task modifier.
1712         (scan_omp_task): Handle taskwait with depend clauses.
1713         (finish_taskreg_scan): Move field corresponding to _reductemp_ clause
1714         first.  Move also OMP_CLAUSE__REDUCTEMP_ clause in front if present.
1715         Handle GIMPLE_OMP_TEAMS like GIMPLE_OMP_PARALLEL.
1716         (scan_omp_for): Fix comment formatting.
1717         (scan_omp_teams): Handle host teams constructs.
1718         (check_omp_nesting_restrictions): Allow teams with no outer
1719         OpenMP context.  Adjust diagnostics for teams strictly nested into
1720         some explicit OpenMP construct other than target.  Allow OpenMP atomics
1721         inside of simd regions.
1722         (scan_omp_1_stmt): Call scan_sharing_clauses for taskgroups.
1723         (scan_omp_1_stmt) <case GIMPLE_OMP_TEAMS>: Temporarily bump
1724         taskreg_nesting_level while scanning host teams construct.
1725         (task_reduction_read): New function.
1726         (lower_rec_input_clauses): Handle OMP_CLAUSE_REDUCTION on taskloop
1727         construct.  Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE__REDUCTEMP_
1728         clauses.  Handle OMP_CLAUSE_REDUCTION with task modifier.  Remove
1729         second argument create_tmp_var if it is NULL.  Don't ignore shared
1730         clauses in is_host_teams_ctx contexts.  Handle
1731         OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE on OMP_CLAUSE_FIRSTPRIVATE
1732         clauses.
1733         (lower_reduction_clauses): Ignore reduction clauses with task
1734         modifier.  Remove second argument create_tmp_var if it is NULL.
1735         Initialize OMP_ATOMIC_MEMORY_ORDER to relaxed.
1736         (lower_send_clauses): Ignore reduction clauses with task modifier.
1737         Handle OMP_CLAUSE__REDUCTEMP_.  Don't send anything for
1738         OMP_CLAUSE_REDUCTION on taskloop.  Handle OMP_CLAUSE_IN_REDUCTION.
1739         (maybe_add_implicit_barrier_cancel): Add OMP_RETURN argument, don't
1740         rely that it is the last stmt in body so far.  Ignore outer taskgroup
1741         contexts.
1742         (omp_task_reductions_find_first, omp_task_reduction_iterate,
1743         lower_omp_task_reductions): New functions.
1744         (lower_omp_sections): Handle reduction clauses with taskgroup
1745         modifiers.  Adjust maybe_add_implicit_barrier_cancel caller.
1746         (lower_omp_single): Adjust maybe_add_implicit_barrier_cancel caller.
1747         (lower_omp_for): Likewise.  Handle reduction clauses with taskgroup
1748         modifiers.
1749         (lower_omp_taskgroup): Handle taskgroup reductions.
1750         (create_task_copyfn): Copy over OMP_CLAUSE__REDUCTEMP_ pointer.
1751         Handle OMP_CLAUSE_IN_REDUCTION and OMP_CLAUSE_REDUCTION clauses.
1752         (lower_depend_clauses): If there are any
1753         OMP_CLAUSE_DEPEND_DEPOBJ or OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1754         depend clauses, use a new array format.  If OMP_CLAUSE_DEPEND_LAST is
1755         seen, assume lowering is done already and return early.  Set kind
1756         on artificial depend clause to OMP_CLAUSE_DEPEND_LAST.
1757         (lower_omp_taskreg): Handle reduction clauses with task modifier on
1758         parallel construct.  Handle reduction clause on taskloop construct.
1759         Handle taskwait with depend clauses.
1760         (lower_omp_1): Use lower_omp_taskreg instead of lower_omp_teams
1761         for host teams constructs.
1762         * tree.c (omp_clause_num_ops): Add in_reduction, task_reduction,
1763         nontemporal and _reductemp_ clause entries.
1764         (omp_clause_code_name): Likewise.
1765         (walk_tree_1): Handle OMP_CLAUSE_{IN,TASK}_REDUCTION,
1766         OMP_CLAUSE_NONTEMPORAL and OMP_CLAUSE__REDUCTEMP_.
1767         * tree-core.h (enum omp_clause_code): Add
1768         OMP_CLAUSE_{{IN,TASK}_REDUCTION,NONTEMPORAL,_REDUCTEMP_}.
1769         (enum omp_clause_defaultmap_kind, enum omp_memory_order): New.
1770         (struct tree_base): Add omp_atomic_memory_order field into union.
1771         Remove OMP_ATOMIC_SEQ_CST comment.
1772         (enum omp_clause_depend_kind): Add OMP_CLAUSE_DEPEND_MUTEXINOUTSET
1773         and OMP_CLAUSE_DEPEND_DEPOBJ.
1774         (struct tree_omp_clause): Add subcode.defaultmap_kind.
1775         * tree.def (OMP_TASKGROUP): Add another operand, move next to other
1776         OpenMP constructs with body and clauses operands.
1777         * tree.h (OMP_BODY): Use OMP_MASTER instead of OMP_TASKGROUP.
1778         (OMP_CLAUSES): Use OMP_TASKGROUP instead of OMP_SINGLE.
1779         (OMP_TASKGROUP_CLAUSES): Define.
1780         (OMP_CLAUSE_DECL): Use OMP_CLAUSE__REDUCTEMP_ instead of
1781         OMP_CLAUSE__LOOPTEMP_.
1782         (OMP_ATOMIC_SEQ_CST): Remove.
1783         (OMP_ATOMIC_MEMORY_ORDER, OMP_CLAUSE_FIRSTPRIVATE_NO_REFERENCE,
1784         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL): Define.
1785         (OMP_CLAUSE_REDUCTION_CODE, OMP_CLAUSE_REDUCTION_INIT,
1786         OMP_CLAUSE_REDUCTION_MERGE, OMP_CLAUSE_REDUCTION_PLACEHOLDER,
1787         OMP_CLAUSE_REDUCTION_DECL_PLACEHOLDER,
1788         OMP_CLAUSE_REDUCTION_OMP_ORIG_REF): Handle
1789         OMP_CLAUSE_{,IN_,TASK_}REDUCTION.
1790         (OMP_CLAUSE_REDUCTION_TASK, OMP_CLAUSE_REDUCTION_INSCAN,
1791         OMP_CLAUSE_DEFAULTMAP_KIND, OMP_CLAUSE_DEFAULTMAP_CATEGORY,
1792         OMP_CLAUSE_DEFAULTMAP_BEHAVIOR, OMP_CLAUSE_DEFAULTMAP_SET_KIND):
1793         Define.
1794         * tree-inline.c (remap_gimple_stmt): Remap taskgroup clauses.
1795         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
1796         OMP_CLAUSE__REDUCTEMP_, OMP_CLAUSE_NONTEMPORAL.
1797         (convert_local_omp_clauses): Likewise.  Remove useless test.
1798         * tree-parloops.c (create_call_for_reduction_1): Pass
1799         OMP_MEMORY_ORDER_RELAXED as new argument to
1800         dump_gimple_omp_atomic_load and dump_gimple_omp_atomic_store.
1801         * tree-pretty-print.c (dump_omp_iterators): New function.
1802         (dump_omp_clause): Handle OMP_CLAUSE__REDUCTEMP_,
1803         OMP_CLAUSE_NONTEMPORAL, OMP_CLAUSE_{TASK,IN}_REDUCTION.  Print
1804         reduction modifiers.  Handle OMP_CLAUSE_DEPEND_DEPOBJ and
1805         OMP_CLAUSE_DEPEND_MUTEXINOUTSET.  Print iterators in depend clauses.
1806         Print __internal__ for OMP_CLAUSE_DEPEND_LAST.  Handle cancel and
1807         simd OMP_CLAUSE_IF_MODIFIERs.  Handle new kinds of
1808         OMP_CLAUSE_DEFAULTMAP. Print conditional: for
1809         OMP_CLAUSE_LASTPRIVATE_CONDITIONAL.
1810         (dump_omp_atomic_memory_order): New function.
1811         (dump_generic_node): Use it.  Print taskgroup clauses.  Print
1812         taskwait with depend clauses.
1813         * tree-pretty-print.h (dump_omp_atomic_memory_order): Declare.
1814         * tree-streamer-in.c (unpack_ts_omp_clause_value_fields):
1815         Handle OMP_CLAUSE_{TASK,IN}_REDUCTION.
1816         * tree-streamer-out.c (pack_ts_omp_clause_value_fields,
1817         write_ts_omp_clause_tree_pointers): Likewise.
1819 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1821         PR ipa/86395
1822         * doc/invoke.texi (-fdump-ipa-): Document the "-optimized",
1823         "-missed", "-note", and "-all" sub-options.
1824         * ipa-inline.c (caller_growth_limits): Port from fprintf to dump
1825         API.
1826         (can_early_inline_edge_p): Likewise.
1827         (want_early_inline_function_p): Likewise.
1828         (want_inline_self_recursive_call_p): Likewise.
1829         (recursive_inlining): Likewise.
1830         (inline_small_functions): Likewise.
1831         (flatten_function): Likewise.
1832         (ipa_inline): Likewise.
1833         (inline_always_inline_functions): Likewise.
1834         (early_inline_small_functions): Likewise.
1835         (early_inliner): Likewise.
1836         * tree-inline.c (expand_call_inline): Likewise.
1838 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1840         * pretty-print.c (pp_format): Handle %f.
1841         (selftest::test_pp_format): Add test of %f.
1842         * pretty-print.h (pp_double): New macro.
1844 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1846         * dump-context.h (ASSERT_IS_CGRAPH_NODE): New macro.
1847         * dumpfile.c (make_item_for_dump_cgraph_node): Move to before...
1848         (dump_pretty_printer::decode_format): Implement "%C" for
1849         cgraph_node *.
1850         (selftest::test_capture_of_dump_calls): Rename "where" to
1851         "stmt_loc".  Convert test_decl to a function decl and set its
1852         location.  Add a symbol_table_test RAII instance and a
1853         cgraph_node, using it to test "%C" and dump_symtab_node.
1855 2018-11-08  Eric Botcazou  <ebotcazou@adacore.com>
1857         PR middle-end/87916
1858         * cgraphclones.c (duplicate_thunk_for_node): Also set DECL_IGNORED_P.
1860 2018-11-08  David Malcolm  <dmalcolm@redhat.com>
1862         * cgraph.c: Include "selftest.h".
1863         (saved_symtab): New variable.
1864         (selftest::symbol_table_test::symbol_table_test): New ctor.
1865         (selftest::symbol_table_test::~symbol_table_test): New dtor.
1866         (selftest::test_symbol_table_test): New test.
1867         (selftest::cgraph_c_tests): New.
1868         * cgraph.h (saved_symtab): New decl.
1869         (selftest::symbol_table_test): New class.
1870         * selftest-run-tests.c (selftest::run_tests): Call
1871         selftest::cgraph_c_tests.
1872         * selftest.h (selftest::cgraph_c_tests): New decl.
1874 2018-11-08  Richard Biener  <rguenther@suse.de>
1876         * tree-data-ref.h (lambda_int): New typedef.
1877         (lambda_vector_gcd): Adjust.
1878         (lambda_vector_new): Likewise.
1879         (lambda_matrix_new): Likewise.
1880         * tree-data-ref.c  (print_lambda_vector): Adjust.
1882 2018-11-08  Richard Biener  <rguenther@suse.de>
1884         PR tree-optimization/87929
1885         * tree-complex.c (expand_complex_comparison): Clean EH.
1887 2018-11-08  Martin Liska  <mliska@suse.cz>
1889         * doc/extend.texi: Reword.
1890         * predict.c (expr_expected_value_1): Likewise.
1892 2018-11-08  Richard Biener  <rguenther@suse.de>
1894         PR tree-optimization/87913
1895         * tree-ssa-phiopt.c (minmax_replacement): Turn EQ/NE compares
1896         of extreme values to ordered comparisons.
1898 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1900         PR middle-end/42726
1901         * doc/invoke.texi (Code Gen Options): Clarify -fno-common behavior.
1903 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1905         * doc/invoke.texi: Remove leading dash from @opindex entries
1906         throughout the file.
1908 2018-11-07  Sandra Loosemore  <sandra@codesourcery.com>
1910         PR driver/80828
1911         * doc/invoke.texi (Option Summary): Add -e and --entry.
1912         (Link Options): Likewise.
1914 2018-11-07  Nathan Sidwell  <nathan@acm.org>
1916         PR 87926
1917         * Makefile.in (bitmap.o-warn): Add -Wno-error to unbreak
1918         --disable-checking bootstrap.
1920 2018-11-07  Hafiz Abid Qadeer  <abidh@codesourcery.com>
1922         * configure: Regenerated.
1924 2018-11-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
1926         PR c/87691
1927         * stor-layout.c (compute_record_mode): Set TYPE_MODE of UNION_TYPE
1928         to the mode of the widest field iff the widest field has mode class
1929         MODE_INT, or MODE_PARTIAL_INT and the union would be passed by
1930         reference.
1932 2018-11-07  Nikolai Merinov  <n.merinov@inango-systems.com>
1934         * common.opt: Add -Wattribute-warning.
1935         * doc/invoke.texi: Add documentation for -Wno-attribute-warning.
1936         * expr.c (expand_expr_real_1): Add new attribute to warning_at
1937         call to allow user configure behavior of "warning" attribute.
1939 2018-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
1941         * target.def: Put @: after every vs., e.g., and i.e. where it is
1942         followed by whitespace.
1943         * doc/extend.texi: Ditto.
1944         * doc/fragments.texi: Ditto.
1945         * doc/gimple.texi: Ditto.
1946         * doc/implement-c.texi: Ditto.
1947         * doc/install.texi: Ditto.
1948         * doc/invoke.texi: Ditto.
1949         * doc/md.texi: Ditto.
1950         * doc/plugins.texi: Ditto.
1951         * doc/rtl.texi: Ditto.
1952         * doc/sourcebuild.texi: Ditto.
1953         * doc/tm.texi.in: Ditto.
1954         * doc/ux.texi: Ditto.
1955         * doc/tm.texi: Regenerate.
1957 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1959         * config/arm/arm-cpus.in (ares): New entry.
1960         * config/arm/arm-tables.opt: Regenerate.
1961         * config/arm/arm-tune.md: Likewise.
1962         * doc/invoke.texi (ARM Options): Document ares.
1964 2018-11-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1966         * config/aarch64/aarch64-cores.def (ares): Define.
1967         * config/aarch64/aarch64-tune.md: Regenerate.
1968         * doc/invoke.texi (AArch64 Options): Document ares value for mtune.
1970 2018-11-07  Jan Hubicka  <jh@suse.cz>
1972         * tree.c (fld_incomplete_type_of): Clear TREE_ADDRESSABLE.
1973         (free_lang_data_in_decl): Set TREE_ADDRESSABLE for public vars and
1974         functions; clear TYPE_DECL_SUPPRESS_DEBUG and DECL_MODE for
1975         TYPE_DECL.
1977 2018-11-07  Richard Biener  <rguenther@suse.de>
1979         PR tree-optimization/87914
1980         * tree-vect-loop.c (vect_is_simple_reduction): Improve detection
1981         of nested cycles.
1982         (vectorizable_reduction): Handle shifts and rotates by dispatching
1983         to vectorizable_shift.
1984         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Handle
1985         in-loop uses of vect_nested_cycle defs.  Merge cycle and internal
1986         def cases.
1987         (vectorizable_shift): Export and handle being called as
1988         vect_nested_cycle.
1989         (vect_analyze_stmt): Call vectorizable_shift after
1990         vectorizable_reduction.
1991         * tree-vectorizer.h (vectorizable_shift): Declare.
1993 2018-11-07  Jan Hubicka  <jh@suse.cz>
1995         * ipa-devirt.c (odr_types_equivalent_p): Expect constants
1996         than const decls in TREE_VALUE of enum.
1997         (dump_type_inheritance_graph): Improve duplicate dumping.
1998         (free_enum_values): New.
1999         (build_type_inheritance_graph): Use it.
2000         * tree.c (free_lang_data_in_type): Free TYPE_VALUES of enums
2001         which are not main variants or not ODR types.
2002         (verify_type_variant): Expect variants to have no TYPE_VALUES.
2004 2018-11-07  Richard Biener  <rguenther@suse.de>
2006         * ipa-inline.c (want_inline_small_function_p): Compute
2007         big_speedup_p lazily and last.
2009 2018-11-07  Jan Hubicka  <jh@suse.cz>
2011         * tree.c (fld_type_variant_equal_p): Skip TYPE_ALIGN check when
2012         building incomplete variant of complete type.
2013         (fld_type_variant): Do not copy TYPE_ALIGN when building incomplete
2014         variant of complete type.
2016 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2018         * config/mips/mips.c: Fix typo in documentation of
2019         mips_loongson_ext2_prefetch_cookie.
2020         (mips_option_override): fix brain twister logical.
2021         * config/mips/mips.h: Fix typo in documentation of
2022         ISA_HAS_CTZ_CTO and define pattern.
2023         * config/mips/mips.md (prefetch): Hoist EXT2 above
2024         the 2EF/EXT block.
2025         (prefetch_indexed): Hoist EXT2 above the EXT block.
2027 2018-11-07  Jan Hubicka  <jh@suse.cz>
2029         * tree.c (free_lang_data_in_type): Add fld parameter; simplify
2030         return and parameter types of function and method types.
2031         (free_lang_data_in_cgraph): Update.
2033 2018-11-07  Martin Liska  <mliska@suse.cz>
2035         PR rtl-optimization/87868
2036         * postreload-gcse.c (eliminate_partially_redundant_load): Set
2037         threshold to max_count if we would overflow.
2038         * profile-count.h: Make max_count a public constant.
2040 2018-11-07  Martin Liska  <mliska@suse.cz>
2042         * mem-stats.h: Fix GNU coding style.
2044 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2046         * config/mips/gs264e.md: New.
2047         * config/mips/mips-cpus.def: Define gs264e.
2048         * config/mips/mips-tables.opt: Regenerate.
2049         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
2050         gs264e.
2051         (mips_issue_rate): Add support for gs264e.
2052         (mips_multipass_dfa_lookahead): Likewise.
2053         * config/mips/mips.h: Define TARGET_GS264E and TUNE_GS264E.
2054         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs264e.
2055         (MIPS_ASE_MSA_SPEC): New.
2056         (BASE_DRIVER_SELF_SPECS): march=gs264e implies -mmsa.
2057         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS264E.
2058         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS264E.
2059         * config/mips/mips.md: Include gs264e.md.
2060         (processor): Add gs264e.
2061         * config/mips/mips.opt (MSA): Use Mask instead of Var.
2062         * doc/invoke.texi: Add gs264e to supported architectures.
2064 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2066         * config/mips/gs464e.md: New.
2067         * config/mips/mips-cpus.def: Define gs464e.
2068         * config/mips/mips-tables.opt: Regenerate.
2069         * config/mips/mips.c (mips_rtx_cost_data): Add DEFAULT_COSTS for
2070         gs464e.
2071         (mips_issue_rate): Add support for gs464e.
2072         (mips_multipass_dfa_lookahead): Likewise.
2073         (mips_option_override): Enable MMI, EXT and EXT2 for gs464e.
2074         * config/mips/mips.h: Define TARGET_GS464E and TUNE_GS464E.
2075         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464e.
2076         (ISA_HAS_FUSED_MADD4): Enable for TARGET_GS464E.
2077         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_GS464E.
2078         * config/mips/mips.md: Include gs464e.md.
2079         (processor): Add gs464e.
2080         * doc/invoke.texi: Add gs464e to supported architectures.
2082 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2084         * config/mips/loongson3a.md: Rename to ...
2085         * config/mips/gs464.md: ... here.
2086         * config/mips/mips-cpus.def: Define gs464; Add loongson3a
2087         as an alias of gs464 processor.
2088         * config/mips/mips-tables.opt: Regenerate.
2089         * config/mips/mips.c (mips_issue_rate): Use PROCESSOR_GS464
2090         instead of PROCESSOR_LOONGSON_3A.
2091         (mips_multipass_dfa_lookahead): Use TUNE_GS464 instead of
2092         TUNE_LOONGSON_3A.
2093         (mips_option_override): Enable MMI and EXT for gs464.
2094         * config/mips/mips.h: Rename TARGET_LOONGSON_3A to TARGET_GS464;
2095         Rename TUNE_LOONGSON_3A to TUNE_GS464.
2096         (MIPS_ISA_LEVEL_SPEC): Infer mips64r2 from gs464.
2097         (ISA_HAS_ODD_SPREG, ISA_AVOID_DIV_HILO, ISA_HAS_FUSED_MADD4,
2098         ISA_HAS_UNFUSED_MADD4): Use TARGET_GS464 instead of
2099         TARGET_LOONGSON_3A.
2100         * config/mips/mips.md: Include gs464.md instead of loongson3a.md.
2101         (processor): Add gs464;
2102         * doc/invoke.texi: Add gs464 to supported architectures.
2104 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2106         * config/mips/mips-protos.h
2107         (mips_loongson_ext2_prefetch_cookie): New prototype.
2108         * config/mips/mips.c (mips_loongson_ext2_prefetch_cookie): New.
2109         (mips_option_override): Enable TARGET_LOONGSON_EXT when
2110         TARGET_LOONGSON_EXT2 is true.
2111         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Define
2112         __mips_loongson_ext2, __mips_loongson_ext_rev=2.
2113         (ISA_HAS_CTZ_CTO): New, true if TARGET_LOONGSON_EXT2.
2114         (ISA_HAS_PREFETCH): Include TARGET_LOONGSON_EXT and
2115         TARGET_LOONGSON_EXT2.
2116         (ASM_SPEC): Add mloongson-ext2 and mno-loongson-ext2.
2117         (define_insn "ctz<mode>2"): New insn pattern.
2118         (define_insn "prefetch"): Include TARGET_LOONGSON_EXT2.
2119         (define_insn "prefetch_indexed_<mode>"): Include
2120         TARGET_LOONGSON_EXT and TARGET_LOONGSON_EXT2.
2121         * config/mips/mips.opt (-mloongson-ext2): Add option.
2122         * gcc/doc/invoke.texi (-mloongson-ext2): Document.
2124 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2126         * config/mips/mips.h (TARGET_CPU_CPP_BUILTINS): Add
2127         __mips_loongson_ext.
2128         (MIPS_ASE_LOONGSON_EXT_SPEC): New.
2129         (BASE_DRIVER_SELF_SPECS): march=loongson3a implies
2130         -mloongson-ext.
2131         (ASM_SPEC): Add mloongson-ext and mno-loongson-ext.
2132         * config/mips/mips.md (mul<mode>3, mul<mode>3_mul3_nohilo,
2133         <u>div<mode>3, <u>mod<mode>3, prefetch): Use TARGET_LOONGSON_EXT
2134         instead of TARGET_LOONGSON_3A.
2135         * config/mips/mips.opt (-mloongson-ext): Add option.
2136         * gcc/doc/invoke.texi (-mloongson-ext): Document.
2138 2018-11-07  Chenghua Xu  <paul.hua.gm@gmail.com>
2140         * config.gcc (extra_headers): Add loongson-mmiintrin.h.
2141         * config/mips/loongson.md: Move to ...
2142         * config/mips/loongson-mmi.md: here; Adjustment.
2143         * config/mips/loongson.h: Move to ...
2144         State as deprecated. Include loongson-mmiintrin.h for back
2145         compatibility and warning.
2146         * config/mips/loongson-mmiintrin.h: ... here.
2147         * config/mips/mips.c (mips_hard_regno_mode_ok_uncached,
2148         mips_vector_mode_supported_p, AVAIL_NON_MIPS16): Use
2149         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
2150         (mips_option_override): Make sure MMI use hard float;
2151         (mips_shift_truncation_mask, mips_expand_vpc_loongson_even_odd,
2152         mips_expand_vpc_loongson_pshufh, mips_expand_vpc_loongson_bcast,
2153         mips_expand_vector_init): Use TARGET_LOONGSON_MMI instead of
2154         TARGET_LOONGSON_VECTORS.
2155         * gcc/config/mips/mips.h (TARGET_LOONGSON_VECTORS): Delete.
2156         (TARGET_CPU_CPP_BUILTINS): Add __mips_loongson_mmi.
2157         (MIPS_ASE_DSP_SPEC, MIPS_ASE_LOONGSON_MMI_SPEC): New.
2158         (BASE_DRIVER_SELF_SPECS): march=loongson2e/2f/3a implies
2159         -mloongson-mmi.
2160         (SHIFT_COUNT_TRUNCATED): Use TARGET_LOONGSON_MMI instead of
2161         TARGET_LOONGSON_VECTORS.
2162         * gcc/config/mips/mips.md (MOVE64, MOVE128): Use
2163         TARGET_LOONGSON_MMI instead of TARGET_LOONGSON_VECTORS.
2164         (Loongson MMI patterns): Include loongson-mmi.md instead of
2165         loongson.md.
2166         * gcc/config/mips/mips.opt (-mloongson-mmi): New option.
2167         * gcc/doc/invoke.texi (-mloongson-mmi): Document.
2169 2018-11-07  Richard Biener  <rguenther@suse.de>
2171         PR lto/87906
2172         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Fixup
2173         BLOCK_ABSTRACT_ORIGIN to be the ultimate origin.
2175 2018-11-07  Alexandre Oliva <aoliva@redhat.com>
2177         PR rtl-optimization/87874
2178         * lra.c (lra_substitute_pseudo): Do not create a subreg for
2179         const wide ints.
2181 2018-11-06  Aaron Sawdey  <acsawdey@linux.ibm.com>
2183         * config/rs6000/rs6000.md (bswap<mode>2): Force address into register
2184         if not in indexed or indirect form.
2185         (bswap<mode>2_load): Change predicate to indexed_or_indirect_operand.
2186         (bswap<mode>2_store): Ditto.
2188 2018-11-06  Richard Earnshaw  <rearnsha@arm.com>
2190         * config/aarch64/aarch64.md (speculation_tracker): Set the mode for
2191         the UNSPEC.
2193 2018-11-06  Richard Biener  <rguenther@suse.de>
2195         PR tree-optimization/86850
2196         * vec.h (vec<T, va_heap, vl_ptr>::splice): Check src.length ()
2197         instead of src.m_vec.
2199 2018-11-06  Jan Hubicka  <jh@suse.cz>
2201         * tree.c (fld_simplified_type_name): Break out form ...
2202         (free_lang_data_in_type): ... here.
2203         (fld_type_variant_equal_p): Use it.
2205 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2207         * config/default-d.c: Include memmodel.h.
2209         * config/sol2-d.c: New file.
2210         * config/t-sol2 (sol2-d.o): New rule.
2211         * config.gcc <*-*-solaris2*>: Set d_target_objs,
2212         target_has_targetdm.
2214 2018-11-06  Jan Hubicka  <jh@suse.cz>
2216         * tree.c (fld_type_variant): Also copy alignment; be sure that
2217         new variant is equal.
2219 2018-11-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2221         PR target/87762
2222         * config/s390/s390.md: Add relative_long attribute.
2224 2018-11-06  Jan Hubicka  <jh@suse.cz>
2226         * ipa-pure-const.c (check_decl): Do not test TYPE_NEEDS_CONSTRUCTING.
2227         * lto-streamer-out.c (hash_tree): Do not hash TYPE_NEEDS_CONSTRUCTING.
2228         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Do not
2229         stream TYPE_NEEDS_CONSTRUCTING.
2230         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
2231         * tree.c (free_lang_data_in_type): Clear TYPE_NEEDS_CONSTRUCTING.
2233 2018-11-06  Richard Biener  <rguenther@suse.de>
2235         * tree-vect-slp.c (vect_slp_bb): Move opening of vect_slp_analyze_bb
2236         dump-scope ...
2237         (vect_slp_analyze_bb_1): ... here to avoid hiding optimized locations.
2239 2018-11-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
2241         * gcc/config/msp430/msp430.h (REG_CLASS_CONTENTS): Add R0 to
2242         REG_CLASS_CONTENTS[GEN_REGS].
2243         (REGNO_REG_CLASS): Return NO_REGS for R2 and R3.
2245 2018-11-06  Jan Hubicka  <jh@suse.cz>
2247          * tree.c (fld_simplified_type_of): Clear TYPELESS_STORAGE flag.
2249 2018-11-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2251         PR sanitizer/80953
2252         * config/sol2.h (ASAN_CC1_SPEC): Define.
2253         (LD_WHOLE_ARCHIVE_OPTION): Define.
2254         (LD_NO_WHOLE_ARCHIVE_OPTION): Define.
2255         (ASAN_REJECT_SPEC): Provide default.
2256         (LIBASAN_EARLY_SPEC): Define.
2257         (LIBTSAN_EARLY_SPEC): Define.
2258         (LIBLSAN_EARLY_SPEC): Define.
2259         * config/i386/sol2.h (CC1_SPEC): Redefine.
2260         (ASAN_REJECT_SPEC): Define.
2262         * config/sparc/sparc.c (sparc_asan_shadow_offset): Declare.
2263         (TARGET_ASAN_SHADOW_OFFSET): Define.
2264         (sparc_asan_shadow_offset): New function.
2265         * config/sparc/sol2.h (CC1_SPEC): Append ASAN_CC1_SPEC.
2266         (ASAN_REJECT_SPEC): Define.
2268 2018-11-06  Jan Hubicka  <jh@suse.cz>
2270         * tree.c (fld_type_variant): Copy canonical type.
2271         (fld_incomplete_type_of): Check that canonical types looks sane;
2272         copy canonical type.
2273         (verify_type): Accept when incomplete type has complete canonical type.
2275 2018-11-06  Jan Hubicka  <jh@suse.cz>
2277         * tree.c (free_lang_data): Reset overwite_assembler_name,
2278         print_xnode, print_decl, print_type and print_identifier of
2279         langhooks.
2281 2018-11-06  Richard Biener  <rguenther@suse.de>
2283         PR tree-optimization/87889
2284         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2285         Do nothing if old and new arg are the same
2287 2018-11-06  Andreas Krebbel  <krebbel@linux.ibm.com>
2289         PR target/87723
2290         * config/s390/s390.md ("*r<noxa>sbg_di_rotl"): Remove mode
2291         attributes for operands 3 and 4.
2293 2018-11-06  Richard Biener  <rguenther@suse.de>
2295         PR middle-end/18041
2296         * simplify-rtx.c (simplify_binary_operation_1): Add pattern
2297         matching bitfield insertion.
2299 2018-11-06  Alexandre Oliva <aoliva@redhat.com>
2301         * auto-inc-dec.c: Include valtrack.h.  Improve comments.
2302         (reg_next_debug_use): New.
2303         (attempt_change): Propagate adjusted expression into affected
2304         debug insns.
2305         (merge_in_block): Track uses in debug insns.
2306         (pass_inc_dec::execute): Allocate and release
2307         reg_next_debug_use.
2309 2018-11-06 Wei Xiao <wei3.xiao@intel.com>
2311         * config/i386/avx512fintrin.h: Update VFIXUPIMM* intrinsics.
2312         (_mm512_fixupimm_round_pd): Update parameters and builtin.
2313         (_mm512_maskz_fixupimm_round_pd): Ditto.
2314         (_mm512_fixupimm_round_ps): Ditto.
2315         (_mm512_maskz_fixupimm_round_ps): Ditto.
2316         (_mm_fixupimm_round_sd): Ditto.
2317         (_mm_maskz_fixupimm_round_sd): Ditto.
2318         (_mm_fixupimm_round_ss): Ditto.
2319         (_mm_maskz_fixupimm_round_ss): Ditto.
2320         (_mm512_fixupimm_pd): Ditto.
2321         (_mm512_maskz_fixupimm_pd): Ditto.
2322         (_mm512_fixupimm_ps): Ditto.
2323         (_mm512_maskz_fixupimm_ps): Ditto.
2324         (_mm_fixupimm_sd): Ditto.
2325         (_mm_maskz_fixupimm_sd): Ditto.
2326         (_mm_fixupimm_ss): Ditto.
2327         (_mm_maskz_fixupimm_ss): Ditto.
2328         (_mm512_mask_fixupimm_round_pd): Update builtin.
2329         (_mm512_mask_fixupimm_round_ps): Ditto.
2330         (_mm_mask_fixupimm_round_sd): Ditto.
2331         (_mm_mask_fixupimm_round_ss): Ditto.
2332         (_mm512_mask_fixupimm_pd): Ditto.
2333         (_mm512_mask_fixupimm_ps): Ditto.
2334         (_mm_mask_fixupimm_sd): Ditto.
2335         (_mm_mask_fixupimm_ss): Ditto.
2336         * config/i386/avx512vlintrin.h:
2337         (_mm256_fixupimm_pd): Update parameters and builtin.
2338         (_mm256_maskz_fixupimm_pd): Ditto.
2339         (_mm256_fixupimm_ps): Ditto.
2340         (_mm256_maskz_fixupimm_ps): Ditto.
2341         (_mm_fixupimm_pd): Ditto.
2342         (_mm_maskz_fixupimm_pd): Ditto.
2343         (_mm_fixupimm_ps): Ditto.
2344         (_mm_maskz_fixupimm_ps): Ditto.
2345         (_mm256_mask_fixupimm_pd): Update builtin.
2346         (_mm256_mask_fixupimm_ps): Ditto.
2347         (_mm_mask_fixupimm_pd): Ditto.
2348         (_mm_mask_fixupimm_ps): Ditto.
2349         * config/i386/i386-builtin-types.def: Add new types and remove useless ones.
2350         * config/i386/i386-builtin.def: Update builtin definitions.
2351         * config/i386/i386.c: Handle new builtin types and remove useless ones.
2352         * config/i386/sse.md: Update VFIXUPIMM* patterns.
2353         (<avx512>_fixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2354         (<avx512>_fixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2355         (<avx512>_fixupimm<mode>_mask<round_saeonly_name>): Update.
2356         (avx512f_sfixupimm<mode>_maskz<round_saeonly_expand_name>): Update.
2357         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>): Update.
2358         (avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Update.
2359         * config/i386/subst.md:
2360         (round_saeonly_sd_mask_operand4): Add new subst_attr.
2361         (round_saeonly_sd_mask_op4): Ditto.
2362         (round_saeonly_expand_operand5): Ditto.
2363         (round_saeonly_expand): Update.
2365 2018-11-05  Max Filippov  <jcmvbkbc@gmail.com>
2367         * config/xtensa/uclinux.h (XTENSA_ALWAYS_PIC): Change to 0.
2369 2018-11-05  Segher Boessenkool  <segher@kernel.crashing.org>
2371         PR rtl-optimization/87871
2372         * combine.c (make_more_copies): Skip if dest is frame_pointer_rtx.
2374 2018-11-05  Paul Koning  <ni1d@arrl.net>
2376         * doc/sourcebuild.texi (target attributes): Document new "inf"
2377         effective target keyword.
2379 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2381         * config/s390/s390.c (s390_register_move_cost): Increase costs for
2382         moves involving the CC reg.
2384 2018-11-05  Richard Biener  <rguenther@suse.de>
2386         PR tree-optimization/87873
2387         * tree-ssa-loop-manip.h (split_loop_exit_edge): Add copy_constants_p
2388         argument.
2389         * tree-ssa-loop-manip.c (split_loop_exit_edge): Likewise.
2390         * tree-vect-loop.c (vect_transform_loop): When splitting the
2391         loop exit also create forwarder PHIs for constants.
2392         * tree-vect-loop-manip.c (slpeel_duplicate_current_defs_from_edges):
2393         Handle constant to_arg, add extra checking we match up the correct
2394         PHIs.
2396 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2398         * config/s390/s390.md: QImode and HImode for load on condition.
2400 2018-11-05  Robin Dapp  <rdapp@linux.ibm.com>
2402         * config/s390/predicates.md: Fix typo.
2403         * config/s390/s390.md: Allow immediates for load on condition.
2405 2018-11-05  Martin Liska  <mliska@suse.cz>
2407         * alloc-pool.h (struct pool_usage): Use SIZE_AMOUNT.
2408         * bitmap.h (struct bitmap_usage): Likewise.
2409         * ggc-common.c (SCALE): Remove.
2410         (LABEL): Likewise.
2411         (struct ggc_usage): Use SIZE_AMOUNT. And update
2412         compare method.
2413         * ggc-page.c (SCALE): Remove.
2414         (STAT_LABEL): Remove.
2415         (ggc_print_statistics): Use SIZE_AMOUNT.
2416         * gimple.h (SCALE): Remove.
2417         (LABEL): Likewise.
2418         * input.c (ONE_K): Remove.
2419         (ONE_M): Likewise.
2420         (SCALE): Likewise.
2421         (STAT_LABEL): Likewise.
2422         (FORMAT_AMOUNT): Likewise.
2423         (dump_line_table_statistics): Use SIZE_AMOUNT.
2424         * mem-stats.h (struct mem_usage): Likewise.
2425         * rtl.c (dump_rtx_statistics): Likewise.
2426         (rtx_alloc_counts): Change type to size_t.
2427         (rtx_alloc_sizes): Likewise.
2428         (rtx_count_cmp): New.
2429         (dump_rtx_statistics): Sort first based on counts.
2430         * tree.c (tree_nodes_cmp): New.
2431         (tree_codes_cmp): New.
2432         (dump_tree_statistics): Sort first based on counts.
2433         * system.h (ONE_K): New.
2434         (ONE_M): Likewise.
2435         (SIZE_SCALE): Likewise.
2436         (SIZE_LABEL): Likewise.
2437         (SIZE_AMOUNT): Likewise.
2438         * tree-cfg.c (dump_cfg_stats): Use SIZE_AMOUNT.
2439         * tree-dfa.c (dump_dfa_stats): Likewise.
2440         * tree-phinodes.c (phinodes_print_statistics): Likewise.
2441         * tree-ssanames.c (ssanames_print_statistics): Likewise.
2442         * tree.c (dump_tree_statistics): Likewise.
2443         * vec.c (struct vec_usage): Likewise.
2444         * trans-mem.c (tm_mangle): Enlarge buffer in order to not
2445         trigger a -Werror=format-overflow with
2446         --enable-gather-detailed-stats.
2448 2018-11-05  Martin Liska  <mliska@suse.cz>
2450         * mem-stats.h (mem_alloc_description::release_instance_overhead):
2451         Return T *.
2452         * vec.c (struct vec_usage): Register m_element_size.
2453         (vec_prefix::register_overhead): New arguments: elements and
2454         element_size.
2455         (vec_prefix::release_overhead): Subtract elements.
2456         * vec.h (struct vec_prefix): Change signature.
2457         (va_heap::reserve): Pass proper arguments.
2458         (va_heap::release): Likewise.
2460 2018-11-05  Martin Liska  <mliska@suse.cz>
2462         * mem-stats.h (mem_alloc_description::get_list): Fix GNU coding
2463         style.
2464         * vec.c: Likewise.
2466 2018-11-05  Richard Biener  <rguenther@suse.de>
2468         * tree-scalar-evolution.h (final_value_replacement_loop): Update
2469         prototype.
2470         * tree-scalar-evolution.c (final_value_replacement_loop): Return
2471         whether anything was done.
2472         (scev_const_prop): Remove constant propagation part, fold
2473         remains into ...
2474         * tree-ssa-loop.c (pass_scev_cprop::execute): ... here.
2475         (pass_data_scev_cprop): TODO_cleanup_cfg is now done
2476         conditionally.
2478 2018-11-05  Jakub Jelinek  <jakub@redhat.com>
2480         PR tree-optimization/87859
2481         * gimple-ssa-store-merging.c (struct merged_store_group): Add
2482         only_constants and first_nonmergeable_order members.
2483         (merged_store_group::merged_store_group): Initialize them.
2484         (merged_store_group::do_merge): Clear only_constants member if
2485         adding something other than INTEGER_CST store.
2486         (imm_store_chain_info::coalesce_immediate_stores): Don't merge
2487         stores with order >= first_nonmergeable_order.  Use
2488         merged_store->only_constants instead of always recomputing it.
2489         Set merged_store->first_nonmergeable_order if we've skipped any
2490         stores.  Attempt to merge overlapping INTEGER_CST stores that
2491         we would otherwise skip.
2493         PR sanitizer/87837
2494         * match.pd (X + Y < X): Don't optimize if TYPE_OVERFLOW_SANITIZED.
2496 2018-11-05  Xuepeng Guo  <xuepeng.guo@intel.com>
2498         PR target/87853
2499         * config/i386/emmintrin.h (__v16qs): New to cope with option
2500         -funsigned-char.
2501         (_mm_cmpeq_epi8): Replace __v16qi with __v16qs.
2502         (_mm_cmplt_epi8): Likewise.
2503         (_mm_cmpgt_epi8): Likewise.
2505 2018-11-05  Richard Biener  <rguenther@suse.de>
2507         PR rtl-optimization/87852
2508         * fwprop.c (use_killed_between): Only consider single-defs of the
2509         use whose definition statement dominates the use.
2511 2018-11-05  Martin Liska  <mliska@suse.cz>
2513         PR web/87829
2514         * doc/invoke.texi: Remove options that are
2515         not disabled with -Os.
2517 2018-11-05  Martin Liska  <mliska@suse.cz>
2519         PR c/87811
2520         * doc/extend.texi: Update constrain about the last argument
2521         of __builtin_expect_with_probability.
2523 2018-11-05  Martin Liska  <mliska@suse.cz>
2525         PR c/87811
2526         * predict.c (expr_expected_value_1): Verify
2527         that last argument is a real constants and emit
2528         error.
2530 2018-11-05  Martin Liska  <mliska@suse.cz>
2532         PR gcov-profile/77698
2533         * ipa-profile.c (ipa_profile): Adjust hotness threshold
2534         only in LTO mode.
2536 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2538         PR tree-optimization/86572
2539         * builtins.c (c_strlen): Handle negative offsets in a safe way.
2541 2018-11-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2543         PR tree-optimization/87672
2544         * gimple-fold.c (gimple_fold_builtin_stxcpy_chk): Gimplify.
2545         * tree-ssa-strlen.c (handle_builtin_strcat): Adjust object size.
2547 2018-11-04  Uros Bizjak  <ubizjak@gmail.com>
2549         PR middle-end/58372
2550         * cfgexpand.c (pass_expand::execute): Move the call to
2551         finish_eh_generation in front of the call to expand_stack_alignment.
2553 2018-11-04  Venkataramanan Kumar <venkataramanan.kumar@amd.com>
2555         * common/config/i386/i386-common.c (processor_alias_table): Add
2556         znver2 entry.
2557         * config.gcc (i[34567]86-*-linux* | ...): Add znver2.
2558         (case ${target}): Add znver2.
2559         * config/i386/driver-i386.c: (host_detect_local_cpu): Let
2560         -march=native recognize znver2 processors.
2561         * config/i386/i386-c.c (ix86_target_macros_internal): Add znver2.
2562         * config/i386/i386.c (m_znver2): New definition.
2563         (m_ZNVER): New definition.
2564         (m_AMD_MULTIPLE): Includes m_znver2.
2565         (processor_cost_table): Add znver2 entry.
2566         (processor_target_table): Add znver2 entry.
2567         (get_builtin_code_for_version): Set priority for
2568         PROCESSOR_ZNVER2.
2569         (processor_model): Add M_AMDFAM17H_ZNVER2.
2570         (arch_names_table): Ditto.
2571         (ix86_reassociation_width): Include znver2.
2572         * config/i386/i386.h (TARGET_znver2): New definition.
2573         (struct ix86_size_cost): Add TARGET_ZNVER2.
2574         (enum processor_type): Add PROCESSOR_ZNVER2.
2575         * config/i386/i386.md (define_attr "cpu"): Add znver2.
2576         * config/i386/x86-tune-costs.h: (processor_costs) Add znver2 costs.
2577         * config/i386/x86-tune-sched.c: (ix86_issue_rate): Add znver2.
2578         (ix86_adjust_cost): Add znver2.
2579         * config/i386/x86-tune.def:  Replace m_ZNVER1 by m_ZNVER.
2580         * gcc/doc/extend.texi: Add details about znver2.
2581         * gcc/doc/invoke.texi: Add details about znver2.
2583 2018-11-03  Sandra Loosemore  <sandra@codesourcery.com>
2585         PR target/87079
2587         * config/nios2/nios2.c (nios2_rtx_costs): Recognize <mul>sidi3
2588         pattern.
2590 2018-11-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
2592         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Pay
2593         attention to TARGET_AVOID_XFORM and BYTES_BIG_ENDIAN.
2595 2018-11-02  Richard Earnshaw  <rearnsha@arm.com>
2597         * config/aarch64/aarch64.c ((aarch64_override_options): Disable
2598         shrink-wrapping when -mtrack-speculation.
2600 2018-11-02  Richard Biener  <rguenther@suse.de>
2602         * tree-ssa-coalesce.c (struct coalesce_list): Add obstack member.
2603         (pop_cost_one_pair): Do not free pair.
2604         (pop_best_coalesce): Likewise.
2605         (create_coalesce_list): Initialize obstack.
2606         (delete_coalesce_list): Free obstack.
2607         (find_coalesce_pair): Obstack-allocate coalesce pairs.
2608         (add_cost_one_coalesce): Likewise.
2609         (struct live_track): Remove bitmap pointer indirections.
2610         (new_live_track): Adjust.
2611         (delete_live_track): Likewise.
2612         (live_track_remove_partition): Likewise.
2613         (live_track_add_partition): Likewise.
2614         (live_track_live_p): Likewise.
2615         (live_track_process_def): Likewise.
2616         (live_track_clear_base_vars): Likewise.
2618 2018-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
2620         * configure.ac (gcc_cv_as_sparc_register_op): Remove.
2621         * configure: Regenerate.
2622         * config.in: Regenerate.
2623         * config/sparc/sparc.c (sparc_output_scratch_registers): Remove
2624         HAVE_AS_REGISTER_PSEUDO_OP guard.
2625         * config/sparc/sparc.h (ASM_ARCH64_SPEC): Remove
2626         !HAVE_AS_REGISTER_PSEUDO_OP support.
2627         (ASM_DECLARE_REGISTER_GLOBAL): Remove HAVE_AS_REGISTER_PSEUDO_OP
2628         guard.
2630 2018-11-02  Richard Biener  <rguenther@suse.de>
2632         * dwarf2out.c (build_abbrev_table): Guard lookup_external_ref call
2633         with is_type_die.
2635 2018-11-02  Richard Biener  <rguenther@suse.de>
2637         PR tree-optimization/87776
2638         * tree-ssa-sccvn.c (do_rpo_vn): Do not mark backedges
2639         executable when iterating but running into rpo-vn-max-loop-depth
2640         and not eliding the iteration.
2642 2018-11-30  Jan Hubicka  <jh@suse.cz>
2644         * tree.c (free_lang_data_in_decl): Free TREE_TYPE and alignment of
2645         TYPE_DECL.
2647 2018-11-01  Aaron Sawdey  <acsawdey@linux.ibm.com>
2649         * config/rs6000/rs6000-protos.h (rs6000_address_for_fpconvert): Remove
2650         prototype.
2651         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem):
2652         Combine with rs6000_address_for_fpconvert.
2653         (rs6000_address_for_fpconvert): Combine with
2654         rs6000_force_indexed_or_indirect_mem.
2655         (rs6000_expand_vector_init): Change function call from
2656         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2657         * config/rs6000/rs6000.md (floatsi<mode>2_lfiwax): Change call from
2658         rs6000_address_for_fpconvert to rs6000_force_indexed_or_indirect_mem.
2659         (floatsi<mode>2_lfiwax_mem): Ditto.
2660         (floatunssi<mode>2_lfiwzx): Ditto.
2661         (floatunssi<mode>2_lfiwzx_mem): Ditto.
2662         (float<QHI:mode><FP_ISA3:mode>2): Ditto.
2663         (floatuns<QHI:mode><FP_ISA3:mode>2): Ditto.
2664         (fix_trunc<mode>si2_stfiwx): Ditto.
2665         (fixuns_trunc<mode>si2_stfiwx): Ditto.
2666         (float_<mode>si2_hw): Ditto.
2667         (floatuns_<mode>si2_hw): Ditto.
2668         * config/rs6000/vsx.md (*vsx_extract_si): Ditto.
2669         (vsx_splat_<mode>): Ditto.
2671 2018-11-01  Joseph Myers  <joseph@codesourcery.com>
2673         * configure.ac (gcc_cv_as_ia64_ltoffx_ldxmov_relocs): Add newline
2674         at end of assembler input text.
2675         * configure: Regenerate.
2677 2018-11-01  Jakub Jelinek  <jakub@redhat.com>
2679         PR tree-optimization/87826
2680         * gimple-ssa-store-merging.c (do_shift_rotate): Punt if count is
2681         negative or larger or equal to type's precision.
2683 2018-10-31  Alexandre Oliva <aoliva@redhat.com>
2685         * opts.c (default_options_table): Do not enable
2686         OPT_fdelayed_branch at -Og.
2687         * doc/invoke.texi (-fdelayed-branch): Document it.
2689 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2691         * optabs-libfuncs.c (build_libfunc_function_visibility):
2692         New, split out from...
2693         (build_libfunc_function): ... here.
2694         (init_one_libfunc_visibility): New, split out from ...
2695         (init_one_libfunc): ... here.
2697         * config/aarch64/atomics.md (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse):
2698         scratch register need not be early-clobber.  Document the reason
2699         why we cannot use ST<OP>.
2701 2018-10-31  Joseph Myers  <joseph@codesourcery.com>
2703         PR bootstrap/82856
2704         * configure.ac: Remove AC_PREREQ.  Use AC_LANG_SOURCE.  Use single
2705         line for second argument of AC_DEFINE_UNQUOTED.
2706         * doc/install.texi (Tools/packages necessary for modifying GCC):
2707         Update to autoconf 2.69 and automake 1.15.1.
2708         * aclocal.m4, config.in, configure: Regenerate.
2710 2018-10-31  Pat Haugen  <pthaugen@us.ibm.com>
2712         * tree-ssa-loop-ivcanon.c (try_peel_loop): Fix typo and remove dead
2713         initialization.
2715 2018-10-31  Martin Liska  <mliska@suse.cz>
2717         PR driver/83193
2718         * common/config/aarch64/aarch64-common.c (aarch64_parse_extension):
2719         Add new argument invalid_extension.
2720         (aarch64_get_all_extension_candidates): New function.
2721         (aarch64_rewrite_selected_cpu): Add NULL to function call.
2722         * config/aarch64/aarch64-protos.h (aarch64_parse_extension): Add
2723         new argument.
2724         (aarch64_get_all_extension_candidates): New function.
2725         * config/aarch64/aarch64.c (aarch64_parse_arch): Add new
2726         argument invalid_extension.
2727         (aarch64_parse_cpu): Likewise.
2728         (aarch64_print_hint_for_extensions): New function.
2729         (aarch64_validate_mcpu): Provide hint about invalid extension.
2730         (aarch64_validate_march): Likewise.
2731         (aarch64_handle_attr_arch): Pass new argument.
2732         (aarch64_handle_attr_cpu): Provide hint about invalid extension.
2733         (aarch64_handle_attr_isa_flags): Likewise.
2735 2018-10-31  Richard Biener  <rguenther@suse.de>
2737         PR middle-end/70359
2738         PR middle-end/86270
2739         * tree-outof-ssa.c (insert_backedge_copies): Restrict
2740         copy generation to useful cases.  Place the copy before
2741         the definition of the backedge value when possible.
2743 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2745         * config/arc/arc-arch.h (ARC_TUNE_ARC7XX): New tune value.
2746         * config/arc/arc.c (arc_active_insn): New function.
2747         (check_store_cacheline_hazard): Likewise.
2748         (workaround_arc_anomaly): Use check_store_cacheline_hazard.
2749         (arc_override_options): Disable delay slot scheduler for older
2750         A7.
2751         (arc_store_addr_hazard_p): New implementation, old one renamed to
2752         ...
2753         (arc_store_addr_hazard_internal_p): Renamed.
2754         (arc_reorg): Don't combine into brcc instructions which are part
2755         of hardware hazard solution.
2756         * config/arc/arc.md (attr tune): Consider new arc7xx tune value.
2757         (tune_arc700): Likewise.
2758         * config/arc/arc.opt (arc7xx): New tune value.
2759         * config/arc/arc700.md: Improve A7 scheduler.
2761 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2763         * config/arc/arc.c (arc_override_options): Remove
2764         TARGET_COMPACT_CASESI.
2765         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Update.
2766         (CASE_VECTOR_MODE): Likewise.
2767         (CASE_VECTOR_PC_RELATIVE): Likewise.
2768         (CASE_VECTOR_SHORTEN_MODE): Likewise.
2769         (CASE_VECTOR_SHORTEN_MODE1): Delete.
2770         (ADDR_VEC_ALIGN): Update.
2771         (ASM_OUTPUT_CASE_LABEL): Undefine.
2772         (ASM_OUTPUT_BEFORE_CASE_LABEL): Undefine.
2773         (TARGET_BI_BIH): Define.
2774         (DEFAULT_BRANCH_INDEX): Likewise.
2775         * config/arc/arc.md (casesi): Rework to accept BI/BIH
2776         instructions, remove compact_casesi use case.
2777         (casesi_compact_jump): Remove.
2778         (casesi_dispatch): New pattern.
2779         * config/arc/arc.opt: Add mbranch-index option. Deprecate
2780         compact_casesi option.
2781         * doc/invoke.texi: Document mbranch-index option.
2783 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2785         * config/arc/arc.c (arc_get_tp): Remove function.
2786         (arc_emit_call_tls_get_addr): Likewise.
2787         (arc_call_tls_get_addr): New function.
2788         (arc_legitimize_tls_address): Make use of arc_call_tls_get_addr.
2789         * config/arc/arc.md (tls_load_tp_soft): Remove.
2790         (tls_gd_get_addr): Likewise.
2792 2018-10-31  Claudiu Zissulescu  <claziss@synopsys.com>
2794         * config/arc/arc.md (mulsi3): Remove call to mulsi_600_lib.
2795         (mulsi3_600_lib): Remove pattern.
2796         (umulsi3_highpart_600_lib_le): Likewise.
2797         (umulsi3_highpart): Remove call to umulsi3_highpart_600_lib_le.
2798         (umulsidi3): Remove call to umulsidi3_600_lib.
2799         (umulsidi3_600_lib): Remove pattern.
2800         (peephole2): Remove peephole using the above deprecated patterns.
2802 2018-10-31  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
2804         PR target/87374
2805         * config/arm/arm.c (arm_option_check_internal): Disable the combined
2806         use of -mslow-flash-data and -mword-relocations.
2807         (arm_option_override): Enable -mword-relocations if -fpic or -fPIC.
2808         * config/arm/arm.md (SYMBOL_REF MOVT splitter): Stop checking for
2809         flag_pic.
2810         * doc/invoke.texi (-mword-relocations): Mention conflict with
2811         -mslow-flash-data.
2812         (-mslow-flash-data): Reciprocally.
2814 2018-10-31  Richard Henderson  <richard.henderson@linaro.org>
2816         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Force
2817         16-byte modes held in GP registers to use an even regno.
2819         * config/aarch64/aarch64.c (aarch64_emit_bic): Remove.
2820         (aarch64_atomic_ldop_supported_p): Remove.
2821         (aarch64_gen_atomic_ldop): Remove.
2822         * config/aarch64/atomic.md (atomic_<atomic_optab><ALLI>):
2823         Fully expand LSE operations here.
2824         (atomic_fetch_<atomic_optab><ALLI>): Likewise.
2825         (atomic_<atomic_optab>_fetch<ALLI>): Likewise.
2826         (aarch64_atomic_<ATOMIC_LDOP><ALLI>_lse): Drop atomic_op iterator
2827         and use ATOMIC_LDOP instead; use register_operand for the input;
2828         drop the split and emit insns directly.
2829         (aarch64_atomic_fetch_<ATOMIC_LDOP><ALLI>_lse): Likewise.
2830         (aarch64_atomic_<atomic_op>_fetch<ALLI>_lse): Remove.
2831         (@aarch64_atomic_load<ATOMIC_LDOP><ALLI>): Remove.
2833         * config/aarch64/aarch64.c (aarch64_emit_atomic_swap): Remove.
2834         (aarch64_gen_atomic_ldop): Don't call it.
2835         * config/aarch64/atomics.md (atomic_exchange<ALLI>):
2836         Use aarch64_reg_or_zero.
2837         (aarch64_atomic_exchange<ALLI>): Likewise.
2838         (aarch64_atomic_exchange<ALLI>_lse): Remove split; remove & from
2839         operand 0; use aarch64_reg_or_zero for input; merge ...
2840         (@aarch64_atomic_swp<ALLI>): ... this and remove.
2842         * config/aarch64/aarch64.c (aarch64_gen_compare_reg_maybe_ze): New.
2843         (aarch64_split_compare_and_swap): Use it.
2844         (aarch64_expand_compare_and_swap): Likewise.  Remove convert_modes;
2845         test oldval against the proper predicate.
2846         * config/aarch64/atomics.md (@atomic_compare_and_swap<ALLI>):
2847         Use nonmemory_operand for expected.
2848         (cas_short_expected_pred): New.
2849         (@aarch64_compare_and_swap<SHORT>): Use it; use "rn" not "rI" to match.
2850         (@aarch64_compare_and_swap<GPI>): Use "rn" not "rI" for expected.
2851         * config/aarch64/predicates.md (aarch64_plushi_immediate): New.
2852         (aarch64_plushi_operand): New.
2854         * config/aarch64/aarch64.c (aarch64_expand_compare_and_swap):
2855         Force oldval into the rval register for TARGET_LSE; emit the compare
2856         during initial expansion so that it may be deleted if unused.
2857         (aarch64_gen_atomic_cas): Remove.
2858         * config/aarch64/atomics.md (@aarch64_compare_and_swap<SHORT>_lse):
2859         Change =&r to +r for operand 0; use match_dup for operand 2;
2860         remove is_weak and mod_f operands as unused.  Drop the split
2861         and merge with...
2862         (@aarch64_atomic_cas<SHORT>): ... this pattern's output; remove.
2863         (@aarch64_compare_and_swap<GPI>_lse): Similarly.
2864         (@aarch64_atomic_cas<GPI>): Similarly.
2866 2018-10-31  Richard Biener  <rguenther@suse.de>
2868         * tree-eh.c (replace_trapping_overflow): Simplify ABS_EXPR case
2869         using ABSU_EXPR.
2871 2018-10-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2873         * config/aarch64/aarch64-cores.def (saphira): Use saphira pipeline.
2874         * config/aarch64/aarch64.md: Include saphira.md
2875         * config/aarch64/saphira.md: New file for pipeline description.
2877 2018-10-30  Martin Sebor  <msebor@redhat.com>
2879         PR middle-end/87041
2880         * gimple-ssa-sprintf.c (format_directive): Use %G to include
2881         inlining context.
2882         (sprintf_dom_walker::compute_format_length):
2883         Avoid setting POSUNDER4K here.
2884         (get_destination_size): Handle null argument values.
2885         (get_user_idx_format): New function.
2886         (sprintf_dom_walker::handle_gimple_call): Handle all printf-like
2887         functions, including user-defined with attribute format printf.
2888         Use %G to include inlining context.
2889         Set POSUNDER4K here.
2891 2018-10-30  Jan Hubicka  <jh@suse.cz>
2893         * params.def (lto-partitions): Bump from 32 to 128.
2895 2018-10-30  Jan Hubicka  <jh@suse.cz>
2897         * tree.c
2898         (free_lang_data_d, add_tree_to_fld_list, fld_worklist_push): Move
2899         head in file.
2900         (free_lang_data_in_type): Forward declare.
2901         (fld_type_variant_equal_p): New function.
2902         (fld_type_variant): New function
2903         (fld_incomplete_types): New hash.
2904         (fld_incomplete_type_of): New function
2905         (fld_simplfied-type): New function.
2906         (free_lang_data_in_decl): Add fld parameter; simplify type of FIELD_DECL
2907         (free_lang_data): Allocate and free fld_incomplete_type; update call
2908         of free_lang_data_in_decl.
2910 2018-10-30  Eric Botcazou  <ebotcazou@adacore.com>
2912         * gcov.c (output_lines): Remove duplicate line.
2914 2018-10-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
2916         * config/rs6000/rs6000.md (bswapdi2): Force address into register
2917         if not in indexed or indirect form.
2918         (bswapdi2_load): Change predicate to indexed_or_indirect_operand.
2919         (bswapdi2_store): Ditto.
2920         * config/rs6000/rs6000.c (rs6000_force_indexed_or_indirect_mem): New
2921         helper function.
2922         * config/rs6000/rs6000-protos.h (rs6000_force_indexed_or_indirect_mem):
2923         Prototype for helper function.
2925 2018-10-30  Martin Sebor  <msebor@redhat.com>
2927         * doc/extend.texi (optimize): Clarify/expand attribute documentation.
2928         (target, pragma GCC optimize, pragma GCC target): Ditto.
2930 2018-10-30  Jonathan Wakely  <jwakely@redhat.com>
2932         * doc/extend.texi: Fix prototype and description of
2933         __builtin_expect_with_probability.
2935 2018-10-30  Michael Ploujnikov  <michael.ploujnikov@oracle.com>
2937         * cgraph.h (clone_function_name_1): Replaced by new
2938           clone_function_name_numbered that takes name as string; for
2939           privatize_symbol_name_1 use only.
2940           (clone_function_name): Renamed to
2941           clone_function_name_numbered to be explicit about numbering.
2942           (clone_function_name): New two-argument function that does
2943           not number its output.
2944           (clone_function_name): New three-argument function that
2945           takes a number to append to its output.
2946         * cgraphclones.c (duplicate_thunk_for_node):
2947           (clone_function_name_1): Renamed.
2948           (clone_function_name_numbered): Two new functions.
2949           (clone_function_name): Improved documentation.
2950           (cgraph_node::create_virtual_clone): Use clone_function_name_numbered.
2951         * config/rs6000/rs6000.c (make_resolver_func): Ditto.
2952         * final.c (final_scan_insn_1): Use the new clone_function_name
2953           without numbering.
2954         * multiple_target.c (create_dispatcher_calls): Ditto.
2955           (create_target_clone): Ditto.
2956         * omp-expand.c (grid_expand_target_grid_body): Ditto.
2957         * omp-low.c (create_omp_child_function_name): Ditto.
2958         * omp-simd-clone.c (simd_clone_create): Ditto.
2959         * symtab.c (simd_symtab_node::noninterposable_alias): Use the
2960           new clone_function_name without numbering.
2962 2018-10-30  Richard Earnshaw  <rearnsha@arm.com>
2964         * alloc-pool.h (base_pool_allocator <TBlockAllocator>::initialize):
2965         Assert that the allocation size is not zero.
2967 2018-10-30  Richard Biener  <rguenther@suse.de>
2969         PR tree-optimization/87800
2970         * tree-vect-slp.c (vect_build_slp_tree_2): Reject any
2971         non-induction or reduction PHIs.
2973 2018-10-30 Sameera Deshpande <sameera.deshpande@linaro.org
2975         * config/aarch64/falkor-tag-collision-avoidance.c
2976         (execute_tag_collision_avoidance): Call df_note_add_problem.
2978 2018-10-30  Martin Liska  <mliska@suse.cz>
2980         * doc/extend.texi: Fix typo in documentation
2981         of __builtin_expect_with_probability.
2983 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2985         PR c++/87721
2986         * input.c (get_substring_ranges_for_loc): Detect if
2987         linemap_resolve_location gives us a NULL map, and reject
2988         this case.
2990 2018-10-29  Iain Buclaw  <ibuclaw@gdcproject.org>
2992         * config.gcc (xstormy16-*-elf): Set tm_d_file.
2994 2018-10-29  David Malcolm  <dmalcolm@redhat.com>
2995             Martin Sebor  <msebor@redhat.com>
2996             Manuel Lopez-Ibanez  <manu@gcc.gnu.org>
2998         * doc/ux.texi (Quoting): New subsection, adapted from material at
2999         https://gcc.gnu.org/wiki/DiagnosticsGuidelines written by
3000         MartinSebor and ManuelLopezIbanez.
3001         (Fix-it hints): Note that fix-it hints shouldn't be marked for
3002         translation.
3004 2018-10-29  Kugan Vivekanandarajah  <kuganv@linaro.org>
3006         PR middle-end/87469
3007         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Fix niter
3008         max value.
3010 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
3012         * gcc.target/powerpc/ssse3-check.h: Remove inaccurate comment.
3014 2018-10-29  Paul Koning  <ni1d@arrl.net>
3016         * doc/rtl.texi (CONST_WIDE_INT_ELT): Give correct macro name.
3018 2018-10-29  Paul A. Clarke  <pc@us.ibm.com>
3020         * gcc/config/rs6000/mmintrin.h (_mm_packs_pi16, _mm_packs_pi32,
3021         _mm_packs_pu16, _mm_unpackhi_pi8, _mm_unpacklo_pi8, _mm_add_pi8,
3022         _mm_add_pi16, _mm_add_pi32, _mm_sub_pi8, _mm_sub_pi16, _mm_sub_pi32,
3023         _mm_cmpgt_pi8, _mm_cmpeq_pi16, _mm_cmpgt_pi16, _mm_cmpeq_pi32,
3024         _mm_cmpgt_pi32, _mm_adds_pi8, _mm_adds_pi16, _mm_adds_pu8,
3025         _mm_adds_pu16, _mm_subs_pi8, _mm_subs_pi16, _mm_subs_pu8,
3026         _mm_subs_pu16, _mm_madd_pi16, _mm_mulhi_pi16, _mm_mullo_pi16,
3027         _mm_sll_pi16, _mm_sra_pi16, _mm_srl_pi16, _mm_set1_pi16, _mm_set1_pi8):
3028         Change 'vector' to '__vector'.
3029         * gcc/config/rs6000/xmmintrin.h (_mm_cvtps_pi32, _mm_cvttps_pi32,
3030         _mm_cvtps_pi16, _mm_cvtps_pi8, _mm_max_pi16, _mm_max_pu8, _mm_min_pi16,
3031         _mm_min_pu8, _mm_mulhi_pu16, _mm_shuffle_pi16, _mm_avg_pu8,
3032         _mm_avg_pu16): Likewise.  And, whitespace corrections.
3034 2018-10-29  Richard Biener  <rguenther@suse.de>
3036         PR tree-optimization/87785
3037         * tree-vect-slp.c (vect_gather_slp_loads): Only gather
3038         internal defs.
3040 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3042         * VXWORKS_NET_LIBS_RTP: New macro, network part of VXWORKS_LIBS_RTP.
3044 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3046         * config/vxworks.h (VXWORKS_PERSONALITY): New VxWorks
3047         ports configuration macro, defaults to "gnu".
3048         (VXWORKS_OS_CPP_BUILTINS): Use it. Feed TOOL and
3049         TOOL_FAMILY instead of _VX_TOOL/_VXTOOL_FAMILY.
3051 2018-10-29  Olivier Hainque  <hainque@adacore.com>
3053         * config/vxworks.h (CLEAR_INSN_CACHE): #undef before #define.
3055 2018-10-29  Richard Biener  <rguenther@suse.de>
3057         PR tree-optimization/87790
3058         * tree-vect-slp.c (vect_mark_slp_stmts): Simplify.
3059         (vect_make_slp_decision): Adjust.
3060         (vect_slp_analyze_bb_1): Likewise.
3061         (vect_detect_hybrid_slp_stmts): Properly union SLP type over
3062         edges.
3064 2018-10-29  Richard Biener  <rguenther@suse.de>
3066         PR tree-optimization/87785
3067         * tree-vect-slp.c (vect_build_slp_tree_2): Remove loads argument
3068         and processing.
3069         (vect_build_slp_tree): Likewise.
3070         (vect_gather_slp_loads): New function.
3071         (vect_analyze_slp_instance): Gather loads separately from the
3072         SLP tree build.
3074 2018-10-29  Martin Liska  <mliska@suse.cz>
3076         * Makefile.in: Make dependency to json.o.
3077         * doc/gcov.texi: Document new JSON format, remove
3078         old intermediate format documentation.
3079         * gcov.c (struct function_info): Come up with m_name and
3080         m_demangled_name.
3081         (function_info::function_info): Initialize it.
3082         (function_info::~function_info): Release it.
3083         (main): Rename flag_intermediate_format to flag_json_format.
3084         (print_usage): Describe --json-format.
3085         (process_args): Set flag_json_format.
3086         (output_intermediate_line): Remove.
3087         (output_intermediate_json_line): Likewise.
3088         (get_gcov_intermediate_filename): Return new extension
3089         ".gcov.json.gz".
3090         (output_intermediate_file): Implement JSON emission.
3091         (output_json_intermediate_file): Implement JSON emission.
3092         (generate_results): Use ::get_name for function name.
3093         Handle JSON output file.
3094         (read_graph_file): Use ::get_name instead of cplus_demangle.
3095         (read_count_file): Likewise.
3096         (solve_flow_graph): Likewise.
3097         (add_line_counts): Likewise.
3098         (accumulate_line_counts): Use new flag_json_format.
3099         (output_function_details): Use ::get_name instead of cplus_demangle.
3100         (output_lines): Likewise.
3101         * json.cc (test_writing_literals): Add new tests.
3102         * json.h (class literal): Add new boolean constructor.
3104 2018-10-29  Segher Boessenkool  <segher@kernel.crashing.org>
3106         PR rtl-optimization/87701
3107         PR rtl-optimization/87780
3108         * combine.c (make_more_copies): Rewrite.
3110 2018-10-28  Kugan Vivekanandarajah  <kuganv@linaro.org>
3112         * doc/generic.texi (ABSU_EXPR): Document.
3113         * match.pd (absu(x)*absu(x) -> x*x): Handle.
3114         (absu(absu(X)) -> absu(X)): Likewise.
3115         (absu(-X) -> absu(X)): Likewise.
3116         (absu(X)  where X is nonnegative -> X): Likewise.
3118 2018-10-28  Iain Buclaw  <ibuclaw@gdcproject.org>
3120         * Makefile.in (tm_d_file_list, tm_d_include_list): New variables.
3121         (TM_D_H, D_TARGET_DEF, D_TARGET_H, D_TARGET_OBJS): New variables.
3122         (tm_d.h, cs-tm_d.h, default-d.o): New rules.
3123         (d/d-target-hooks-def.h, s-d-target-hooks-def-h): New rules.
3124         (s-tm-texi): Also check timestamp on d-target.def.
3125         (generated_files): Add TM_D_H and d-target-hooks-def.h.
3126         (build/genhooks.o): Also depend on D_TARGET_DEF.
3127         * config.gcc (tm_d_file, d_target_objs, target_has_targetdm): New
3128         variables.
3129         * config/aarch64/aarch64-d.c: New file.
3130         * config/aarch64/aarch64-linux.h (GNU_USER_TARGET_D_CRITSEC_SIZE):
3131         Define.
3132         * config/aarch64/aarch64-protos.h (aarch64_d_target_versions): New
3133         prototype.
3134         * config/aarch64/aarch64.h (TARGET_D_CPU_VERSIONS): Define.
3135         * config/aarch64/t-aarch64 (aarch64-d.o): New rule.
3136         * config/arm/arm-d.c: New file.
3137         * config/arm/arm-protos.h (arm_d_target_versions): New prototype.
3138         * config/arm/arm.h (TARGET_D_CPU_VERSIONS): Define.
3139         * config/arm/linux-eabi.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3140         * config/arm/t-arm (arm-d.o): New rule.
3141         * config/default-d.c: New file.
3142         * config/glibc-d.c: New file.
3143         * config/gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3144         * config/i386/i386-d.c: New file.
3145         * config/i386/i386-protos.h (ix86_d_target_versions): New prototype.
3146         * config/i386/i386.h (TARGET_D_CPU_VERSIONS): Define.
3147         * config/i386/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3148         (GNU_USER_TARGET_D_CRITSEC_SIZE): Define.
3149         * config/i386/t-i386 (i386-d.o): New rule.
3150         * config/kfreebsd-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3151         * config/kopensolaris-gnu.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3152         * config/linux-android.h (ANDROID_TARGET_D_OS_VERSIONS): Define.
3153         * config/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3154         * config/mips/linux-common.h (EXTRA_TARGET_D_OS_VERSIONS): Define.
3155         * config/mips/mips-d.c: New file.
3156         * config/mips/mips-protos.h (mips_d_target_versions): New prototype.
3157         * config/mips/mips.h (TARGET_D_CPU_VERSIONS): Define.
3158         * config/mips/t-mips (mips-d.o): New rule.
3159         * config/powerpcspe/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3160         * config/powerpcspe/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3161         * config/powerpcspe/powerpcspe-d.c: New file.
3162         * config/powerpcspe/powerpcspe-protos.h (rs6000_d_target_versions):
3163         New prototype.
3164         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
3165         Support GNU D by using 0 as the language type.
3166         * config/powerpcspe/powerpcspe.h (TARGET_D_CPU_VERSIONS): Define.
3167         * config/powerpcspe/t-powerpcspe (powerpcspe-d.o): New rule.
3168         * config/riscv/riscv-d.c: New file.
3169         * config/riscv/riscv-protos.h (riscv_d_target_versions): New
3170         prototype.
3171         * config/riscv/riscv.h (TARGET_D_CPU_VERSIONS): Define.
3172         * config/riscv/t-riscv (riscv-d.o): New rule.
3173         * config/rs6000/linux.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3174         * config/rs6000/linux64.h (GNU_USER_TARGET_D_OS_VERSIONS): Define.
3175         * config/rs6000/rs6000-d.c: New file.
3176         * config/rs6000/rs6000-protos.h (rs6000_d_target_versions): New
3177         prototype.
3178         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
3179         Support GNU D by using 0 as the language type.
3180         * config/rs6000/rs6000.h (TARGET_D_CPU_VERSIONS): Define.
3181         * config/rs6000/t-rs6000 (rs6000-d.o): New rule.
3182         * config/s390/s390-d.c: New file.
3183         * config/s390/s390-protos.h (s390_d_target_versions): New prototype.
3184         * config/s390/s390.h (TARGET_D_CPU_VERSIONS): Define.
3185         * config/s390/t-s390 (s390-d.o): New rule.
3186         * config/sparc/sparc-d.c: New file.
3187         * config/sparc/sparc-protos.h (sparc_d_target_versions): New
3188         prototype.
3189         * config/sparc/sparc.h (TARGET_D_CPU_VERSIONS): Define.
3190         * config/sparc/t-sparc (sparc-d.o): New rule.
3191         * config/t-glibc (glibc-d.o): New rule.
3192         * configure: Regenerated.
3193         * configure.ac (tm_d_file): New variable.
3194         (tm_d_file_list, tm_d_include_list, d_target_objs): Add substitutes.
3195         * doc/contrib.texi (Contributors): Add self for the D frontend.
3196         * doc/frontends.texi (G++ and GCC): Mention D as a supported language.
3197         * doc/install.texi (Configuration): Mention libphobos as an option for
3198         --enable-shared.  Mention d as an option for --enable-languages.
3199         (Testing): Mention check-d as a target.
3200         * doc/invoke.texi (Overall Options): Mention .d, .dd, and .di as file
3201         name suffixes.  Mention d as a -x option.
3202         * doc/sourcebuild.texi (Top Level): Mention libphobos.
3203         * doc/standards.texi (Standards): Add section on D language.
3204         * doc/tm.texi: Regenerated.
3205         * doc/tm.texi.in: Add @node for D language and ABI, and @hook for
3206         TARGET_CPU_VERSIONS, TARGET_D_OS_VERSIONS, and TARGET_D_CRITSEC_SIZE.
3207         * dwarf2out.c (is_dlang): New function.
3208         (gen_compile_unit_die): Use DW_LANG_D for D.
3209         (declare_in_namespace): Return module die for D, instead of adding
3210         extra declarations into the namespace.
3211         (gen_namespace_die): Generate DW_TAG_module for D.
3212         (gen_decl_die): Handle CONST_DECLSs for D.
3213         (dwarf2out_decl): Likewise.
3214         (prune_unused_types_walk_local_classes): Handle DW_tag_interface_type.
3215         (prune_unused_types_walk): Handle DW_tag_interface_type same as other
3216         kinds of aggregates.
3217         * gcc.c (default_compilers): Add entries for .d, .dd and .di.
3218         * genhooks.c: Include d/d-target.def.
3220 2018-10-28  Iain Sandoe  <iain@sandoe.co.uk>
3222         PR target/85669
3223         * config/rs6000/darwin.h (STACK_BOUNDARY): New.
3224         (RS6000_STARTING_FRAME_OFFSET): Adjust to preserve 16byte alignment.
3225         (STACK_DYNAMIC_OFFSET): Likewise.
3227 2018-10-27  Sandra Loosemore  <sandra@codesourcery.com>
3229         PR target/80024
3230         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Fix
3231         error message.
3233 2018-10-26  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>
3235         * doc/extend.texi (PowerPC builtins): Fix __builtin_unpack_ibm128
3236         return type and other typos.
3238 2018-10-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
3240         * config/rs6000/rs6000-string.c (expand_strncmp_gpr_sequence): Change to
3241         a shorter sequence with fewer branches.
3242         (emit_final_str_compare_gpr): Ditto.
3244 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3246         * config/rs6000/tmmintrin.h: New file.
3247         * config.gcc (powerpc*-*-*): Add tmmintrin.h to extra_headers.
3249 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3251         * config/rs6000/mmintrin.h: Enable 32bit compilation.
3252         * config/rs6000/xmmintrin.h: Likewise.
3254 2018-10-26  Paul A. Clarke  <pc@us.ibm.com>
3256         * config/rs6000/xmmintrin.h (_mm_extract_pi16): Fix for big-endian.
3258 2018-10-26  Richard Biener  <rguenther@suse.de>
3260         * tree-vect-slp.c (vect_mark_slp_stmts): Add visited hash_set
3261         and wrapper.
3262         (vect_mark_slp_stmts_relevant): Likewise.
3263         (vect_detect_hybrid_slp_stmts): Likewise.
3264         (vect_bb_slp_scalar_cost): Likewise.
3265         (vect_remove_slp_scalar_calls): Likewise.
3267 2018-10-26  Jan Hubicka  <jh@suse.cz>
3269         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix recursion.
3270         (warn_types_mismatch): Fix walk of DECL_NAME.
3271         (odr_types_equivalent_p): Fix overactive assert.
3273 2018-10-26  Richard Biener  <rguenther@suse.de>
3275         PR tree-optimization/87105
3276         * tree-vectorizer.h (_slp_tree::refcnt): New member.
3277         * tree-vect-slp.c (vect_free_slp_tree): Decrement and honor
3278         refcnt.
3279         (vect_create_new_slp_node): Initialize refcnt to one.
3280         (bst_traits): Move.
3281         (scalar_stmts_set_t, bst_fail): Remove.
3282         (vect_build_slp_tree_2): Add bst_map argument and adjust calls.
3283         (vect_build_slp_tree): Add bst_map argument and lookup
3284         already created SLP nodes.
3285         (vect_print_slp_tree): Handle a SLP graph, print SLP node
3286         addresses.
3287         (vect_slp_rearrange_stmts): Handle a SLP graph.
3288         (vect_analyze_slp_instance): Adjust and free SLP nodes from
3289         the CSE map.  Fix indenting.
3290         (vect_schedule_slp_instance): Add short-cut.
3292 2018-10-26  Martin Liska  <mliska@suse.cz>
3294         PR testsuite/86158
3295         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Hash with
3296         addr_expr and not with pointers.
3298 2018-10-26  Jan Hubicka  <jh@suse.cz>
3300         * tree.c (free_lang_data_in_type): Only check main variants.
3301         * ipa-devirt.c (warn_odr): Make static.
3302         (types_same_for_odr): Drop strict variant.
3303         (types_odr_comparable): Likewise.
3304         (odr_or_derived_type_p): Look for main variants.
3305         (odr_name_hasher::equal): Cleanup comment.
3306         (odr_subtypes_equivalent): Add warn and warned arguments; check main
3307         variants.
3308         (type_variants_equivalent_p): break out from ...
3309         (odr_types_equivalent): ... here; go for main variants where needed.
3310         (warn_odr): ... here; turn static.
3311         (warn_types_mismatch): Compare mangled names of main variants.
3312         * ipa-utils.h (types_odr_comparable): Drop strict parameter.
3313         (type_with_linkage_p): Sanity check that we look at main variant.
3314         * lto.c (lto_read_decls): Only consider main variant to be ODR type.
3315         * tree.h (types_same_for_odr): Drop strict argument.
3317 2018-10-26  Richard Biener  <rguenther@suse.de>
3319         PR tree-optimization/87746
3320         * tree-vect-data-refs.c (vect_update_misalignment_for_peel):
3321         Simplify and fix WRT strided store groups with size not
3322         equal to step in element count.
3323         (vect_analyze_group_access_1): Dump the whole group.
3325 2018-10-25  Carl Love  <cel@us.ibm.com>
3327         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_VSCEDPGT,
3328         P9V_BUILTIN_VEC_VSCEDPLT, P9V_BUILTIN_VEC_VSCEDPEQ,
3329         P9V_BUILTIN_VEC_VSCEDPUO): Rename base overloaded name.  Add quad
3330         precicion entry for each overloaded builtin.
3331         * config/rs6000/rs6000-builtin.def (VSCEDPGT, VSCEDPLT, VSCEDPEQ,
3332         VSCEDPUO): Rename overloaded name.
3333         (VSCEDPGT, VSCEQPGT, VSCEDPLT, VSCEQPLT, VSCEDPEQ, VSCEQPEQ,
3334         VSCEDPUO, VSCEQPUO): Add defitions for overloaded builtins.
3335         * config/rs6000/vsx.md (xscmpexpqp_<code>_<mode>): Add
3336         define_expand for xscmpexqp instruction.
3337         (*xscmpexpqp): Add define_insn for the xscmpexqp instruction.
3339 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3340             Jinsong Ji <jji@us.ibm.com>
3342         * config/rs6000/emmintrin.h (_mm_slli_epi16): Replace deprecated
3343         function with vec_sl.
3344         (_mm_slli_epi32): Likewise.
3345         (_mm_slli_epi64): Likewise.
3346         (_mm_srai_epi16): Replace deprecated function with vec_sra.
3347         (_mm_srai_epi32): Likewise.
3348         (_mm_srli_epi16): Replace deprecated function with vec_sr.
3349         (_mm_srli_epi32): Likewise.
3350         (_mm_srli_epi64): Likewise.
3351         (_mm_sll_epi16): Replace deprecated function with vec_sl.
3352         (_mm_sll_epi32): Likewise.
3353         (_mm_sll_epi64): Likewise.
3354         (_mm_sra_epi16): Replace deprecated function with vec_sra.
3355         (_mm_sra_epi32): Likewise.
3356         (_mm_srl_epi16): Replace deprecated function with vec_sr.
3357         (_mm_srl_epi32): Likewise.
3358         (_mm_srl_epi64): Likewise.
3360 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3361             Jinsong Ji <jji@us.ibm.com>
3363         * gcc/config/rs6000/emmintrin.h (_mm_sll_epi16): Replace
3364         comparison operators with vec_cmp* for compatibility due to
3365         unfortunate history; clean up formatting and use types more
3366         appropriately.
3367         (_mm_sll_epi32): Likewise.
3368         (_mm_sll_epi64): Likewise.
3369         (_mm_srl_epi16): Likewise.
3370         (_mm_srl_epi32): Likewise.
3371         (_mm_srl_epi64): Likewise.
3373 2018-10-25  Bill Schmidt  <wschmidt@linux.ibm.com>
3374             Jinsong Ji <jji@us.ibm.com>
3376         * config/rs6000/emmintrin.h (_mm_sll_epi64): Remove wrong cast.
3377         * config/rs6000/xmmintrin.h (_mm_min_ps): Change m's type to
3378         __vector __bool int.  Use vec_cmpgt in preference to deprecated
3379         function vec_vcmpgtfp.
3380         (_mm_max_ps): Likewise.
3382 2018-10-25  Jeff Law  <law@redhat.com>
3384         * config/rl78/rl78.c (insn_ok_now): Always re-recognize the insn
3385         if returning false.
3387 2018-10-25  Martin Sebor  <msebor@redhat.com>
3389         * doc/extend.texi (aligned): Expand attribute description.
3390         (Alignment): Rename section.  Discuss function arguments.
3392 2018-10-25  Jan Hubicka  <jh@suse.cz>
3394         * ipa-devirt.c (main_odr_variant): Remove.
3395         (hash_odr_name, types_same_for_odr, types_odr_comparable,
3396         odr_name_hasher::equal, odr_subtypes_equivalent_p):
3397         Drop use of main_odr_variant.
3398         (add_type_duplicate): Silence confused warnings on integer types.
3399         (get_odr_type): Always look for main variant.
3400         (register_odr_type): Simplify.
3402 2018-10-25  Richard Biener  <rguenther@suse.de>
3404         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses):
3405         Initialize ng to silence error with release checking bootstrap.
3407 2018-10-25  Richard Biener  <rguenther@suse.de>
3409         * tree-if-conv.c: Include tree-ssa-sccvn.h.
3410         (tree_if_conversion): Run CSE on the if-converted loop body.
3412 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3414         * config/s390/constraints.md (ZL): New constraint.
3415         * config/s390/s390.c (legitimate_pic_operand_p): Accept LARL
3416         operands.
3417         * config/s390/s390.md (movdi_larl): Remove.
3418         (movdi_64): Add the LARL alternative.
3420 2018-10-25  Ilya Leoshkevich  <iii@linux.ibm.com>
3422         PR bootstrap/87747
3423         * rtl.c (RTX_CODE_HWINT_P_1): New helper macro.
3424         (RTX_CODE_HWINT_P): New macro.
3425         (rtx_code_size): Use RTX_CODE_HWINT_P ().
3427 2018-10-25  Jan Hubicka  <jh@suse.cz>
3429         * ipa-devirt.c (odr_types_equivalent_p): Do not ICE if one of types
3430         is anonymous.
3432 2018-10-25  Richard Biener  <rguenther@suse.de>
3434         PR tree-optimization/87665
3435         PR tree-optimization/87745
3436         * tree-vectorizer.h (get_earlier_stmt): Remove.
3437         (get_later_stmt): Pick up UID from the original non-pattern stmt.
3439 2018-10-25  Sam Tebbs  <sam.tebbs@arm.com>
3441         * options.texi (Deprecated): Move list to Var section.
3443 2018-10-24  Bill Schmidt  <wschmidt@linux.ibm.com>
3444             Jinsong Ji <jji@us.ibm.com>
3446         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Change deprecated
3447         __vector long to __vector long long.
3448         (_mm_cvtpd_ps): Likewise.
3449         (_mm_cvttpd_epi32): Likewise.
3450         (_mm_cvtpi32_pd): Likewise.
3451         (_mm_unpackhi_epi64): Likewise.
3452         (_mm_unpacklo_epi64): Likewise.
3454 2018-10-24  Segher Boessenkool  <segher@kernel.crashing.org>
3456         PR rtl-optimization/87720
3457         * combine.c (make_more_copies): Skip if the dest is pc_rtx.
3459 2018-10-24  Alexandre Oliva <aoliva@redhat.com>
3461         * gimple-ssa-isolate-paths.c
3462         (find_implicit_erroneous_behavior): Do not change code if the
3463         pass is running for warnings only.
3464         (find_explicit_erroneous_behavior): Likewise.
3466 2018-10-24  Michael Meissner  <meissner@linux.ibm.com>
3468         * config/rs6000/rs6000.c (TARGET_MANGLE_DECL_ASSEMBLER_NAME):
3469         Define as rs6000_mangle_decl_assembler_name.
3470         (rs6000_mangle_decl_assembler_name): If the user switched from IBM
3471         long double to IEEE long double, switch the names of the long
3472         double built-in functions to be <func>f128 instead of <func>l.
3474 2018-10-24  Martin Sebor  <msebor@redhat.com>
3476         * doc/extend.texi (nonnull): List no-argument form.  Reference
3477         -fno-delete-null-pointer-checks and -fisolate-erroneous-paths-attribute.
3479 2018-10-24  Richard Biener  <rguenther@suse.de>
3481         * tree-ssa-sccvn.c (do_rpo_vn): Free rpo_state.
3483 2018-10-24  Martin Liska  <mliska@suse.cz>
3485         PR tree-optimization/84436
3486         * tree-switch-conversion.c (switch_conversion::contains_same_values_p):
3487         Remove.
3488         (switch_conversion::contains_linear_function_p): New.
3489         (switch_conversion::build_one_array): Support linear
3490         transformation on input.
3491         * tree-switch-conversion.h (struct switch_conversion): Add
3492         contains_linear_function_p declaration.
3494 2018-10-24  Richard Biener  <rguenther@suse.de>
3496         * varasm.c (const_hash_1): Return hash of ADDR_EXPR
3497         if its argument is CONSTANT_CLASS_P.
3499 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
3501         * ipa-utils.h (type_with_linkage_p): No longer check for TYPE_STUB_DECL;
3502         it is wrong for forward declarations.
3504 2018-10-24  Ilya Leoshkevich  <iii@linux.ibm.com>
3506         * config/s390/s390.c (s390_check_qrst_address): Add the missing
3507         SYMBOL_REF_P () check.
3509 2018-10-24  Richard Biener  <rguenther@suse.de>
3511         PR tree-optimization/87105
3512         * tree-vect-data-refs.c (vect_analyze_group_access_1): Adjust
3513         dump classification.
3514         (vect_analyze_data_ref_accesses): Handle duplicate loads and
3515         stores by splitting the affected group after the fact.
3516         * tree-vect-slp.c (vect_build_slp_tree_2): Dump when we
3517         fail the SLP build because of size constraints.
3519 2018-10-24  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
3521         * configure.ac (gcc_cv_ld_aligned_shf_merge): New test.
3522         * configure: Regenerate.
3523         * config.in: Regenerate.
3524         * varasm.c (mergeable_string_section): Use readonly_data_section
3525         if linker doesn't support SHF_MERGE with alignment > 8.
3526         (mergeable_constant_section): Likewise.
3528 2018-10-24  Richard Biener  <rguenther@suse.de>
3530         PR tree-optimization/84013
3531         * tree-ssa-structalias.c (struct msdi_data): New struct for
3532         marshalling data to walk_stmt_load_store_ops.
3533         (maybe_set_dependence_info): Refactor as callback for
3534         walk_stmt_load_store_ops.
3535         (compute_dependence_clique): Set restrict info on all stmt kinds.
3537 2018-10-24  Martin Liska  <mliska@suse.cz>
3539         * cgraph.c (cgraph_node::dump):
3540         Remove reduntant dumps and make tp_first_run dump more compact.
3542 2018-10-24  Richard Biener  <rguenther@suse.de>
3544         PR tree-optimization/87665
3545         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Adjust
3546         to reflect reality.
3548 2018-10-12  Jeff Law  <law@redhat.com>
3550         * config/h8300/h8300.c (h8300_expand_prologue): Fix stm generation
3551         for H8/S.
3553 2018-10-23  Richard Biener  <rguenther@suse.de>
3555         * tree-vrp.c (add_assert_info): Guard dump_printf with
3556         dump_enabled_p.
3557         * gimple-ssa-evrp-analyze.c
3558         (evrp_range_analyzer::record_ranges_from_incoming_edge):
3559         Use value_range::ignore_equivs_equal_p.
3561 2018-10-23  Richard Biener  <rguenther@suse.de>
3563         PR tree-optimization/87105
3564         PR tree-optimization/87608
3565         * passes.def (pass_all_early_optimizations): Add early phi-opt
3566         after dce.
3567         * tree-ssa-phiopt.c (value_replacement): Ignore NOPs and predicts in
3568         addition to debug stmts.
3569         (tree_ssa_phiopt_worker): Add early_p argument, do only min/max
3570         and abs replacement early.
3571         * tree-cfg.c (gimple_empty_block_p): Likewise.
3573 2018-10-23  Richard Earnshaw  <rearnsha@arm.com>
3575         PR target/86383
3576         * config.gcc (arm*-*-netbsdelf*): Default to StrongARM if no CPU
3577         specified to configure.
3578         (arm*-*-*): Use ARM7TDMI as the target CPU if no default provided.
3580 2018-10-23  Richard Biener  <rguenther@suse.de>
3582         PR tree-optimization/87700
3583         * tree-ssa-copy.c (set_copy_of_val): Fix change detection logic.
3585 2018-10-23  Jakub Jelinek  <jakub@redhat.com>
3587         PR target/87674
3588         * config/i386/avx512vlintrin.h (_mm_mask_mullo_epi32): Change type of
3589         second argument from __mmask16 to __mmask8.
3590         * config/i386/avx512vlbwintrin.h (_mm_mask_packus_epi32,
3591         _mm_mask_packs_epi32): Likewise.
3592         * config/i386/avx512pfintrin.h (_mm512_mask_prefetch_i64scatter_ps):
3593         Likewise.
3594         (_mm512_mask_prefetch_i64scatter_pd): Likewise.  Formatting fix.
3596 2018-10-23  Richard Biener  <rguenther@suse.de>
3598         * tree-vect-stmts.c (vect_analyze_stmt): Fix typo in comment.
3600 2018-10-23  Richard Biener  <rguenther@suse.de>
3602         PR tree-optimization/86144
3603         * tree-vect-stmts.c (vect_analyze_stmt): Prefer -mveclibabi
3604         over simd attribute.
3606 2018-10-23  Richard Biener  <rguenther@suse.de>
3608         PR tree-optimization/87693
3609         * tree-ssa-threadedge.c (thread_around_empty_blocks): Handle
3610         the case we do not find the taken edge.
3612 2018-10-22  Bill Schmidt  <wschmidt@linux.ibm.com>
3613             Jinsong Ji  <jji@us.ibm.com>
3615         * config/rs6000/emmintrin.h (_MM_SHUFFLE2): Comment cleanup.
3616         (_mm_store_pd): Use unaligned vector type for pointer cast.
3617         (_mm_maskmoveu_si128): Likewise.
3618         * config/rs6000/xmmintrin.h (__m128_u): New typedef.
3619         (_mm_store_ps): Use unaligned vector type for pointer cast.
3621 2018-10-22  Paul Koning  <ni1d@arrl.net>
3623         * symtab.c (symtab_node::increase_alignment): Correct max
3624         alignment check.
3626 2018-10-22  Yury Gribov  <tetra2005@gmail.com>
3628         PR tree-optimization/87633
3629         * match.pd: Do not generate unordered integer comparisons.
3631 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3633         PR rtl-optimization/87600
3634         * combine.c: Add include of expr.h.
3635         (cant_combine_insn_p): Do not combine moves from any hard non-fixed
3636         register to a pseudo.
3637         (make_more_copies): New function, add a copy to a new pseudo after
3638         the moves from hard registers into pseudos.
3639         (rest_of_handle_combine): Declare rebuild_jump_labels_after_combine
3640         later.  Call make_more_copies.
3642 2018-10-22  Andrew Stubbs  <ams@codesourcery.com>
3644         * lra-constraints.c (process_alt_operands): New local array,
3645         matching_early_clobber.  Check matching_early_clobber before
3646         decrementing reject, and set matching_early_clobber after.
3648 2018-10-22  Segher Boessenkool  <segher@kernel.crashing.org>
3650         PR target/87598
3651         * config/rs6000/rs6000.c (print_operand_address): For unexpected RTL
3652         call output_addr_const and hope for the best.
3654 2018-10-22  Richard Biener  <rguenther@suse.de>
3656         * gimple-ssa-evrp-analyze.c
3657         (evrp_range_analyzer::record_ranges_from_incoming_edge): Be
3658         smarter about what ranges to use.
3659         * tree-vrp.c (add_assert_info): Dump here.
3660         (register_edge_assert_for_2): Instead of here at multiple but
3661         not all places.
3663         * gcc.dg/tree-ssa/evrp12.c: New testcase.
3664         * gcc.dg/predict-6.c: Adjust.
3665         * gcc.dg/tree-ssa/vrp33.c: Disable EVRP.
3666         * gcc.dg/tree-ssa/vrp02.c: Likewise.
3667         * gcc.dg/tree-ssa/cunroll-9.c: Likewise.
3669 2018-10-22  Steven Bosscher <steven@gcc.gnu.org>
3670         Richard Biener  <rguenther@suse.de>
3672         * bitmap.h: Update data structure documentation, including a
3673         description of bitmap views as either linked-lists or splay trees.
3674         (struct bitmap_element_def): Update comments for splay tree bitmaps.
3675         (struct bitmap_head_def): Likewise.
3676         (bitmap_list_view, bitmap_tree_view): New prototypes.
3677         (bitmap_initialize_stat): Initialize a bitmap_head's indx and
3678         tree_form fields.
3679         (bmp_iter_set_init): Assert the iterated bitmaps are in list form.
3680         (bmp_iter_and_init, bmp_iter_and_compl_init): Likewise.
3681         * bitmap.c (bitmap_elem_to_freelist): Unregister overhead of a
3682         released bitmap element here.
3683         (bitmap_element_free): Remove.
3684         (bitmap_elt_clear_from): Work on splay tree bitmaps.
3685         (bitmap_list_link_element): Renamed from bitmap_element_link.  Move
3686         this function similar ones such that linked-list bitmap implementation
3687         functions are grouped.
3688         (bitmap_list_unlink_element): Renamed from bitmap_element_unlink,
3689         and moved for grouping.
3690         (bitmap_list_insert_element_after): Renamed from
3691         bitmap_elt_insert_after, and moved for grouping.
3692         (bitmap_list_find_element): New function spliced from bitmap_find_bit.
3693         (bitmap_tree_link_left, bitmap_tree_link_right,
3694         bitmap_tree_rotate_left, bitmap_tree_rotate_right, bitmap_tree_splay,
3695         bitmap_tree_link_element, bitmap_tree_unlink_element,
3696         bitmap_tree_find_element): New functions for splay-tree bitmap
3697         implementation.
3698         (bitmap_element_link, bitmap_element_unlink, bitmap_elt_insert_after):
3699         Renamed and moved, see above entries.
3700         (bitmap_tree_listify_from): New function to convert part of a splay
3701         tree bitmap to a linked-list bitmap.
3702         (bitmap_list_view): Convert a splay tree bitmap to linked-list form.
3703         (bitmap_tree_view): Convert a linked-list bitmap to splay tree form.
3704         (bitmap_find_bit): Remove.
3705         (bitmap_clear, bitmap_clear_bit, bitmap_set_bit,
3706         bitmap_single_bit_set_p, bitmap_first_set_bit, bitmap_last_set_bit):
3707         Handle splay tree bitmaps.
3708         (bitmap_copy, bitmap_count_bits, bitmap_and, bitmap_and_into,
3709         bitmap_elt_copy, bitmap_and_compl, bitmap_and_compl_into,
3710         bitmap_compl_and_into, bitmap_elt_ior, bitmap_ior, bitmap_ior_into,
3711         bitmap_xor, bitmap_xor_into, bitmap_equal_p, bitmap_intersect_p,
3712         bitmap_intersect_compl_p, bitmap_ior_and_compl,
3713         bitmap_ior_and_compl_into, bitmap_set_range, bitmap_clear_range,
3714         bitmap_hash): Reject trying to act on splay tree bitmaps.  Make
3715         corresponding changes to use linked-list specific bitmap_element
3716         manipulation functions as applicable for efficiency.
3717         (bitmap_tree_to_vec): New function.
3718         (debug_bitmap_elt_file): New function split out from ...
3719         (debug_bitmap_file): ... here.  Handle splay tree bitmaps.
3720         (bitmap_print): Likewise.
3722         PR tree-optimization/63155
3723         * tree-ssa-propagate.c (ssa_prop_init): Use tree-view for the
3724         SSA edge worklists.
3725         * tree-ssa-coalesce.c (coalesce_ssa_name): Populate used_in_copies
3726         in tree-view.
3728 2018-10-22  Martin Liska  <mliska@suse.cz>
3730         PR tree-optimization/87686
3731         Revert
3732         2018-08-29  Martin Liska  <mliska@suse.cz>
3734         * tree-switch-conversion.c (switch_conversion::expand):
3735         Strenghten assumption about gswitch statements.
3737 2018-10-22  Martin Liska  <mliska@suse.cz>
3739         * ipa-icf.c (sem_item::compare_attributes): Remove.
3740         (sem_item::compare_referenced_symbol_properties): Use
3741         attribute_list_equal instead.
3742         (sem_function::equals_wpa): Likewise.
3743         * ipa-icf.h: Remove compare_attributes.
3745 2018-10-22  Richard Biener  <rguenther@suse.de>
3747         PR middle-end/87682
3748         * mem-stats.h (mem_usage::operator==): Fix pasto.
3750 2018-10-22  Richard Biener  <rguenther@suse.de>
3752         PR tree-optimization/87640
3753         * tree-vrp.c (set_value_range_with_overflow): Decompose
3754         incomplete result.
3755         (extract_range_from_binary_expr_1): Adjust.
3757 2018-10-22  Martin Jambor  <mjambor@suse.cz>
3759         * tree-eh.h (stmt_could_throw_p): Add function parameter.
3760         (stmt_can_throw_external): Likewise.
3761         (stmt_can_throw_internal): Likewise.
3762         * tree-eh.c (lower_eh_constructs_2): Pass cfun to stmt_could_throw_p.
3763         (lower_eh_constructs_2): Likewise.
3764         (stmt_could_throw_p): Add fun parameter, use it instead of cfun.
3765         (stmt_can_throw_external): Likewise.
3766         (stmt_can_throw_internal): Likewise.
3767         (maybe_clean_eh_stmt_fn): Pass cfun to stmt_could_throw_p.
3768         (maybe_clean_or_replace_eh_stmt): Pass cfun to stmt_could_throw_p.
3769         (maybe_duplicate_eh_stmt_fn): Pass new_fun to stmt_could_throw_p.
3770         (maybe_duplicate_eh_stmt): Pass cfun to stmt_could_throw_p.
3771         (pass_lower_eh_dispatch::execute): Pass cfun to
3772         stmt_can_throw_external.
3773         (cleanup_empty_eh): Likewise.
3774         (verify_eh_edges): Pass cfun to stmt_could_throw_p.
3775         * cgraph.c (cgraph_edge::set_call_stmt): Pass a function to
3776         stmt_can_throw_external instead of pushing it to cfun.
3777         (symbol_table::create_edge): Likewise.
3778         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Pass cfun to
3779         stmt_can_throw_internal.
3780         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Pass cfun
3781         to stmt_could_throw_p.
3782         * gimple-ssa-store-merging.c (handled_load): Pass cfun to
3783         stmt_can_throw_internal.
3784         (pass_store_merging::execute): Likewise.
3785         * gimple-ssa-strength-reduction.c
3786         (find_candidates_dom_walker::before_dom_children): Pass cfun to
3787         stmt_could_throw_p.
3788         * gimplify-me.c (gimple_regimplify_operands): Pass cfun to
3789         stmt_can_throw_internal.
3790         * ipa-pure-const.c (check_call): Pass cfun to stmt_could_throw_p and
3791         to stmt_can_throw_external.
3792         (check_stmt): Pass cfun to stmt_could_throw_p.
3793         (check_stmt): Pass cfun to stmt_can_throw_external.
3794         (pass_nothrow::execute): Likewise.
3795         * trans-mem.c (expand_call_tm): Pass cfun to stmt_can_throw_internal.
3796         * tree-cfg.c (is_ctrl_altering_stmt): Pass cfun to
3797         stmt_can_throw_internal.
3798         (verify_gimple_in_cfg): Pass cfun to stmt_could_throw_p.
3799         (stmt_can_terminate_bb_p): Pass cfun to stmt_can_throw_external.
3800         (gimple_purge_dead_eh_edges): Pass cfun to stmt_can_throw_internal.
3801         * tree-complex.c (expand_complex_libcall): Pass cfun to
3802         stmt_could_throw_p and to stmt_can_throw_internal.
3803         (expand_complex_multiplication): Pass cfun to stmt_can_throw_internal.
3804         * tree-inline.c (copy_edges_for_bb): Likewise.
3805         (maybe_move_debug_stmts_to_successors): Likewise.
3806         * tree-outof-ssa.c (ssa_is_replaceable_p): Pass cfun to
3807         stmt_could_throw_p.
3808         * tree-parloops.c (oacc_entry_exit_ok_1): Likewise.
3809         * tree-sra.c (scan_function): Pass cfun to stmt_can_throw_external.
3810         * tree-ssa-alias.c (stmt_kills_ref_p): Pass cfun to
3811         stmt_can_throw_internal.
3812         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Likewise.
3813         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Pass cfun to
3814         stmt_could_throw_p.
3815         (mark_aliased_reaching_defs_necessary_1): Pass cfun to
3816         stmt_can_throw_internal.
3817         * tree-ssa-forwprop.c (pass_forwprop::execute): Likewise.
3818         * tree-ssa-loop-im.c (movement_possibility): Pass cfun to
3819         stmt_could_throw_p.
3820         * tree-ssa-loop-ivopts.c (find_givs_in_stmt_scev): Likewise.
3821         (add_autoinc_candidates): Pass cfun to stmt_can_throw_internal.
3822         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
3823         (convert_mult_to_fma_1): Likewise.
3824         (convert_to_divmod): Likewise.
3825         * tree-ssa-phiprop.c (propagate_with_phi): Likewise.
3826         * tree-ssa-pre.c (compute_avail): Pass cfun to stmt_could_throw_p.
3827         * tree-ssa-propagate.c
3828         (substitute_and_fold_dom_walker::before_dom_children): Likewise.
3829         * tree-ssa-reassoc.c (suitable_cond_bb): Likewise.
3830         (maybe_optimize_range_tests): Likewise.
3831         (linearize_expr_tree): Likewise.
3832         (reassociate_bb): Likewise.
3833         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
3834         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Likewise.
3835         * tree-ssa-strlen.c (adjust_last_stmt): Likewise.
3836         (handle_char_store): Likewise.
3837         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Pass cfun to
3838         stmt_can_throw_internal.
3839         * tree-vect-patterns.c (check_bool_pattern): Pass cfun to
3840         stmt_could_throw_p.
3841         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
3842         (vectorizable_call): Pass cfun to stmt_can_throw_internal.
3843         (vectorizable_simd_clone_call): Likewise.
3844         * value-prof.c (gimple_ic): Pass cfun to stmt_could_throw_p.
3845         (gimple_stringop_fixed_value): Likewise.
3847 2018-10-22  Ilya Leoshkevich  <iii@linux.ibm.com>
3849         * config/s390/s390.c (s390_loadrelative_operand_p): Accept
3850         literal pool references.
3851         (s390_check_qrst_address): Adapt to the new behavior of
3852         s390_loadrelative_operand_p ().
3854 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3856         PR target/72782
3857         * config/i386/sse.md (*andnot<mode>3_bcst): New.
3859 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3861         PR target/72782
3862         * config/i386/sse.md (*<code><mode>3_bcst): New.
3864 2018-10-22  H.J. Lu  <hongjiu.lu@intel.com>
3866         PR target/72782
3867         * config/i386/sse.md (avx512bcst): Updated for V4SI, V2DI, V8SI,
3868         V4DI, V16SI and V8DI.
3869         (*sub<mode>3<mask_name>_bcst): New.
3870         (*add<mode>3<mask_name>_bcst): Likewise.
3872 2018-10-21  Bill Schmidt  <wschmidt@linux.ibm.com>
3873             Jinsong Ji  <jji@us.ibm.com>
3875         * config/rs6000/emmintrin.h (_mm_movemask_pd): Replace __vector
3876         __m64 with __vector unsigned long long for compatibility.
3877         (_mm_movemask_epi8): Likewise.
3878         * config/rs6000/xmmintrin.h (_mm_cvtps_pi32): Likewise.
3879         (_mm_cvttps_pi32): Likewise.
3880         (_mm_cvtpi32_ps): Likewise.
3881         (_mm_cvtps_pi16): Likewise.
3882         (_mm_loadh_pi): Likewise.
3883         (_mm_storeh_pi): Likewise.
3884         (_mm_movehl_ps): Likewise.
3885         (_mm_movelh_ps): Likewise.
3886         (_mm_loadl_pi): Likewise.
3887         (_mm_storel_pi): Likewise.
3888         (_mm_movemask_ps): Likewise.
3889         (_mm_shuffle_pi16): Likewise.
3891 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3893         PR target/72782
3894         * config/i386/avx512fintrin.h (_mm512_fnmsub_round_pd): Use
3895         __builtin_ia32_vfnmsubpd512_mask.
3896         (_mm512_mask_fnmsub_round_pd): Likewise.
3897         (_mm512_fnmsub_pd): Likewise.
3898         (_mm512_mask_fnmsub_pd): Likewise.
3899         (_mm512_maskz_fnmsub_round_pd): Use
3900         __builtin_ia32_vfnmsubpd512_maskz.
3901         (_mm512_maskz_fnmsub_pd): Likewise.
3902         (_mm512_fnmsub_round_ps): Use __builtin_ia32_vfnmsubps512_mask.
3903         (_mm512_mask_fnmsub_round_ps): Likewise.
3904         (_mm512_fnmsub_ps): Likewise.
3905         (_mm512_mask_fnmsub_ps): Likewise.
3906         (_mm512_maskz_fnmsub_round_ps): Use
3907         __builtin_ia32_vfnmsubps512_maskz.
3908         (_mm512_maskz_fnmsub_ps): Likewise.
3909         * config/i386/avx512vlintrin.h (_mm256_mask_fnmsub_pd): Use
3910         __builtin_ia32_vfnmsubpd256_mask.
3911         (_mm256_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256_maskz.
3912         (_mm_mask_fnmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
3913         (_mm_maskz_fnmsub_pd): Use __builtin_ia32_vfnmsubpd128_maskz.
3914         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3915         (_mm256_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_mask.
3916         (_mm256_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps256_maskz.
3917         (_mm_mask_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_mask.
3918         (_mm_maskz_fnmsub_ps): Use __builtin_ia32_vfnmsubps128_maskz.
3919         * config/i386/fmaintrin.h (_mm_fnmsub_pd): Use
3920         __builtin_ia32_vfnmsubpd.
3921         (_mm256_fnmsub_pd): Use __builtin_ia32_vfnmsubpd256.
3922         (_mm_fnmsub_ps): Use __builtin_ia32_vfnmsubps.
3923         (_mm256_fnmsub_ps): Use __builtin_ia32_vfnmsubps256.
3924         (_mm_fnmsub_sd): Use __builtin_ia32_vfnmsubsd3.
3925         (_mm_fnmsub_ss): Use __builtin_ia32_vfnmsubss3.
3926         * config/i386/i386-builtin.def: Add
3927         __builtin_ia32_vfnmsubpd256_mask,
3928         __builtin_ia32_vfnmsubpd256_maskz,
3929         __builtin_ia32_vfnmsubpd128_mask,
3930         __builtin_ia32_vfnmsubpd128_maskz,
3931         __builtin_ia32_vfnmsubps256_mask,
3932         __builtin_ia32_vfnmsubps256_maskz,
3933         __builtin_ia32_vfnmsubps128_mask,
3934         __builtin_ia32_vfnmsubps128_maskz,
3935         __builtin_ia32_vfnmsubpd512_mask,
3936         __builtin_ia32_vfnmsubpd512_maskz,
3937         __builtin_ia32_vfnmsubps512_mask,
3938         __builtin_ia32_vfnmsubps512_maskz, __builtin_ia32_vfnmsubss3,
3939         __builtin_ia32_vfnmsubsd3, __builtin_ia32_vfnmsubps,
3940         __builtin_ia32_vfnmsubpd, __builtin_ia32_vfnmsubps256 and.
3941         __builtin_ia32_vfnmsubpd256.
3942         * config/i386/sse.md (fma4i_fnmsub_<mode>): New.
3943         (<avx512>_fnmsub_<mode>_maskz<round_expand_name>): Likewise.
3944         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_1):
3945         Likewise.
3946         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_2):
3947         Likewise.
3948         (*<sd_mask_codefor>fma_fnmsub_<mode><sd_maskz_name>_bcst_3):
3949         Likewise.
3950         (fmai_vmfnmsub_<mode><round_name>): Likewise.
3952 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
3954         PR target/72782
3955         * config/i386/avx512fintrin.h (_mm512_fnmadd_round_pd): Use
3956         __builtin_ia32_vfnmaddpd512_mask.
3957         (_mm512_mask_fnmadd_round_pd): Likewise.
3958         (_mm512_fnmadd_pd): Likewise.
3959         (_mm512_mask_fnmadd_pd): Likewise.
3960         (_mm512_maskz_fnmadd_round_pd): Use
3961         __builtin_ia32_vfnmaddpd512_maskz.
3962         (_mm512_maskz_fnmadd_pd): Likewise.
3963         (_mm512_fnmadd_round_ps): Use __builtin_ia32_vfnmaddps512_mask.
3964         (_mm512_mask_fnmadd_round_ps): Likewise.
3965         (_mm512_fnmadd_ps): Likewise.
3966         (_mm512_mask_fnmadd_ps): Likewise.
3967         (_mm512_maskz_fnmadd_round_ps): Use
3968         __builtin_ia32_vfnmaddps512_maskz.
3969         (_mm512_maskz_fnmadd_ps): Likewise.
3970         * config/i386/avx512vlintrin.h (_mm256_mask_fnmadd_pd): Use
3971         __builtin_ia32_vfnmaddpd256_mask.
3972         (_mm256_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256_maskz.
3973         (_mm_mask_fnmadd_pd): Use __builtin_ia32_vfmaddpd128_mask
3974         (_mm_maskz_fnmadd_pd): Use __builtin_ia32_vfnmaddpd128_maskz.
3975         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3976         (_mm256_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_mask.
3977         (_mm256_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps256_maskz.
3978         (_mm_mask_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_mask.
3979         (_mm_maskz_fnmadd_ps): Use __builtin_ia32_vfnmaddps128_maskz.
3980         * config/i386/fmaintrin.h (_mm_fnmadd_pd): Use
3981         __builtin_ia32_vfnmaddpd.
3982         (_mm256_fnmadd_pd): Use __builtin_ia32_vfnmaddpd256.
3983         (_mm_fnmadd_ps): Use __builtin_ia32_vfnmaddps.
3984         (_mm256_fnmadd_ps): Use __builtin_ia32_vfnmaddps256.
3985         (_mm_fnmadd_sd): Use __builtin_ia32_vfnmaddsd3.
3986         (_mm_fnmadd_ss): Use __builtin_ia32_vfnmaddss3.
3987         * config/i386/i386-builtin.def: Add
3988         __builtin_ia32_vfnmaddpd256_mask,
3989         __builtin_ia32_vfnmaddpd256_maskz,
3990         __builtin_ia32_vfnmaddpd128_mask,
3991         __builtin_ia32_vfnmaddpd128_maskz,
3992         __builtin_ia32_vfnmaddps256_mask,
3993         __builtin_ia32_vfnmaddps256_maskz,
3994         __builtin_ia32_vfnmaddps128_mask,
3995         __builtin_ia32_vfnmaddps128_maskz,
3996         __builtin_ia32_vfnmaddpd512_mask,
3997         __builtin_ia32_vfnmaddpd512_maskz,
3998         __builtin_ia32_vfnmaddps512_mask,
3999         __builtin_ia32_vfnmaddps512_maskz, __builtin_ia32_vfnmaddss3,
4000         __builtin_ia32_vfnmaddsd3, __builtin_ia32_vfnmaddps,
4001         __builtin_ia32_vfnmaddpd, __builtin_ia32_vfnmaddps256 and.
4002         __builtin_ia32_vfnmaddpd256.
4003         * config/i386/sse.md (fma4i_fnmadd_<mode>): New.
4004         (<avx512>_fnmadd_<mode>_maskz<round_expand_name>): Likewise.
4005         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_1):
4006         Likewise.
4007         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_2):
4008         Likewise.
4009         (*<sd_mask_codefor>fma_fnmadd_<mode><sd_maskz_name>_bcst_3):
4010         Likewise.
4011         (fmai_vmfnmadd_<mode><round_name>): Likewise.
4013 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4015         PR target/72782
4016         * config/i386/avx512fintrin.h (_mm512_fmsub_round_pd): Use
4017         __builtin_ia32_vfmsubpd512_mask.
4018         (_mm512_mask_fmsub_round_pd): Likewise.
4019         (_mm512_fmsub_pd): Likewise.
4020         (_mm512_mask_fmsub_pd): Likewise.
4021         (_mm512_maskz_fmsub_round_pd): Use
4022         __builtin_ia32_vfmsubpd512_maskz.
4023         (_mm512_maskz_fmsub_pd): Likewise.
4024         (_mm512_fmsub_round_ps): Use __builtin_ia32_vfmsubps512_mask.
4025         (_mm512_mask_fmsub_round_ps): Likewise.
4026         (_mm512_fmsub_ps): Likewise.
4027         (_mm512_mask_fmsub_ps): Likewise.
4028         (_mm512_maskz_fmsub_round_ps): Use
4029         __builtin_ia32_vfmsubps512_maskz.
4030         (_mm512_maskz_fmsub_ps): Likewise.
4031         * config/i386/avx512vlintrin.h (_mm256_mask_fmsub_pd): Use
4032         __builtin_ia32_vfmsubpd256_mask.
4033         (_mm256_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd256_maskz.
4034         (_mm_mask_fmsub_pd): Use __builtin_ia32_vfmaddpd128_mask
4035         (_mm_maskz_fmsub_pd): Use __builtin_ia32_vfmsubpd128_maskz.
4036         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
4037         (_mm256_mask_fmsub_ps): Use __builtin_ia32_vfmsubps256_mask.
4038         (_mm256_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps256_maskz.
4039         (_mm_mask_fmsub_ps): Use __builtin_ia32_vfmsubps128_mask.
4040         (_mm_maskz_fmsub_ps): Use __builtin_ia32_vfmsubps128_maskz.
4041         * config/i386/fmaintrin.h (_mm_fmsub_pd): Use
4042         __builtin_ia32_vfmsubpd.
4043         (_mm256_fmsub_pd): Use __builtin_ia32_vfmsubpd256.
4044         (_mm_fmsub_ps): Use __builtin_ia32_vfmsubps.
4045         (_mm256_fmsub_ps): Use __builtin_ia32_vfmsubps256.
4046         (_mm_fmsub_sd): Use __builtin_ia32_vfmsubsd3.
4047         (_mm_fmsub_ss): Use __builtin_ia32_vfmsubss3.
4048         * config/i386/i386-builtin.def: Add
4049         __builtin_ia32_vfmsubpd256_mask,
4050         __builtin_ia32_vfmsubpd256_maskz,
4051         __builtin_ia32_vfmsubpd128_mask,
4052         __builtin_ia32_vfmsubpd128_maskz,
4053         __builtin_ia32_vfmsubps256_mask,
4054         __builtin_ia32_vfmsubps256_maskz,
4055         __builtin_ia32_vfmsubps128_mask,
4056         __builtin_ia32_vfmsubps128_maskz,
4057         __builtin_ia32_vfmsubpd512_mask,
4058         __builtin_ia32_vfmsubpd512_maskz,
4059         __builtin_ia32_vfmsubps512_mask,
4060         __builtin_ia32_vfmsubps512_maskz, __builtin_ia32_vfmsubss3,
4061         __builtin_ia32_vfmsubsd3, __builtin_ia32_vfmsubps,
4062         __builtin_ia32_vfmsubpd, __builtin_ia32_vfmsubps256 and.
4063         __builtin_ia32_vfmsubpd256.
4064         * config/i386/sse.md (fma4i_fmsub_<mode>): New.
4065         (<avx512>_fmsub_<mode>_maskz<round_expand_name>): Likewise.
4066         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_1):
4067         Likewise.
4068         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_2):
4069         Likewise.
4070         (*<sd_mask_codefor>fma_fmsub_<mode><sd_maskz_name>_bcst_3):
4071         Likewise.
4072         (fmai_vmfmsub_<mode><round_name>): Likewise.
4074 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4076         * config/i386/sse.md (*<plusminus_insn><mode>3<mask_name>_bcst_1):
4077         Remove plus.  Renamed to ...
4078         (*sub<mode>3<mask_name>_bcst): This.
4079         (*add<mode>3<mask_name>_bcst_2): Renamede to ...
4080         (*add<mode>3<mask_name>_bcst): This.
4082 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4084         PR target/72782
4085         * config/i386/sse.md (*mul<mode>3<mask_name>_bcst): New.
4087 2018-10-21  H.J. Lu  <hongjiu.lu@intel.com>
4089         PR target/87662
4090         * i386/avx512vlintrin.h (_mm256_or_epi32): New.
4091         (_mm_or_epi32): Likewise.
4092         (_mm256_xor_epi32): Likewise.
4093         (_mm_xor_epi32): Likewise.
4094         (_mm256_or_epi64): Likewise.
4095         (_mm_or_epi64): Likewise.
4096         (_mm256_xor_epi64): Likewise.
4097         (_mm_xor_epi64): Likewise.
4099 2018-10-20  H.J. Lu  <hongjiu.lu@intel.com>
4101         PR target/72782
4102         * config/i386/sse.md (*<avx512>_div<mode>3<mask_name>_bcst): New.
4104 2018-10-20  Jakub Jelinek  <jakub@redhat.com>
4106         PR middle-end/87647
4107         * varasm.c (decode_addr_const): Handle COMPOUND_LITERAL_EXPR.
4109 2018-10-20  Andreas Schwab  <schwab@linux-m68k.org>
4111         * doc/ux.texi: Move @section directly after @node.
4113 2018-10-19  Jakub Jelinek  <jakub@redhat.com>
4115         PR middle-end/85488
4116         PR middle-end/87649
4117         * omp-low.c (check_omp_nesting_restrictions): Diagnose ordered without
4118         depend closely nested inside of loop with ordered clause with
4119         a parameter.
4121 2018-10-19  David Malcolm  <dmalcolm@redhat.com>
4123         * Makefile.in (TEXI_GCCINT_FILES): Add ux.texi.
4124         * doc/gccint.texi: Include ux.texi and use it in top-level menu.
4125         * doc/ux.texi: New file.
4127 2018-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
4129         * config/rs6000/rs6000.h (REG_ALLOC_ORDER): Move 68 (that is, CR0) to
4130         be the first CR field allocated.
4132 2018-10-19  Richard Biener  <rguenther@suse.de>
4134         PR target/87657
4135         * config/i386/i386.c (ix86_builtin_vectorization_cost): Use
4136         TYPE_VECTOR_SUBPARTS and avoid relying on vector mode.
4138 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
4140         PR target/72782
4141         * config/i386/sse.md
4142         (*<plusminus_insn><mode>3<mask_name>_bcst_1): New.
4143         (*add<mode>3<mask_name>_bcst_2): Likewise.
4145 2018-10-19  H.J. Lu  <hongjiu.lu@intel.com>
4147         * config/i386/sse.md
4148         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
4149         Replace nonimmediate_operand with register_operand.
4150         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
4151         Likewise.
4152         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
4153         Likewise.
4155 2018-10-19  Ilya Leoshkevich  <iii@linux.ibm.com>
4157         PR rtl-optimization/87596
4158         * lra-constraints.c (spill_hard_reg_in_range): Use INSN_P () +
4159         lra_get_insn_recog_data () instead of lra_insn_recog_data[]
4160         for instructions in FROM..TO range.
4162 2018-10-19  Eric Botcazou  <ebotcazou@adacore.com>
4164         * cfgexpand.c (expand_one_var): Use specific wording in error message
4165         for non-local frame variables.
4166         * stor-layout.c (layout_decl): Do not issue a warning for them.
4168 2018-10-19  Robin Dapp  <rdapp@linux.ibm.com>
4170         * haifa-sched.c (priority): Add force_recompute parameter.
4171         (apply_replacement): Call priority () with force_recompute = true.
4172         (restore_pattern): Likewise.
4174 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4176         * simplify-rtx.c (simplify_subreg): Limit mask of vec_merge to
4177         HOST_BITS_PER_WIDE_INT.
4178         (test_vector_ops_duplicate): Likewise.
4180 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4182         PR target/72782
4183         * config/i386/sse.md (VF_AVX512): New.
4184         (avx512bcst): Likewise.
4185         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_1):
4186         Likewise.
4187         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_2):
4188         Likewise.
4189         (*<sd_mask_codefor>fma_fmadd_<mode><sd_maskz_name>_bcst_3):
4190         Likewise.
4192 2018-10-18  Jonathan Wakely  <jwakely@redhat.com>
4194         * doc/invoke.texi (-dumpversion): Improve grammar.
4195         (-dumpfullversion): Make more consistent with -dumpversion.
4197 2018-10-18  Uros Bizjak  <ubizjak@gmail.com>
4199         * config/i386/i386.c (ix86_emit_fp_unordered_jump):
4200         Set JUMP_LABEL to the jump insn.
4201         (ix86_emit_i387_log1p): Use ix86_expand_branch to expand branch.
4202         Predict emitted jump and add label to jump insn.
4204 2018-10-18  David Malcolm  <dmalcolm@redhat.com>
4206         PR tree-optimization/87562
4207         * input.c (get_substring_ranges_for_loc): Use
4208         LRK_SPELLING_LOCATION rather than LRK_MACRO_EXPANSION_POINT when
4209         getting the linemap for the endpoint.  Verify that it's either
4210         in the same linemap as the start point's spelling location, or
4211         at least in the same file.
4213 2018-10-18  Richard Biener  <rguenther@suse.de>
4215         * config/i386/i386.c (ix86_builtin_vectorization_cost): Do not
4216         feed width-specific load/store costs through ix86_vec_cost.
4217         * config/i386/x86-tune-costs.h (athlon_cost): Adjust.
4218         (k8_cost): Likewise.
4219         (bdver_cost): Likewise.
4220         (znver1_cost): Likewise.
4221         (btver1_cost): Likewise.
4222         (btver2_cost): Likewise.
4224 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4226         * simplify-rtx.c (simplify_subreg): Call simplify_gen_subreg
4227         to simplify subreg of vec_merge.
4229 2018-10-18  Richard Biener  <rguenther@suse.de>
4231         * config/i386/i386.c: Fix costing of vector FMA.
4233 2018-10-18  Richard Biener  <rguenther@suse.de>
4235         * config/i386/i386.c (ix86_vec_cost): Remove !parallel path
4236         and argument.
4237         (ix86_builtin_vectorization_cost): For vec_construct properly
4238         cost insertion into SSE regs.
4239         (...): Adjust calls to ix86_vec_cost.
4241 2018-10-18  Richard Biener  <rguenther@suse.de>
4243         PR middle-end/87087
4244         Revert
4245         2018-02-07  Richard Biener  <rguenther@suse.de>
4247         PR tree-optimization/84204
4248         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
4249         this place.
4251 2018-10-18  H.J. Lu  <hongjiu.lu@intel.com>
4253         PR target/87537
4254         * simplify-rtx.c (simplify_subreg): Simplify subreg of vec_merge
4255         of vec_duplicate.
4256         (test_vector_ops_duplicate): Add test for a scalar subreg of a
4257         VEC_MERGE of a VEC_DUPLICATE.
4259 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
4261         * doc/cpp.texi (__STDC_VERSION__): Document C2X handling.
4262         * doc/invoke.texi (-std=c2x, -std=gnu2x): Document new options.
4263         * doc/standards.texi (C Language): Document C2X.
4264         * dwarf2out.c (highest_c_language), config/rl78/rl78.c
4265         (rl78_option_override): Handle "GNU C2X" language name.
4267 2018-10-17  Joseph Myers  <joseph@codesourcery.com>
4269         * doc/invoke.texi (-std=c17), doc/standards.texi (C Language):
4270         Document C17 as published in 2018.
4272 2018-10-17  Eric Botcazou  <ebotcazou@adacore.com>
4274         PR middle-end/87623
4275         * fold-const.c (fold_truth_andor_1): If the right side is not constant,
4276         bail out if both sides do not have the same storage order.
4278 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4280         * bitmap.c (bitmap_head::dump): New.
4281         * bitmap.h (bitmap_head): Add dump().
4282         * gimple-ssa-evrp-analyze.c
4283         (evrp_range_analyzer::try_find_new_range): Adjust for value_range API.
4284         (evrp_range_analyzer::set_ssa_range_info): Same.
4285         (evrp_range_analyzer::record_ranges_from_phis): Same.
4286         (evrp_range_analyzer::record_ranges_from_stmt): Same.
4287         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Same.
4288         * gimple-ssa-sprintf.c (get_int_range): Same.
4289         (format_integer): Same.
4290         (sprintf_dom_walker::handle_gimple_call): Same.
4291         * ipa-cp.c (ipcp_vr_lattice::meet_with_1): Same.
4292         (ipcp_vr_lattice::top_p): Same.
4293         (ipcp_vr_lattice::bottom_p): Same.
4294         (ipcp_vr_lattice::set_to_bottom): Same.
4295         (ipa_vr_operation_and_type_effects): Same.
4296         (propagate_vr_across_jump_function): Same.
4297         (ipcp_store_vr_results): Same.
4298         * ipa-prop.c (struct ipa_vr_ggc_hash_traits): Same.
4299         (ipa_print_node_jump_functions_for_edge): Same.
4300         (ipa_get_value_range): Same.
4301         (ipa_compute_jump_functions_for_edge): Same.
4302         (ipa_write_jump_function): Same.
4303         * tree-ssa-dom.c (simplify_stmt_for_jump_threading): Same.
4304         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis):
4305         Same.
4306         * vr-values.c (set_value_range_to_nonnegative): Same.
4307         (set_value_range_to_truthvalue): Same.
4308         (vr_values::get_value_range): Same.
4309         (vr_values::set_defs_to_varying): Same.
4310         (vr_values::update_value_range): Same.
4311         (symbolic_range_based_on_p): Same.
4312         (vr_values::op_with_boolean_value_range_p): Same.
4313         (vr_values::extract_range_for_var_from_comparison_expr): Same.
4314         (vr_values::extract_range_from_ssa_name): Same.
4315         (vr_values::extract_range_from_binary_expr): Same.
4316         (vr_values::extract_range_from_unary_expr): Same.
4317         (vr_values::extract_range_from_cond_expr): Same.
4318         (vr_values::extract_range_from_comparison): Same.
4319         (vr_values::check_for_binary_op_overflow): Same.
4320         (vr_values::extract_range_basic): Same.
4321         (vr_values::extract_range_from_assignment): Same.
4322         (compare_ranges): Same.
4323         (compare_range_with_value): Same.
4324         (vr_values::adjust_range_with_scev): Same.
4325         (vrp_valueize): Same.
4326         (vrp_valueize_1): Same.
4327         (vr_values::get_vr_for_comparison): Same.
4328         (vr_values::compare_name_with_value): Same.
4329         (vr_values::compare_names): Same.
4330         (vr_values::vrp_evaluate_conditional): Same.
4331         (find_case_label_ranges): Same.
4332         (vr_values::vrp_visit_switch_stmt): Same.
4333         (vr_values::extract_range_from_phi_node): Same.
4334         (vr_values::simplify_div_or_mod_using_ranges): Same.
4335         (vr_values::simplify_bit_ops_using_ranges): Same.
4336         (test_for_singularity): Same.
4337         (range_fits_type_p): Same.
4338         (vr_values::simplify_cond_using_ranges_1): Same.
4339         (vr_values::simplify_switch_using_ranges): Same.
4340         (vr_values::simplify_float_conversion_using_ranges): Same.
4341         (vr_values::two_valued_val_range_p): Same.
4342         (vr_values::add_equivalence): Move to value_range::equiv_add.
4343         * vr-values.h (vr_values::add_equivalence): Remove.
4344         (VR_INITIALIZER): Remove.
4345         * tree-vrp.c (value_range::set): New.
4346         (value_range::equiv_add): New.
4347         (value_range::value_range): New.
4348         (value_range::deep_copy): New.
4349         (value_range::check): New.
4350         (value_range::equal_p): New.
4351         (value_range::ignore_equivs_equal_p): New.
4352         (value_range::operator==): New.
4353         (value_range::operator!=): New.
4354         (value_range::symbolic_p): New.
4355         (value_range::numeric_p): New.
4356         (value_range::set_undefined): New.
4357         (value_range::set_varying): New.
4358         (value_range::may_contain_p): New.
4359         (value_range::equiv_clear): New.
4360         (value_range::singleton_p): New.
4361         (value_range::intersect): New.
4362         (value_range::dump): New.
4363         (value_range::set_and_canonicalize): New.
4364         (set_value_range): Adjust for value_range API.
4365         (set_value_range_to_undefined): Same.
4366         (set_value_range_to_varying): Same.
4367         (set_and_canonicalize_value_range): Same.
4368         (set_value_range_to_nonnull): Same.
4369         (set_value_range_to_null): Same.
4370         (range_is_null): Same.
4371         (range_is_nonnull): Same.
4372         (range_int_cst_p): Same.
4373         (range_int_cst_singleton_p): Same.
4374         (symbolic_range_p): Same.
4375         (range_includes_zero_p): Same.
4376         (value_range_constant_singleton): Same.
4377         (vrp_set_zero_nonzero_bits): Same.
4378         (ranges_from_anti_range): Same.
4379         (extract_range_into_wide_ints): Same.
4380         (extract_range_from_multiplicative_op): Same.
4381         (set_value_range_with_overflow): Same.
4382         (extract_range_from_binary_expr_1): Same.
4383         (extract_range_from_unary_expr): Same.
4384         (dump_value_range): Same.
4385         (debug_value_range): Same.
4386         (vrp_prop::check_array_ref): Same.
4387         (vrp_prop::check_mem_ref): Same.
4388         (vrp_prop::vrp_initialize): Same.
4389         (vrp_prop::visit_stmt): Same.
4390         (intersect_ranges): Same.
4391         (vrp_prop::visit_phi): Same.
4392         (vrp_prop::vrp_finalize): Same.
4393         (determine_value_range_1): Same.
4394         (determine_value_range): Same.
4395         (vrp_intersect_ranges_1): Rename to...
4396         (vrp_intersect_1): this.
4397         (vrp_intersect_ranges): Rename to...
4398         (value_range::intersect_helper): ...this.
4399         (vrp_meet_1): Rename to...
4400         (value_range::union_helper): ...this.
4401         (vrp_meet): Rename to...
4402         (value_range::union_): ...this.
4403         (copy_value_range): Remove.
4404         * tree-vrp.h (struct value_range): Rewrite into a proper class.
4405         (value_range::vrtype): New.
4406         (value_range::type): New.
4407         (value_range::equiv): New.
4408         (value_range::min): New.
4409         (value_range::max): New.
4410         (value_range::varying_p): New.
4411         (value_range::undefined_p): New.
4412         (value_range::null_p): New.
4413         (value_range::equiv_add): New.
4414         (copy_value_range): Remove.
4416 2018-10-17  David Malcolm  <dmalcolm@redhat.com>
4418         * Makefile.in (SELFTEST_TARGETS): New.
4419         (selftest) Change from s-selftest-c to $(SELFTEST_TARGETS).
4420         (C_SELFTEST_FLAGS, C_SELFTEST_DEPS, s-selftest-c, selftest-c-gdb)
4421         (selftest-gdb, selftest-c-valgrind, selftest-valgrind): Move to
4422         c/Make-lang.in.
4423         (CPP_SELFTEST_FLAGS, CPP_SELFTEST_DEPS, s-selftest-c++)
4424         (selftest-c++-gdb, selftest-c++-valgrind): Move to
4425         cp/Make-lang.in.
4426         * configure: Regenerate.
4427         * configure.ac (selftest_languages): New.
4429 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4431         * tree-vrp.c (extract_range_from_multiplicative_op): Remove
4432         overflow wraps argument.
4433         (extract_range_from_binary_expr_1): Do not pass overflow wraps to
4434         wide_int_range_multiplicative_op.
4435         * wide-int-range.cc (wide_int_range_mult_wrapping): Remove
4436         overflow wraps argument.
4437         (wide_int_range_multiplicative_op): Same.
4438         (wide_int_range_lshift): Same.
4439         (wide_int_range_div): Same.
4440         * wide-int-range.h (wide_int_range_multiplicative_op): Same.
4441         (wide_int_range_lshift): Same.
4442         (wide_int_range_div): Same.
4444 2018-10-17  Aldy Hernandez  <aldyh@redhat.com>
4446         * wide-int-range.h (wide_int_range_shift_undefined_p): Adjust to
4447         use sign as argument.
4448         * tree-vrp.c (extract_range_from_binary_expr_1): Pass sign to
4449         wide_int_range_shift_undefined_p.
4451 2018-10-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
4453         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>):
4454         Rename to...
4455         (@despeculate_copy<ALLI_TI:mode>): ... This.
4456         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): Remove
4457         switch statement.
4459 2018-10-16  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4461         * config.gcc: Obsolete *-*-solaris2.10*.
4462         * doc/install.texi (Specific, *-*-solaris2*): Document it.
4464 2018-10-12  Jeff Law  <law@redhat.com>
4466         * config/ft32/ft32.md (ft32_general_movsrc_operand): Disable
4467         reg + sym +- const_int addressing modes.
4469 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4471         * common.opt (fdiagnostics-minimum-margin-width=): New option.
4472         * diagnostic-show-locus.c (layout::layout): Apply the minimum
4473         margin width.
4474         (layout::start_annotation_line): Only print up to 3 of the
4475         margin character, to avoid touching the left-hand side.
4476         (selftest::test_diagnostic_show_locus_fixit_lines): Update for
4477         minimum margin width, as set by test_diagnostic_context's ctor.
4478         (selftest::test_fixit_insert_containing_newline): Likewise.
4479         (selftest::test_fixit_insert_containing_newline_2): Likewise.
4480         (selftest::test_line_numbers_multiline_range): Clear
4481         dc.min_margin_width.
4482         * diagnostic.c (diagnostic_initialize): Initialize
4483         min_margin_width.
4484         * diagnostic.h (struct diagnostic_context): Add field
4485         "min_margin_width".
4486         * doc/invoke.texi: Add -fdiagnostics-minimum-margin-width=.
4487         * opts.c (common_handle_option): Handle
4488         OPT_fdiagnostics_minimum_margin_width_.
4489         * selftest-diagnostic.c
4490         (selftest::test_diagnostic_context::test_diagnostic_context):
4491         Initialize min_margin_width to 6.
4492         * toplev.c (general_init): Initialize global_dc->min_margin_width.
4494 2018-10-15  David Malcolm  <dmalcolm@redhat.com>
4496         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
4497         Fix usage of "error_at_rich_loc" in the comment.
4499 2018-10-15  Renlin Li  <renlin.li@arm.com>
4501         PR target/87563
4502         * tree-vectorizer.c (try_vectorize_loop_1): Don't use
4503         if-conversioned loop when it contains ifn with types not
4504         supported by backend.
4505         * internal-fn.c (expand_direct_optab_fn): Add an assert.
4506         (direct_internal_fn_supported_p): New helper function.
4507         * internal-fn.h (direct_internal_fn_supported_p): Declare.
4509 2018-10-15  Jakub Jelinek  <jakub@redhat.com>
4511         PR target/87572
4512         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4513         Remove OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4514         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4516 2018-10-15  Bin Cheng  <bin.cheng@linux.alibaba.com>
4518         PR tree-optimization/87022
4519         * tree-loop-distribution.c (pg_add_dependence_edges): Check all
4520         bits in dist vector rather than the first one.
4522 2018-10-15  Richard Biener  <rguenther@suse.de>
4524         PR middle-end/87610
4525         * tree-ssa-structalias.c (struct vls_data): Add escaped_p member.
4526         (visit_loadstore): When a used restrict tag escaped verify that
4527         the points-to solution of "other" pointers do not include
4528         escaped.
4529         (compute_dependence_clique): If a used restrict tag escaped
4530         communicated that down to visit_loadstore.
4532 2018-10-15  Andreas Krebbel  <krebbel@linux.ibm.com>
4534         * config/s390/s390.c (s390_expand_vec_init): Force vector element
4535         into reg if it isn't a general operand.
4537 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4539         PR target/87599
4540         * config/i386/sse.md (*vec_dupv2di): Add register source to
4541         movddup.
4543 2018-10-14  H.J. Lu  <hongjiu.lu@intel.com>
4545         PR target/87572
4546         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512F_UNSET):
4547         Add OPTION_MASK_ISA_AVX512IFMA_UNSET,
4548         OPTION_MASK_ISA_AVX5124FMAPS_UNSET and
4549         OPTION_MASK_ISA_AVX5124VNNIW_UNSET.
4551 2018-10-13  Eric Botcazou  <ebotcazou@adacore.com>
4553         * dwarf2cfi.c (struct dw_trace_info): Add args_size_defined_for_eh.
4554         (notice_args_size): Set it in the current trace if no insn that can
4555         throw internally has been seen yet.
4556         (connect_traces): When connecting args_size between traces, allow the
4557         incoming values not to match if there is an insn setting it before the
4558         first insn that can throw internally; in that case, force the creation
4559         of a CFI note on this latter insn.
4561 2018-10-13  Jonathan Wakely  <jwakely@redhat.com>
4563         * opt-problem.h (opt_wrapper): Use template-argument-list when naming
4564         the base class, because using the injected-class-name was not clearly
4565         specified until DR 176.
4567 2018-10-12  Paul Koning  <ni1d@arrl.net>
4569         * config/pdp11/pdp11.md (doloop_end): New expander.
4570         (doloop_end_insn): renamed from "doloop_end".
4571         (addqi3): New pattern.
4572         (subqi3): New pattern.
4573         * config/pdp11/predicates.md (incdec_operand): New predicate.
4575 2018-10-12  Yury Gribov  <tetra2005@gmail.com>
4577         PR middle-end/81376
4578         * real.c (format_helper::can_represent_integral_type_p): New function
4579         * real.h (format_helper::can_represent_integral_type_p): Ditto.
4580         * match.pd: New pattern.
4582 2018-10-12  Alexandre Oliva <oliva@adacore.com>
4584         * configure.ac: Introduce --enable-large-address-aware
4585         to define MINGW_DEFAULT_LARGE_ADDR_AWARE.
4586         * doc/install.texi: Document it.
4587         * configure, config.in: Rebuilt.
4588         * config/i386/mingw32.h (LINK_SPEC_LARGE_ADDR_AWARE): Define,
4589         based on MINGW_DEFAULT_LARGE_ADDR_AWARE.
4590         (LINK_SPEC): Insert it.
4591         * config/i386/mingw-w64.h: Likewise.
4593         * cygming.h (HAVE_GAS_ALIGNED_COMM): Fallback-define.
4595 2018-10-12  Peter Bergner  <bergner@linux.ibm.com>
4597         PR rtl-optimization/87600
4598         * ira-lives (non_conflicting_reg_copy_p): Disable for non LRA targets.
4600 2018-10-12  Paul Koning  <ni1d@arrl.net>
4602         * doc/md.texi (doloop_end): Document that the pattern code may
4603         need to check operand mode.
4605 2018-10-12  Wilco Dijkstra  <wdijkstr@arm.com>
4607         * config/aarch64/aarch64.md (zero_extendsidi2_aarch64): Add alternatives
4608         to zero-extend between int and floating-point registers.
4609         (load_pair_zero_extendsidi2_aarch64): Add alternative for zero-extended
4610         ldp into floating-point registers.  Add type and arch attributes.
4611         (zero_extend<SHORT:mode><GPI:mode>2_aarch64): Add arch attribute.
4612         Use f_loads for type attribute.
4614 2018-10-11  Martin Sebor  <msebor@redhat.com>
4616         * doc/extend.texi (attribute packed): Correct typos.
4618 2018-10-11  Martin Sebor  <msebor@redhat.com>
4620         * doc/extend.texi (attribute flatten): Mention interaction with
4621         noinline.
4623 2018-10-11  Jan Hubicka  <hubicka@ucw.cz>
4625         PR target/87156
4626         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
4627         Set new_decl virtual flag to zero.
4629 2018-10-11  Martin Sebor  <msebor@redhat.com>
4631         PR middle-end/87593
4632         * doc/extend.texi (attribute format_arg): Discuss using multiple
4633         attributes on a single function.
4635 2018-10-11  Giuliano Belinassi  <giuliano.belinassi@usp.br>
4637         PR tree-optimization/86829
4638         * match.pd (sin (atan (x))): New simplification rules.
4639         (cos (atan (x))): Likewise.
4640         * real.c (build_sinatan_real): New function.
4641         * real.h (build_sinatan_real): Prototype.
4643 2018-10-11  Will Schmidt <will_schmidt@vnet.ibm.com>
4645         * config/rs6000/rs6000.c (map_to_integral_tree_type): New helper
4646         function.
4647         (fold_mergeeo_helper): New helper function.
4648         (rs6000_gimple_fold_builtin): Add hooks for vec_mergee and vec_mergeo
4649         intrinsics.  Correct some whitespace indentation issues.
4651 2018-10-11  Wilco Dijkstra  <wdijkstr@arm.com>
4653         PR target/87511
4654         * config/aarch64/aarch64.c (aarch64_mask_and_shift_for_ubfiz_p):
4655         Use HOST_WIDE_INT_1U for shift.
4657 2018-10-11  Doug Rupp  <rupp@adacore.com>
4658             Olivier Hainque  <hainque@adacore.com>
4660         * config/rs6000/vxworks.h (VXWORKS_RELAX_LINK_SPEC): New macro.
4661         Pass --relax to the linker for RTPs.
4662         (LINK_SPEC): Append VXWORKS_RELAX_LINK_SPEC.
4664 2018-10-11  Andrew Stubbs  <ams@codesourcery.com>
4665             Jan Hubicka  <jh@suse.cz>
4666             Martin Jambor  <mjambor@suse.cz>
4668         * print-rtl.c (print_rtx_operand_codes_E_and_V): Print how many times
4669         the same elements are repeated rather than printing all of them.
4670         * read-rtl.c (rtx_reader::read_rtx_operand): Recognize and expand
4671         "repeated" elements.
4672         * read-rtl-function.c (test_loading_repeat): New function.
4673         (read_rtl_function_c_tests): Call test_loading_repeat.
4674         * rtl-tests.c (test_dumping_repeat): New function.
4675         (rtl_tests_c_tests): Call test_dumping_repeat.
4677 2018-10-11  Richard Biener  <rguenther@suse.de>
4679         * config/i386/x86-tune-costs.h (bdver?_memcpy, bdver?_memset,
4680         bdver?_cost): Unify to ...
4681         (bdver_memcpy, bdver_memset, bdver_cost): ... this.
4682         * config/i386/i386.c (processor_cost_table): Adjust.
4684 2018-10-10  Eric Botcazou  <ebotcazou@adacore.com>
4686         PR middle-end/87574
4687         * cgraphunit.c (cgraph_node::expand_thunk): Force DECL_IGNORED_P on
4688         the thunk when expanding to GIMPLE.
4690 2018-10-10  Bernd Edlinger  <bernd.edlinger@hotmail.de>
4692         * varasm.c (mergeable_string_section): Don't try to move zero-length
4693         strings to the merge section.
4695 2018-10-10  Uros Bizjak  <ubizjak@gmail.com>
4697         PR target/87573
4698         * config/i386/mmx.md (const_vector 0 -> mem splitter): New splitter.
4700 2018-10-10  Jakub Jelinek  <jakub@redhat.com>
4702         PR target/87550
4703         * config/i386/i386-builtin.def (IX86_BUILTIN_RDPMC): Move from args set
4704         to special_args set.
4706 2018-10-10  Richard Biener  <rguenther@suse.de>
4708         * config/i386/sse.md (reduc_plus_scal_v8df, reduc_plus_scal_v4df,
4709         reduc_plus_scal_v2df, reduc_plus_scal_v16sf, reduc_plus_scal_v8sf,
4710         reduc_plus_scal_v4sf): Merge into pattern reducing to half width
4711         and recursing and pattern terminating the recursion on SSE
4712         vector width using ix86_expand_reduc.
4713         (reduc_sminmax_scal_<mode>): Split into part reducing to half
4714         width and recursing and SSE2 vector variant doing the final
4715         reduction with ix86_expand_reduc.
4716         (reduc_uminmax_scal_<mode>): Likewise for the AVX512 variants
4717         with terminating the recursion at AVX level, splitting that
4718         to SSE there.
4720 2018-10-09  David Malcolm  <dmalcolm@redhat.com>
4722         * genmatch.c (error_cb): Rename to...
4723         (diagnostic_cb): ...this, converting int params to enums.
4724         (fatal_at): Update for renaming.
4725         (warning_at): Likewise.
4726         (main): Likewise.
4727         * input.c (selftest::ebcdic_execution_charset::apply):
4728         Update for renaming of...
4729         (selftest::ebcdic_execution_charset::on_error): ...this, renaming
4730         to...
4731         (selftest::ebcdic_execution_charset::on_diagnostic): ...this,
4732         converting level and reason to enums.
4733         (class selftest::lexer_error_sink): Rename to...
4734         (class selftest::lexer_test_options): ...this, renaming field
4735         "m_errors" to "m_diagnostics".
4736         (selftest::lexer_test_options::apply): Update for renaming of...
4737         (selftest::lexer_test_options::on_error): ...this, renaming to...
4738         (selftest::lexer_test_options::on_diagnostic): ...this
4739         converting level and reason to enums.
4740         (selftest::test_lexer_string_locations_raw_string_unterminated):
4741         Update for renamings.
4742         * opth-gen.awk (struct cpp_reason_option_codes_t): Use enum for
4743         "reason".
4745 2018-10-09  Paul A. Clarke  <pc@us.ibm.com>
4747         * config.gcc (powerpc*-*-*): Add pmmintrin.h to extra_headers.
4748         * config/rs6000/pmmintrin.h: New file.
4750 2018-10-09  Eric Botcazou  <ebotcazou@adacore.com>
4752         PR tree-optimization/86659
4753         * gimple-match.h (gimple_match_op constructors): Initialize reverse.
4755 2018-10-09  Richard Biener  <rguenther@suse.de>
4757         PR tree-optimization/63155
4758         * tree-ssa-structalias.c: Include tree-ssa.h.
4759         (get_constraint_for_ssa_var): For undefs return nothing_id.
4760         (find_func_aliases): Cleanup PHI handling.
4762 2018-10-09  Richard Biener  <rguenther@suse.de>
4764         * tree-vectorizer.c (dump_stmt_cost): Fix cut&paste missing
4765         replacements.
4767 2018-10-09  Martin Liska  <mliska@suse.cz>
4769         * asan.c (asan_emit_stack_protection): If a stack variable
4770         is located in a same file as current function, then emit
4771         line info into variable definition string.
4773 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4775         * print-rtl.c (rtx_writer::print_rtx_operand_code_i): Print column
4776         information.
4778 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4780         * cgraphunit.c (cgraph_node::expand_thunk): Do not force DECL_IGNORED_P
4781         on the thunk.
4783 2018-10-08  Eric Botcazou  <ebotcazou@adacore.com>
4785         PR tree-optimization/86659
4786         * gimple-match.h (struct gimple_match_op): Add reverse field.
4787         (gimple_match_op::set_op): New overloaded method.
4788         * gimple-match-head.c (maybe_build_generic_op) <BIT_FIELD_REF>: Set
4789         the REF_REVERSE_STORAGE_ORDER flag on the value.
4790         (gimple_simplify) <GIMPLE_ASSIGN>: For BIT_FIELD_REF, propagate the
4791         REF_REVERSE_STORAGE_ORDER flag and avoid simplifying if it is set.
4792         
4793 2018-10-08  Richard Sandiford  <richard.sandiford@arm.com>
4795         PR middle-end/63155
4796         * gimple-ssa-backprop.c (backprop::intersect_uses): Use
4797         FOR_EACH_IMM_USE_FAST instead of FOR_EACH_IMM_USE_STMT.
4799 2018-10-08  H.J. Lu  <hongjiu.lu@intel.com>
4801         PR target/87517
4802         * config/i386/avx512fintrin.h (_mm512_mask_fmaddsub_round_pd):
4803         Defined with __builtin_ia32_vfmaddsubpd512_mask.
4805 2018-10-08  Richard Biener  <rguenther@suse.de>
4807         * config/i386/x86-tune-costs.h (znver1_cost): Make AVX256 vector loads
4808         cost the same as AVX128 ones.
4810 2018-10-08  Paul Koning  <ni1d@arrl.net>
4812         * config/pdp11/pdp11-protos.h (output_block_move): Remove.
4813         (expand_block_move): New function.
4814         * config/pdp11/pdp11.c (output_block_move): Remove.
4815         (expand_block_move): New function.
4816         * config/pdp11/pdp11.h (MOVE_RATIO): New definition.
4817         * config/pdp11/pdp11.md (movmemhi): Use expand_block_move.
4818         (*movmemhi1): Remove.
4820 2018-10-08  Robin Dapp  <rdapp@linux.ibm.com>
4822         * config/s390/2827.md: Increase latencies for some FP instructions.
4824 2018-10-08  Richard Biener  <rguenther@suse.de>
4826         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
4827         Open a dump scope.
4828         * tree-vectorizer.c (dump_stmt_cost): Add cost param and dump it.
4829         * tree-vectorizer.h (dump_stmt_cost): Adjust.
4830         (add_stmt_cost): Dump return value of the hook.
4832 2018-10-08  Richard Biener  <rguenther@suse.de>
4834         PR tree-optimization/63155
4835         * tree-ssa-propagate.c (add_ssa_edge): Do cheap check first.
4836         (ssa_propagation_engine::ssa_propagate): Remove redundant
4837         bitmap bit clearing.
4839 2018-10-05  Peter Bergner  <bergner@linux.ibm.com>
4841         PR rtl-optimization/86939
4842         PR rtl-optimization/87479
4843         * ira.h (non_conflicting_reg_copy_p): New prototype.
4844         * ira-lives.c (ignore_reg_for_conflicts): New static variable.
4845         (make_hard_regno_dead): Don't add conflicts for register
4846         ignore_reg_for_conflicts.
4847         (make_object_dead): Likewise.
4848         (non_conflicting_reg_copy_p): New function.
4849         (process_bb_node_lives): Set ignore_reg_for_conflicts for copies.
4850         Remove special conflict handling of REAL_PIC_OFFSET_TABLE_REGNUM.
4851         * lra-lives.c (ignore_reg_for_conflicts): New static variable.
4852         (make_hard_regno_dead): Don't add conflicts for register
4853         ignore_reg_for_conflicts.  Remove special conflict handling of
4854         REAL_PIC_OFFSET_TABLE_REGNUM.  Remove now unused argument
4855         check_pic_pseudo_p and update callers.
4856         (mark_pseudo_dead): Don't add conflicts for register
4857         ignore_reg_for_conflicts.
4858         (process_bb_lives): Set ignore_reg_for_conflicts for copies.
4860 2018-10-05  Andrew Waterman  <andrew@sifive.com>
4861             Jim Wilson  <jimw@sifive.com>
4863         * config/riscv/riscv.md (f<quiet_pattern>_quiet<ANYF:mode><X:mode>4):
4864         Add define_expand.  Add ! HONOR_SNANS check to current pattern.  Add
4865         new pattern using HONOR_SNANS that emits one extra instruction.
4867 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4869         * config/rs6000/rs6000.md (unnamed mfcr scc_comparison_operator
4870         patterns): Merge SI and DI patterns to a GPR pattern.
4871         (unnamed define_insn and define_split for record form of that): Merge
4872         to a single define_insn_and_split pattern.
4874 2018-10-05  David Malcolm  <dmalcolm@redhat.com>
4876         PR c++/56856
4877         * input.c (expand_location_to_spelling_point): Add param "aspect"
4878         and use rather than hardcoding LOCATION_ASPECT_CARET.
4879         (get_substring_ranges_for_loc): Handle the case of a single token
4880         within a macro expansion.
4881         * input.h (expand_location_to_spelling_point): Add "aspect" param,
4882         defaulting to LOCATION_ASPECT_CARET.
4884 2018-10-05  Paul Koning  <ni1d@arrl.net>
4886         * config/pdp11/pdp11.c (TARGET_CXX_GUARD_TYPE): Define.
4887         (TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT): Define.
4888         (TARGET_CXX_LIBRARY_RTTI_COMDAT): Define.
4889         (pdp11_guard_type): New function.
4891 2018-10-05  Paul Koning  <ni1d@arrl.net>
4893         * config/pdp11/pdp11.h (FLOAT_TYPE_SIZE): Always 32.
4894         * config/pdp11/pdp11.opt (mfloat32): Remove.
4895         (mfloat64): Remove.
4896         * doc/invoke.texi (pdp11 -mfloat32): Remove:
4897         (pdp11 -mfloat64): Remove.
4899 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4901         * config/i386/i386.md (*cmpxf_cc_i387): Remove pattern.
4902         (*cmp<mode>_cc_i387): Ditto.
4903         (*cmpu<mode>_cc_i387): Ditto.
4904         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
4905         * config/i386/i386.c (ix86_expand_fp_compare): Remove
4906         "scratch" argument.
4907         <case IX86_FPCMP_SAHF>: Do not generate pattern with HImode clobber.
4908         Emit x86_sahf_1 pattern.
4909         (ix86_expand_compare): Update call to ix86_expand_fp_compare.
4910         (ix86_expand_carry_flag_compare): Ditto.
4912 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4914         * config/i386/i386.md (*cmpxf_i387): Change operand 2 predicate
4915         to reg_or_0_operand.  Add "C" constraint.
4916         (*cmpxf_cc_i387): Ditto.
4917         (*cmp<mode>_i387): Change operand 2 predicate
4918         to nonimm_or_0_operand.  Add "C" constraint.
4919         (*cmp<mode>_cc_i387): Ditto.
4920         (*cmp<mode>_0_i387): Remove insn pattern.
4921         (*cmp<mode>_0_cc_i387): Ditto.
4923 2018-10-05  Uros Bizjak  <ubizjak@gmail.com>
4925         * config/i386/constraints.md ("C"): Do not depend on TARGET_SSE.
4926         * config/i386/predicates.md (nonimm_or_0_operand): Rename
4927         from vector_move_operand.  Update all uses.
4929 2018-10-05  Martin Sebor  <msebor@redhat.com>
4931         PR tree-optimization/87490
4932         * builtins.c (expand_builtin_strnlen): Handle a null data.decl
4933         consistently.
4935 2018-10-05  Richard Biener  <rguenther@suse.de>
4937         PR tree-optimization/63155
4938         * tree-ssa-ccp.c (ccp_propagate::visit_phi): Avoid excess
4939         vertical space in dumpfiles.
4940         * tree-ssa-propagate.h
4941         (ssa_propagation_engine::process_ssa_edge_worklist): Remove.
4942         * tree-ssa-propagate.c (cfg_blocks_back): New global.
4943         (ssa_edge_worklist_back): Likewise.
4944         (curr_order): Likewise.
4945         (cfg_blocks_get): Remove abstraction.
4946         (cfg_blocks_add): Likewise.
4947         (cfg_blocks_empty_p): Likewise.
4948         (add_ssa_edge): Add to current or next worklist based on
4949         RPO index.
4950         (add_control_edge): Likewise.
4951         (ssa_propagation_engine::process_ssa_edge_worklist): Fold
4952         into ...
4953         (ssa_propagation_engine::ssa_propagate): ... here.  Unify
4954         iteration from CFG and SSA edge worklist so we process
4955         everything in RPO order, prioritizing forward progress
4956         over iteration.
4957         (ssa_prop_init): Allocate new worklists, do not dump
4958         immediate uses.
4959         (ssa_prop_fini): Free new worklists.
4961 2018-10-05  Richard Biener  <rguenther@suse.de>
4963         * tree-core.h (tree_block::abstract_flag): Remove.
4964         (tree_block::block_num): Make full 32bits.
4965         * tree.def (BLOCK): Remove docs about BLOCK_ABSTRACT.
4966         * tree.h (BLOCK_ABSTRACT): Remove.
4967         * dwarf2out.c (gen_lexical_block_die): Remove dead code
4968         resulting from BLOCK_ABSTRACT being always false.
4969         (gen_inlined_subroutine_die): Likewise.
4970         (gen_block_die): Likewise.
4971         * tree.c (block_ultimate_origin): Likewise.
4972         * tree-pretty-print.c (dump_block_node): Remove code dealing
4973         with BLOCK_ABSTRACT.
4974         * tree-ssa-live.c (dump_scope_block): Likewise.
4975         * tree-streamer-in.c (unpack_ts_block_value_fields): Likewise.
4976         * tree-streamer-out.c (pack_ts_block_value_fields): Likewise.
4978 2018-10-05   Richard Biener  <rguenther@suse.de>
4980         * config/i386/i386.c (ix86_add_stmt_cost): When scalar cost
4981         is asked for initialize mode to the component mode of the
4982         vector type.
4984 2018-10-05  H.J. Lu  <hongjiu.lu@intel.com>
4986         PR target/87522
4987         * config/i386/gnu-user.h (ASM_SPEC): Don't pass -msse2avx to
4988         assembler for -mavx.
4989         * config/i386/gnu-user64.h (ASM_SPEC): Likewise.
4991 2018-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
4993         PR target/87509
4994         * config/rs6000/rs6000-builtin.def (RS6000_BUILTIN_SET_FPSCR_DRN): Use
4995         RS6000_BTM_DFP.
4996         * config/rs6000/rs6000.md (rs6000_set_fpscr_rn): Require the operand
4997         to be DImode.  When using mffscrn, force the operand to a register.
4999 2018-10-04  Uros Bizjak  <ubizjak@gmail.com>
5001         * config/i386/i386.md (*fop_<X87MODEF:mode>_2_i387): Macroize insn
5002         from *fop_<MODEF:mode>_2_i387 and *fop_xf_2_i387 using
5003         X87MODEF mode iterator.
5004         (*fop_<X87MODEF:mode>_3_i387): Macroize insn from
5005         *fop_<MODEF:mode>_3_i387 and *fop_xf_3_i387 using
5006         X87MODEF mode iterator.
5008 2018-10-04  Vinay Kumar  <vinay.kumar@blackfigtech.com>
5010         * doc/invoke.texi (-Wno-prio-ctor-dtor): Document new warning
5011         -Wno-prio-ctor-dtor.
5013 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5015         * Makefile.in (OBJS): Add opt-problem.o.
5016         * dump-context.h: Include "selftest.h.
5017         (selftest::temp_dump_context): New forward decl.
5018         (class dump_context): Make friend of class
5019         selftest::temp_dump_context.
5020         (dump_context::dump_loc_immediate): New decl.
5021         (class dump_pretty_printer): Move here from dumpfile.c.
5022         (class temp_dump_context): Move to namespace selftest.
5023         (temp_dump_context::temp_dump_context): Add param
5024         "forcibly_enable_dumping".
5025         (selftest::verify_dumped_text):
5026         (ASSERT_DUMPED_TEXT_EQ): Move here from dumpfile.c.
5027         (selftest::verify_item):
5028         (ASSERT_IS_TEXT): Move here from dumpfile.c.
5029         (ASSERT_IS_TREE): Likewise.
5030         (ASSERT_IS_GIMPLE): Likewise.
5031         * dumpfile.c (dump_context::dump_loc): Move immediate dumping
5032         to...
5033         (dump_context::dump_loc_immediate): ...this new function.
5034         (class dump_pretty_printer): Move to dump-context.h.
5035         (dump_switch_p_1): Don't enable MSG_PRIORITY_REEMITTED.
5036         (opt_info_switch_p_1): Enable MSG_PRIORITY_REEMITTED.
5037         (temp_dump_context::temp_dump_context): Move to "selftest"
5038         namespace.  Add param "forcibly_enable_dumping", and use it to
5039         conditionalize the use of m_pp;
5040         (selftest::verify_dumped_text): Make non-static.
5041         (ASSERT_DUMPED_TEXT_EQ): Move to dump-context.h.
5042         (selftest::verify_item): Make non-static.
5043         (ASSERT_IS_TEXT): Move to dump-context.h.
5044         (ASSERT_IS_TREE): Likewise.
5045         (ASSERT_IS_GIMPLE): Likewise.
5046         (selftest::test_capture_of_dump_calls): Pass "true" for new
5047         param of temp_dump_context.
5048         * dumpfile.h (enum dump_flag): Add MSG_PRIORITY_REEMITTED, adding
5049         it to MSG_ALL_PRIORITIES.  Update values of TDF_COMPARE_DEBUG and
5050         TDF_COMPARE_DEBUG.
5051         * opt-problem.cc: New file.
5052         * opt-problem.h: New file.
5053         * optinfo-emit-json.cc
5054         (selftest::test_building_json_from_dump_calls): Pass "true" for
5055         new param of temp_dump_context.
5056         * optinfo.cc (optinfo_kind_to_dump_flag): New function.
5057         (optinfo::emit_for_opt_problem): New function.
5058         (optinfo::emit): Clarity which emit_item is used.
5059         * optinfo.h (optinfo::get_dump_location): New accessor.
5060         (optinfo::emit_for_opt_problem): New decl.
5061         (optinfo::emit): Make const.
5062         * selftest-run-tests.c (selftest::run_tests): Call
5063         selftest::opt_problem_cc_tests.
5064         * selftest.h (selftest::opt_problem_cc_tests): New decl.
5065         * tree-data-ref.c (dr_analyze_innermost): Convert return type from
5066         bool to opt_result, converting fprintf messages to
5067         opt_result::failure_at calls.  Add "stmt" param for use by the
5068         failure_at calls.
5069         (create_data_ref): Pass "stmt" to the dr_analyze_innermost call.
5070         (runtime_alias_check_p): Convert return type from bool to
5071         opt_result, converting dump_printf calls to
5072         opt_result::failure_at, using the statement DDR_A for their
5073         location.
5074         (find_data_references_in_stmt): Convert return type from bool to
5075         opt_result, converting "return false" to opt_result::failure_at
5076         with a new message.
5077         * tree-data-ref.h: Include "opt-problem.h".
5078         (dr_analyze_innermost): Convert return type from bool to opt_result,
5079         and add a const gimple * param.
5080         (find_data_references_in_stmt): Convert return type from bool to
5081         opt_result.
5082         (runtime_alias_check_p): Likewise.
5083         * tree-predcom.c (find_looparound_phi): Pass "init_stmt" to
5084         dr_analyze_innermost.
5085         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test):
5086         Convert return type from bool to opt_result, adding a message for
5087         the PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS zero case.
5088         (vect_analyze_data_ref_dependence): Convert return type from bool
5089         to opt_result.  Change sense of return type from "false"
5090         effectively meaning "no problems" to "false" meaning a problem,
5091         so that "return false" becomes "return opt_result::success".
5092         Convert "return true" calls to opt_result::failure_at, using
5093         the location of statement A rather than vect_location.
5094         (vect_analyze_data_ref_dependences): Convert return type from bool
5095         to opt_result.
5096         (verify_data_ref_alignment): Likewise, converting dump_printf_loc
5097         calls to opt_result::failure_at, using the stmt location rather
5098         than vect_location.
5099         (vect_verify_datarefs_alignment): Convert return type from bool
5100         to opt_result.
5101         (vect_enhance_data_refs_alignment): Likewise.  Split local "stat"
5102         into multiple more-tightly-scoped copies.
5103         (vect_analyze_data_refs_alignment): Convert return type from bool
5104         to opt_result.
5105         (vect_analyze_data_ref_accesses): Likewise, converting a
5106         "return false" to a "return opt_result::failure_at", adding a
5107         new message.
5108         (vect_prune_runtime_alias_test_list): Convert return type from
5109         bool to opt_result, converting dump_printf_loc to
5110         opt_result::failure_at.  Add a %G to show the pertinent statement,
5111         and use the stmt's location rather than vect_location.
5112         (vect_find_stmt_data_reference): Convert return type from
5113         bool to opt_result, converting dump_printf_loc to
5114         opt_result::failure_at, using stmt's location.
5115         (vect_analyze_data_refs):  Convert return type from bool to
5116         opt_result.  Convert "return false" to "return
5117         opt_result::failure_at", adding messages as needed.
5118         * tree-vect-loop.c (vect_determine_vf_for_stmt_1): Convert return
5119         type from bool to opt_result.
5120         (vect_determine_vf_for_stmt): Likewise.
5121         (vect_determine_vectorization_factor): Likewise, converting
5122         dump_printf_loc to opt_result::failure_at, using location of phi
5123         rather than vect_location.
5124         (vect_analyze_loop_form_1): Convert return type from bool to
5125         opt_result, converting dump_printf_loc calls, retaining the use of
5126         vect_location.
5127         (vect_analyze_loop_form): Convert return type from loop_vec_info
5128         to opt_loop_vec_info.
5129         (vect_analyze_loop_operations): Convert return type from bool to
5130         opt_result, converting dump_printf_loc calls, using the location
5131         of phi/stmt rather than vect_location where available.  Convert
5132         various "return false" to "return opt_result::failure_at" with
5133         "unsupported phi" messages.
5134         (vect_get_datarefs_in_loop): Convert return type from bool to
5135         opt_result.  Add a message for the
5136         PARAM_LOOP_MAX_DATAREFS_FOR_DATADEPS failure.
5137         (vect_analyze_loop_2): Convert return type from bool to
5138         opt_result.  Ensure "ok" is set to a opt_result::failure_at before
5139         each "goto again;", adding new messages where needed.
5140         Add "unsupported grouped {store|load}" messages.
5141         (vect_analyze_loop): Convert return type from loop_vec_info to
5142         opt_loop_vec_info.
5143         * tree-vect-slp.c (vect_analyze_slp): Convert return type from
5144         bool to opt_result.
5145         * tree-vect-stmts.c (process_use): Likewise, converting
5146         dump_printf_loc call and using stmt location, rather than
5147         vect_location.
5148         (vect_mark_stmts_to_be_vectorized): Likeise.
5149         (vect_analyze_stmt): Likewise, adding a %G.
5150         (vect_get_vector_types_for_stmt): Convert return type from bool to
5151         opt_result, converting dump_printf_loc calls and using stmt
5152         location, rather than vect_location.
5153         (vect_get_mask_type_for_stmt): Convert return type from tree to
5154         opt_tree, converting dump_printf_loc calls and using stmt location.
5155         * tree-vectorizer.c: Include "opt-problem.h.
5156         (try_vectorize_loop_1): Flag "Analyzing loop at" dump message as
5157         MSG_PRIORITY_INTERNALS.  Convert local "loop_vinfo" from
5158         loop_vec_info to opt_loop_vec_info.  If if fails, and dumping is
5159         enabled, use it to report at the top level "couldn't vectorize
5160         loop" followed by the problem.
5161         * tree-vectorizer.h (opt_loop_vec_info): New typedef.
5162         (vect_mark_stmts_to_be_vectorized): Convert return type from bool
5163         to opt_result.
5164         (vect_analyze_stmt): Likewise.
5165         (vect_get_vector_types_for_stmt): Likewise.
5166         (tree vect_get_mask_type_for_stmt): Likewise.
5167         (vect_analyze_data_ref_dependences): Likewise.
5168         (vect_enhance_data_refs_alignment): Likewise.
5169         (vect_analyze_data_refs_alignment): Likewise.
5170         (vect_verify_datarefs_alignment): Likewise.
5171         (vect_analyze_data_ref_accesses): Likewise.
5172         (vect_prune_runtime_alias_test_list): Likewise.
5173         (vect_find_stmt_data_reference): Likewise.
5174         (vect_analyze_data_refs): Likewise.
5175         (vect_analyze_loop): Convert return type from loop_vec_info to
5176         opt_loop_vec_info.
5177         (vect_analyze_loop_form): Likewise.
5178         (vect_analyze_slp): Convert return type from bool to opt_result.
5180 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5182         * doc/invoke.texi (-fopt-info): Document new "internals"
5183         sub-option.
5184         * dump-context.h (dump_context::apply_dump_filter_p): New decl.
5185         * dumpfile.c (dump_options): Update for renaming of MSG_ALL to
5186         MSG_ALL_KINDS.
5187         (optinfo_verbosity_options): Add "internals".
5188         (kind_as_string): Update for renaming of MSG_ALL to MSG_ALL_KINDS.
5189         (dump_context::apply_dump_filter_p): New member function.
5190         (dump_context::dump_loc): Use apply_dump_filter_p rather than
5191         explicitly masking the dump_kind.
5192         (dump_context::begin_scope): Increment the scope depth first.  Use
5193         apply_dump_filter_p rather than explicitly masking the dump_kind.
5194         (dump_context::emit_item): Use apply_dump_filter_p rather than
5195         explicitly masking the dump_kind.
5196         (dump_dec): Likewise.
5197         (dump_hex): Likewise.
5198         (dump_switch_p_1): Default to MSG_ALL_PRIORITIES.
5199         (opt_info_switch_p_1): Default to MSG_PRIORITY_USER_FACING.
5200         (opt_info_switch_p): Update handling of default
5201         MSG_OPTIMIZED_LOCATIONS to cope with default of
5202         MSG_PRIORITY_USER_FACING.
5203         (dump_basic_block): Use apply_dump_filter_p rather than explicitly
5204         masking the dump_kind.
5205         (selftest::test_capture_of_dump_calls): Update test_dump_context
5206         instances to use MSG_ALL_KINDS | MSG_PRIORITY_USER_FACING rather
5207         than MSG_ALL.  Generalize scope test to be run at all four
5208         combinations of with/without MSG_PRIORITY_USER_FACING and
5209         MSG_PRIORITY_INTERNALS, adding examples of explicit priority
5210         for each of the two values.
5211         * dumpfile.h (enum dump_flag): Add comment about the MSG_* flags.
5212         Rename MSG_ALL to MSG_ALL_KINDS.  Add MSG_PRIORITY_USER_FACING,
5213         MSG_PRIORITY_INTERNALS, and MSG_ALL_PRIORITIES, updating the
5214         values for TDF_COMPARE_DEBUG and TDF_ALL_VALUES.
5215         (AUTO_DUMP_SCOPE): Add a note to the comment about the interaction
5216         with MSG_PRIORITY_*.
5217         * tree-vect-loop-manip.c (vect_loop_versioning): Mark versioning
5218         dump messages as MSG_PRIORITY_USER_FACING.
5219         * tree-vectorizer.h (DUMP_VECT_SCOPE): Add a note to the comment
5220         about the interaction with MSG_PRIORITY_*.
5222 2018-10-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
5224         * varasm.c (output_constant): Add new parameter merge_strings.
5225         Make strings properly zero terminated in merge string sections.
5226         (mergeable_string_section): Don't fail if the last char is non-zero.
5227         (assemble_variable_contents): Handle merge string sections.
5228         (assemble_variable): Likewise.
5229         (assemble_constant_contents): Likewise.
5230         (output_constant_def_contents): Likewise.
5231         (output_constructor_array_range,
5232         output_constructor_regular_field): Adjust call to output_constant.
5233         (output_object_block): Adjust call to assemble_constant_contents
5234         and assemble_variable_contents.
5236 2018-10-04  Martin Liska  <mliska@suse.cz>
5238         PR c/87483
5239         * cgraphunit.c (process_function_and_variable_attributes):
5240         Warn about a function with alias attribute and a body.
5242 2018-10-04  Martin Liska  <mliska@suse.cz>
5244         PR ipa/82625
5245         * multiple_target.c (redirect_to_specific_clone): New function.
5246         (ipa_target_clone): Use it.
5247         * tree-inline.c: Fix comment.
5249 2018-10-04  David Malcolm  <dmalcolm@redhat.com>
5251         * dumpfile.c (gcc::dump_manager::dump_manager): Initialize new
5252         fields.
5253         (gcc::dump_manager::~dump_manager): Free m_optinfo_filename.
5254         (gcc::dump_manager::register_pass): New member function, adapted
5255         from loop body in gcc::pass_manager::register_pass, adding a
5256         call to update_dfi_for_opt_info.
5257         (gcc::dump_manager::opt_info_enable_passes): Store the
5258         -fopt-info options into the new fields.  Move the loop
5259         bodies into...
5260         (gcc::dump_manager::update_dfi_for_opt_info): ...this new member
5261         function.
5262         * dumpfile.h (struct opt_pass): New forward decl.
5263         (gcc::dump_manager::register_pass): New decl.
5264         (gcc::dump_manager::update_dfi_for_opt_info): New decl.
5265         (class gcc::dump_manager): Add fields "m_optgroup_flags",
5266         "m_optinfo_flags", and "m_optinfo_filename".
5267         * passes.c (gcc::pass_manager::register_pass): Move all of the
5268         dump-handling code to gcc::dump_manager::register_pass.
5270 2018-10-04  Peter Bergner  <bergner@linux.ibm.com>
5272         PR rtl-optimization/87466
5273         * target.def (setjmp_preserves_nonvolatile_regs_p): New target hook.
5274         * doc/tm.texi.in (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P): New hook.
5275         * doc/tm.texi: Regenerate.
5276         * ira-lives.c (process_bb_node_lives): Use the new target hook.
5277         * lra-lives.c (process_bb_lives): Likewise.
5278         * config/rs6000/rs6000.c (TARGET_SETJMP_PRESERVES_NONVOLATILE_REGS_P):
5279         Define.
5281 2018-10-04  Tamar Christina  <tamar.christina@arm.com>
5283         * params.c (add_params): Fix initialization.
5285 2018-10-04  Martin Liska  <mliska@suse.cz>
5287         PR gcov-profile/84107
5288         * tree-profile.c (init_ic_make_global_vars):
5289         Remove ic_void_ptr_var and ic_gcov_type_ptr_var.
5290         Come up with new ic_tuple* variables.  Emit
5291         __gcov_indirect_call{,_topn} variables.
5292         (gimple_gen_ic_profiler): Access the variable
5293         and emit gimple.
5294         (gimple_gen_ic_func_profiler): Access
5295         __gcov_indirect_call.callee field.
5296         (gimple_init_gcov_profiler): Use ptr_type_node.
5297         * value-prof.c (gimple_ic): Use ptr_type_node.
5299 2018-10-04  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5301         PR tree-optimization/85787
5302         * ipa-pure-const.c (malloc_candidate_p_1): Move most of
5303         malloc_candidate_p into this function and add support for
5304         detecting multiple phis.
5305         (DUMP_AND_RETURN): Move from malloc_candidate_p into top-level macro.
5307 2018-10-04  Martin Liska  <mliska@suse.cz>
5309         PR ipa/87491
5310         * ipa-inline.c (inline_to_all_callers_1):
5311         Call ultimate_alias_target for node being inlined.
5313 2018-10-03  Jeff Law  <law@redhat.com>
5315         * gimple-ssa-sprintf.c (format_string): Do not hardcode size of
5316         target's wchar_t.
5317         * tree.c (get_typenode_from_name): Moved from fortran/trans-types.c.
5318         * tree.h (get_typenode_from_name): Prototype.
5320 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
5322         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
5323         Change operand 2 predicate to nonimmediate_operand.
5324         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
5326 2018-10-03  Martin Sebor  <msebor@redhat.com>
5327             Jeff Law  <law@redhat.com>
5329         * gimple-ssa-sprintf.c (struct fmtresult): Add new member and
5330         initialize it.
5331         (get_string_length): Detect unterminated arrays.
5332         (format_string): Same.
5333         (format_directive): Warn about unterminated arrays.
5334         (handle_gimple_call): Mark statements with no_warning as needed.
5336 2018-10-03  Jim Wilson  <jimw@sifive.com>
5338         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): For ABI_ILP32E,
5339         also define __riscv_abi_rve.  Delete trailing white space.
5341 2018-10-03  Paul Koning  <ni1d@arrl.net>
5343         Enable LRA register allocator for PDP11.
5344         * config/pdp11/constraints.md (Q): Use define_memory_constraint.
5345         (R): Likewise.
5346         (D): Likewise.
5347         * config/pdp11/pdp11.c (pdp11_lra_p): New function.
5348         * config/pdp11/pdp11.opt (-mlra): New option.
5349         * doc/invoke.texi (PDP-11 Options): Document -mlra.
5351 2018-10-03  Uros Bizjak  <ubizjak@gmail.com>
5353         * config/i386/i386.md (*<absneg:code>extendsfdf2): Remove.
5354         (*<absneg:code>extend<mode>xf2): Ditto.
5356 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
5358         PR tree-optimization/87415
5359         * tree-vrp.c (set_value_range_with_overflow): Special case one bit
5360         precision fields.
5362 2018-10-02  Jeff Law  <law@redhat.com>
5364         * gimple-fold.c (get_range_strlen): Only set *nonstr when
5365         an unterminated string is discovered.  Bubble up range
5366         even for unterminated strings.
5367         (gimple_fold_builtin_strlen): Do not fold if get_range_strlen
5368         indicates the string was not terminated via NONSTR.
5370 2018-10-03  Aldy Hernandez  <aldyh@redhat.com>
5372         * tree-vrp.c (extract_range_from_unary_expr): Special case all
5373         pointer conversions.
5374         Do not do anything special for anti-ranges.
5376 2018-10-03  Jérôme Lambourg  <lambourg@adacore.com>
5378         * config/arm/vxworks.h (ARM_TARGET2_DWARF_FORMAT): Adjust to
5379         DW_EH_PE_pcrel | DW_EH_PE_indirect for RTPs.
5381 2018-10-03  Martin Liska  <mliska@suse.cz>
5383         PR gcov-profile/86109
5384         * coverage.c (coverage_begin_function): Do not
5385         mark lambdas as artificial.
5386         * tree-core.h (struct GTY): Remove tm_clone_flag
5387         and introduce new lambda_function.
5388         * tree.h (DECL_LAMBDA_FUNCTION): New macro.
5390 2018-10-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
5392         PR target/87474
5393         * config/rs6000/rs6000-string.c (expand_strn_compare): Check that both
5394         P8_VECTOR and VSX are enabled.
5396 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
5398         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
5399         0x3907 as CPU model number.
5401 2018-10-02  Andreas Krebbel  <krebbel@linux.ibm.com>
5403         * common/config/s390/s390-common.c: Rename PF_ARCH12 to PF_Z14.
5404         * config/s390/s390.h (enum processor_flags): Rename PF_ARCH12 to
5405         PF_Z14.  Rename TARGET_CPU_ARCH12 to TARGET_CPU_Z14,
5406         TARGET_CPU_ARCH12_P to TARGET_CPU_Z14_P, TARGET_ARCH12 to
5407         TARGET_Z14, and TARGET_ARCH12_P to TARGET_Z14_P.
5408         * config/s390/s390.md: Likewise. Rename also the cpu attribute
5409         value from arch12 to z14.
5411 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
5413         * config/i386/i386.md (fxam<mode>2_i387_with_temp): Remove.
5414         (isinfxf2): Ditto.
5415         (isinf<mode>2): Ditto.
5417 2018-10-02  Uros Bizjak  <ubizjak@gmail.com>
5419         * config/i386/i386.c (ix86_emit_i387_round): Extend op1 to XFmode
5420         before emitting fxam.  Perform calculations in XFmode.
5422 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5424         * match.pd (((X /[ex] A) +- B) * A): New transformation.
5426 2018-10-02  Marc Glisse  <marc.glisse@inria.fr>
5428         PR middle-end/87319
5429         * fold-const.c (fold_plusminus_mult_expr): Handle complex and vectors.
5430         * tree.c (signed_or_unsigned_type_for): Handle complex.
5432 2018-10-02  Jeff Law  <law@redhat.com>
5434         * gimple-fold.c (get_range_strlen): Remove dead code.
5436 2018-10-02  Martin Sebor  <msebor@redhat.com>
5437             Jeff Law  <law@redhat.com>
5439         * builtins.c (unterminated_array): Add new arguments.
5440         If argument is not terminated, bubble up size and exact
5441         state to callers.
5442         (expand_builtin_strnlen): Detect, avoid expanding
5443         and diagnose unterminated arrays.
5444         (c_strlen): Fill in offset of start of unterminated strings.
5445         * builtins.h (unterminated_array): Update prototype.
5447 2018-10-02  Richard Biener  <rguenther@suse.de>
5449         * config/i386/sse.md (reduc_plus_scal_v4df): Avoid the use
5450         of haddv4df, first reduce to SSE width and exploit the fact
5451         that we only need element zero with the reduction result.
5452         (reduc_plus_scal_v2df): Likewise.
5454 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5456         * dojump.h (do_jump): Delete.
5457         (do_jump_1): Likewise.
5458         (split_comparison): Move around.
5459         * dojump.c (do_jump): Make static.
5460         (do_jump_1): Likewise.
5461         (jumpifnot): Move around.
5462         (jumpifnot_1): Likewise.
5463         (jumpif): Likewise.
5464         (jumpif_1): Likewise.
5465         * expr.c (expand_expr_real_1): Call jumpif[not] instead of do_jump.
5467 2018-10-02  Eric Botcazou  <ebotcazou@adacore.com>
5469         * reorg.c (make_return_insns): Use emit_copy_of_insn_after for the
5470         insns in the delay slot and add_insn_after for the jump insn.
5472 2018-10-02  Richard Biener  <rguenther@suse.de>
5474         * tree-inline.c (expand_call_inline): Use the location of
5475         the callee declaration for the inline-entry marker.
5476         * final.c (notice_source_line): Remove special-casing of
5477         NOTE_INSN_INLINE_ENTRY.
5479 2018-10-01  Carl Love  <cel@us.ibm.com>
5481         PR 69431
5482         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
5483         (__builtin_mtfsb0): New.
5484         (__builtin_mtfsb1): New.
5485         ( __builtin_set_fpscr_rn): New.
5486         (__builtin_set_fpscr_drn): New.
5487         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
5488         (rs6000_expand_set_fpscr_rn_builtin): Add.
5489         (rs6000_expand_set_fpscr_drn_builtin): Add.
5490         (rs6000_expand_builtin): Add case statement entries for
5491         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
5492         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
5493         RS6000_BUILTIN_MFFSL.
5494         (rs6000_init_builtins): Add ftype initialization and def_builtin
5495         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
5496         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
5497         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
5498         rs6000_mffscdrn): Add define_insn.
5499         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
5500         * doc/extend.texi: Add documentation for the builtins.
5502 2018-10-01  Richard Biener  <rguenther@suse.de>
5504         PR tree-optimization/87465
5505         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
5506         causing branch miscounts.
5508 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5510         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
5511         aarch64_option_default_param):  New.
5512         (params.h): Include.
5513         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
5514         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
5515         stack-clash protection validation code.
5517 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5519         * params.c (validate_param): New.
5520         (add_params): Use it.
5521         (set_param_value): Refactor param validation into validate_param.
5522         (diagnostic.h): Include.
5523         * diagnostic.h (diagnostic_ready_p): New.
5525 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5527         * params.c (set_param_value):
5528         Add index of parameter being validated.
5529         * common/common-target.def (option_validate_param): New.
5530         * common/common-targhooks.h (default_option_validate_param): New.
5531         * common/common-targhooks.c (default_option_validate_param): New.
5532         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
5533         * doc/tm.texi: Regenerate.
5535 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5537         PR target/86486
5538         * config/aarch64/aarch64.c (aarch64_override_options_internal):
5539         Add validation for stack-clash parameters and set defaults.
5541 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5543         PR target/86486
5544         * configure.ac: Add stack-clash-protection-guard-size.
5545         * doc/install.texi: Document it.
5546         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
5547         * params.def: Update comment for guard-size.
5548         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
5549         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
5550         * configure: Regenerate.
5552 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5554         PR target/86486
5555         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
5556         STACK_DYNAMIC_OFFSET): New.
5557         * config/aarch64/aarch64.c (aarch64_layout_frame):
5558         Update outgoing args size.
5559         (aarch64_stack_clash_protection_alloca_probe_range,
5560         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5562 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5564         PR target/86486
5565         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
5566         probe ranges.
5567         * target.def (stack_clash_protection_alloca_probe_range): New.
5568         (stack_clash_protection_final_dynamic_probe): Remove.
5569         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
5570         (default_stack_clash_protection_final_dynamic_probe): Remove.
5571         * targhooks.c: Likewise.
5572         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
5573         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
5574         * doc/tm.texi: Regenerate.
5576 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5578         PR target/86486
5579         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
5580         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
5581         aarch64_clamp_to_uimm12_shift): New.
5582         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
5583         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
5585 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
5587         PR target/86486
5588         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
5590 2018-10-01  Jeff Law  <law@redhat.com>
5591             Richard Sandiford <richard.sandiford@linaro.org>
5592             Tamar Christina  <tamar.christina@arm.com>
5594         PR target/86486
5595         * config/aarch64/aarch64.md
5596         (probe_stack_range): Add k (SP) constraint.
5597         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
5598         STACK_CLASH_MAX_UNROLL_PAGES): New.
5599         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
5600         stack probes for stack clash.
5601         (aarch64_allocate_and_probe_stack_space): New.
5602         (aarch64_expand_prologue): Use it.
5603         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
5604         (aarch64_sub_sp): Add emit_move_imm optional param.
5606 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
5608         PR tree-optimization/87261
5609         * match.pd: Remove trailing whitespace.
5610         Add (x & y) | ~(x | y) -> ~(x ^ y),
5611         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
5613 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5615         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
5616         constraints.
5617         (ashlsi3_insn): Update instruction constraints.
5618         (ashrsi3_insn): Likewise.
5619         (rotrsi3): Likewise.
5620         (add_shift): Likewise.
5621         * config/arc/constraints.md (Csz): New 32 bit constraint. It
5622         avoids placing in the limm field small constants which, otherwise,
5623         could end into a small instruction.
5625 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
5627         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
5628         destination register is not odd-even.
5629         (umaddsidi4_split): Likewise.
5631 2018-10-01  Richard Biener  <rguenther@suse.de>
5633         * tree-inline.c (expand_call_inline): Store origin of fn
5634         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
5635         * tree.c (block_ultimate_origin): Simplify and do some
5636         checking.
5638 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
5640         * config/i386/mmx.md (EMMS): New int iterator.
5641         (emms): New int attribute.
5642         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
5643         EMMS int iterator.  Explicitly declare clobbers.
5644         (mmx_emms): Remove expander.
5645         (mmx_femms): Ditto.
5646         * config/i386/predicates.md (emms_operation): Remove predicate.
5647         (vzeroall_pattern): New predicate.
5648         (vzeroupper_pattern): Rename from vzeroupper_operation.
5649         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
5650         vzeroupper_pattern and vzeroall_pattern predicates.
5652 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
5654         PR rtl-optimization/86939
5655         * ira-lives.c (make_hard_regno_born): Rename from this...
5656         (make_hard_regno_live): ... to this.  Remove update to conflict
5657         information.  Update function comment.
5658         (make_hard_regno_dead): Add conflict information update.  Update
5659         function comment.
5660         (make_object_born): Rename from this...
5661         (make_object_live): ... to this.  Remove update to conflict information.
5662         Update function comment.
5663         (make_object_dead):  Add conflict information update.  Update function
5664         comment.
5665         (mark_pseudo_regno_live): Call make_object_live.
5666         (mark_pseudo_regno_subword_live): Likewise.
5667         (mark_hard_reg_dead): Update function comment.
5668         (mark_hard_reg_live): Call make_hard_regno_live.
5669         (process_bb_node_lives): Likewise.
5670         * lra-lives.c (make_hard_regno_born): Rename from this...
5671         (make_hard_regno_live): ... to this.  Remove update to conflict
5672         information.  Remove now uneeded check_pic_pseudo_p argument.
5673         Update function comment.
5674         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
5675         to conflict information.  Update function comment.
5676         (mark_pseudo_live): Remove update to conflict information.  Update
5677         function comment.
5678         (mark_pseudo_dead): Add conflict information update.
5679         (mark_regno_live): Call make_hard_regno_live.
5680         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
5681         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
5683 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
5685         PR target/87370
5686         * config/i386/i386.c (construct_container): Use TImode for
5687         BLKmode values in 2 integer registers.
5689 2018-09-29  Jeff Law  <law@redhat.com>
5691         * builtins.c (unterminated_array): Pass in c_strlen_data * to
5692         c_strlen rather than just a tree *.
5693         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
5694         Update recursive calls appropriately.  If caller did not provide a
5695         suitable data pointer, create a local one.  When a non-terminated
5696         string is discovered, bubble up information about the string via the
5697         c_strlen_data object.
5698         * builtins.h (c_strlen): Update prototype.
5699         (c_strlen_data): New structure.
5700         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
5701         For a type 2 call, if c_strlen indicates a non-terminated string
5702         use the length of the non-terminated string.
5703         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
5705 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
5707         PR target/87467
5708         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
5709         __m512d type for __A argument rather than __m512.
5711 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
5713         * match.pd (simple_comparison): Don't optimize if either operand is
5714         a function pointer when target needs function pointer canonicalization.
5716 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5718         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
5719         power5 .. power9 to remove indirection.
5720         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
5721         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
5722         ASM_CPU_476_SPEC): Delete.
5723         (ASM_CPU_SPEC): Adjust.
5724         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
5725         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
5727 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5729         * config.in: Delete HAVE_AS_DCI.
5730         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
5731         * config/rs6000/rs6000.h: Ditto.
5732         * configure.ac: Delete HAVE_AS_DCI.
5733         * configure: Regenerate.
5735 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
5737         * config.in (HAVE_AS_LWSYNC): Delete.
5738         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5739         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
5740         do it as a .long .
5741         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
5742         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
5743         as a .long .
5744         * configure.ac: Delete HAVE_AS_LWSYNC.
5745         * configure: Regenerate.
5747 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5748             Pierre-Marie de Rodat  <derodat@adacore.com>
5750         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
5751         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
5752         (cgraph_node::create_thunk): Add indirect_offset parameter.
5753         (thunk_adjust): Likewise.
5754         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
5755         and initialize the corresponding field with it.
5756         (cgraph_node::dump): Dump indirect_offset field.
5757         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
5758         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
5759         (thunk_adjust): Add indirect_offset parameter and deal with it.
5760         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
5761         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
5762         or if the thunk is external or local.  Fix formatting.  Do not chain
5763         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
5764         if any, in the GIMPLE representation.
5765         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
5766         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
5767         (input_node): Read indirect_offset field.
5768         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
5769         call to thunk_adjust.
5770         * tree-nested.c (struct nesting_info): Add thunk_p field.
5771         (create_nesting_tree): Set it.
5772         (convert_all_function_calls): Copy static chain from targets to thunks.
5773         (finalize_nesting_tree_1): Return early for thunks.
5774         (unnest_nesting_tree_1): Do not finalize thunks.
5775         (gimplify_all_functions): Do not gimplify thunks.
5777 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5779         * opt-suggestions.c (option_proposer::build_option_suggestions):
5780         Release "option_values".
5782 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
5784         * coverage.c (get_coverage_counts): Convert problem-reporting dump
5785         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
5786         * dumpfile.c (kind_as_string): New function.
5787         (dump_loc): Rather than a hardcoded prefix of "note: ", use
5788         kind_as_string to vary the prefix based on dump_kind.
5789         (selftest::test_capture_of_dump_calls): Update for above.
5791 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5793         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
5794         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
5796 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
5798         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
5799         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
5800         INVALID_REGNUM instead of FPSR_REG.
5801         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
5802         * config/i386/i386.md: Update comment of FP compares.
5803         (fldenv): Do not clobber FPSR_REG.
5805 2018-09-28  Richard Biener  <rguenther@suse.de>
5807         * tree.h (BLOCK_ORIGIN): New.
5808         * omp-expand.c (grid_expand_target_grid_body): Assign
5809         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
5810         * tree-inline.c (remap_block): Likewise.
5811         * auto-profile.c (get_function_decl_from_block): Simplify
5812         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
5813         * langhooks.c (lhd_print_error_function): Likewise.
5814         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
5815         Likewise.
5816         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
5817         * tree.c (block_nonartificial_location): Likewise.
5818         (block_ultimate_origin): Likewise.
5819         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
5820         no longer needed LTO case.
5822 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
5823             Jan Hubicka  <jh@suse.cz>
5824             Martin Jambor  <mjambor@suse.cz>
5826         * simplify-rtx.c (simplify_merge_mask): New function.
5827         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
5828         same masks are used in op1 or op2.
5829         (test_vec_merge): New function.
5830         (test_vector_ops): Call test_vec_merge.
5832 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
5834         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
5835         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
5836         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
5837         (cypress_costs): Set it.
5838         (supersparc_costs): Likewise.
5839         (hypersparc_costs): Likewise.
5840         (leon_cost): Likewise.
5841         (leon3_costs): Likewise.
5842         (sparclet_costs): Likewise.
5843         (ultrasparc_costs): Likewise.
5844         (ultrasparc_costs): Likewise.
5845         (niagara_costs): Likewise.
5846         (niagara2_costs): Likewise.
5847         (niagara3_costs): Likewise.
5848         (niagara4_costs): Likewise.
5849         (niagara7_costs): Likewise.
5850         (m8_costs): Likewise.
5851         (TARGET_CAN_FOLLOW_JUMP): Define.
5852         (pass_work_around_errata::gate): Minor tweak.
5853         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
5854         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
5855         Automaitcally clear MASK_FSMULD mask for V7 processors.
5856         (sparc_can_follow_jump): New static function.
5857         (output_ubranch): Deal with CROSSING_JUMP_P.
5858         (sparc_use_sched_lookahead): Rewrite using switch statement.
5859         (sparc_issue_rate): Reorder.
5860         (sparc_branch_cost): New function.
5862 2018-09-27  Martin Sebor  <msebor@redhat.com>
5864         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
5865         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
5866         (int_fits_type_p): Same.
5868 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
5870         * config/i386/i386.md (FPCR_REG): Remove.
5871         (UNSPEC_FLDCW): Remove.
5872         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
5873         (x86_fldcw_1): Remove insn pattern.
5874         (fnstenv): Do not clobber FPCR_REG.
5875         (fldenv): Ditto.
5876         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
5877         (CALL_USED_REGISTERS): Ditto.
5878         (REG_ALLOC_ORDER): Ditto.
5879         (REG_CLASS_CONTENTS): Ditto.
5880         (HI_REGISTER_NAMES): Ditto.
5881         (ADDITIONAL_REGISTER_NAMES): Use defines instead
5882         of numerical constants.
5883         * config/i386/i386.c (regclass_map): Remove fpsr register.
5884         (dbx_register_map): Ditto.
5885         (dbx64_register_map): Ditto.
5886         (svr4_dbx_register_map): Ditto.
5887         (print_reg): Do not handle FPCR_REG.
5889 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
5891         PR target/87149
5892         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
5893         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
5894         Delete, always treat as true.
5895         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
5896         Ditto.  Simplify remaining code.
5897         * config/powerpcspe/powerpcspe.h: Ditto.
5898         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
5899         Simplify remaining code.
5900         (rs6000_expand_builtin): Ditto.
5901         * config/rs6000/rs6000.h: Ditto.
5902         * configure.ac: Ditto.
5903         * configure: Regenerate.
5905 2018-09-27  Martin Liska  <mliska@suse.cz>
5907         * coverage.c (get_coverage_counts): Revert the formatting
5908         of missing profile opt info.
5910 2018-09-27  Richard Biener  <rguenther@suse.de>
5912         PR debug/37801
5913         PR debug/87440
5914         * dwarf2out.c (set_block_origin_self): Do not mark outermost
5915         block as we do not output that.
5916         (gen_inlined_subroutine_die): Elide the originally outermost
5917         block, matching what we do for concrete instances.
5918         (decls_for_scope): Add parameter specifying whether to recurse
5919         to subblocks.
5921 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
5922             Tom de Vries  <tom@codesourcery.com>
5924         PR 82089
5926         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
5927         STORE_FLAG_VALUE == 1.
5929 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
5931         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
5932         constant definitions.
5933         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
5934         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
5935         ("speculation_barrier"): New expander definition.
5937 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
5939         PR gcov-profile/86957
5940         * common.opt: New warning option -Wmissing-profile.
5941         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
5942         * doc/invoke.texi: Document -Wmissing-profile.
5944 2018-09-26  Jim Wilson  <jimw@sifive.com>
5946         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
5947         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
5949 2018-09-26  Martin Sebor  <msebor@redhat.com>
5951         * tree.c (zerop): Change return type to bool.
5952         (integer_zerop, integer_onep, integer_each_onep): Same.
5953         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5954         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5955         (real_onep, real_minus_onep, chain_index): Same.
5956         (print_type_hash_statistics, type_list_equal): Same.
5957         * tree.h (zerop): Same.
5958         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
5959         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
5960         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
5961         (real_onep, real_minus_onep, chain_index): Same.
5962         (print_type_hash_statistics, type_list_equal): Same.
5964 2018-09-26  Jim Wilson  <jimw@sifive.com>
5966         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
5968 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
5970         PR target/87414
5971         * config/i386/i386.c: Include debug.h and dwarf2out.h.
5972         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
5973         call.
5975 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
5977         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
5979 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
5981         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
5982         and FP_SECOND_SSE_REGS.
5983         (REG_CLASS_NAMES): Ditto.
5984         (REG_CLASS_CONTENTS): Ditto.
5985         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
5986         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
5987         (ix86_preferred_output_reload_class): Ditto.
5988         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
5989         clobber constraint to "=&f".
5990         (fix_truncdi_i387): Ditto.
5991         (lrintxfdi2): Ditto.
5992         (fistdi2_<rounding>): Ditto.
5993         (fpremxf4_i387): Change "=u" constraint to "=f".
5994         (fprem1xf4_i387): Ditto.
5995         (sincosxf3): Ditto.
5996         (fptanxf4_i387): Ditto.
5997         (fxtractxf3_i387): Ditto.
5998         (fscalexf4_i387): Ditto.
5999         (atan2xf3): Change "u" constraint to "f".
6000         (fyl2xxf3_i387): Ditto.
6001         (fyl2xp1xf3_i387): Ditto.
6003 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
6005         PR target/87439
6006         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
6007         for removed I387_MASK_PM entity.
6010 2018-09-26  Jeff Law  <law@redhat.com>
6011         Revert
6012         2018-09-26  Alexey Neyman  <stilor@att.net>
6014         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
6015         headers are no longer pulled in by <isl/val.h>.
6017 2018-09-26  Richard Biener  <rguenther@suse.de>
6019         PR debug/87443
6020         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
6021         or concrete instance DIE to the tree.  Create abstract origin
6022         attributes also for concrete instances.
6024 2018-09-26  Alexey Neyman  <stilor@att.net>
6026         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
6027         headers are no longer pulled in by <isl/val.h>.
6029 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
6031         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
6032         Use new helper functions.
6033         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
6034         Use new helper functions.
6035         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
6036         aarch_mm_needs_release): New declarations.
6037         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
6038         aarch_mm_needs_release): New.
6040 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
6042         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
6043         (arm32_output_mi_thunk): Deal with long calls.
6045 2018-09-26  Richard Biener  <rguenther@suse.de>
6047         PR debug/87428
6048         PR debug/87362
6049         * tree-inline.c (expand_call_inline): When the location
6050         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
6051         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
6052         the inserted BLOCK to make inlined_function_outer_scope_p
6053         recognize it.
6054         * dwarf2out.c (add_call_src_coords_attributes): Do not add
6055         coords for reserved locations.
6057 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6059         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
6060         (*call_indirect_nonlocal_sysv<mode>): Ditto.
6061         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
6062         (*sibcall_nonlocal_sysv<mode>): Ditto.
6063         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
6064         (<bd>_<mode>): Ditto.
6065         (<bd>tf_<mode>): Ditto.
6067 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6069         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
6070         control string as a list of templates instead of as C code.
6071         (*altivec_movti): Ditto.
6072         * config/rs6000/darwin.md (movdf_low_di): Ditto.
6074 2018-09-25  Jim Wilson  <jimw@sifive.com>
6076         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
6077         when target symbol is weak.
6079 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6081         PR c/87387
6082         * builtins.c (unterminated_array): Simplify.
6083         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
6084         where pointer arithmetic is safe.
6086 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
6088         PR target/86987
6089         * config/rs6000/altivec.md (altivec_vspltb): Use
6090         const_0_to_15_operand instead of u5bit_cint_operand.
6091         (*altivec_vspltb_internal): Ditto.
6092         (altivec_vspltb_direct): Ditto.
6093         (altivec_vsplth): Use const_0_to_7_operand instead of
6094         u5bit_cint_operand.
6095         (*altivec_vsplth_internal): Ditto.
6096         (altivec_vsplth_direct): Ditto.
6097         (altivec_vspltw): Use const_0_to_3_operand instead of
6098         u5bit_cint_operand.
6099         (*altivec_vspltw_internal): Ditto.
6100         (altivec_vspltw_direct): Ditto.
6101         (altivec_vspltsf): Ditto.
6102         (*altivec_vspltsf_internal): Ditto.
6103         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
6104         various splats with the proper size immediate.  Reorder the various
6105         cases by ascending size of immediate, and put all such together.
6107 2018-09-25  Richard Biener  <rguenther@suse.de>
6109         PR debug/83941
6110         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
6111         GC-ification.
6112         (maybe_create_die_with_external_ref): Do not create
6113         DW_TAG_imported_unit here.
6114         (add_abstract_origin_attribute): Handle external BLOCK refs.
6115         (dwarf2out_abstract_function): Simplify LTO case.
6116         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
6117         rather than using maybe_create_die_with_external_ref.
6119 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
6121         PR target/71278
6122         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
6123         (frndintxf2_mask_pm_i387): Ditto.
6124         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
6125         Enable for !flag_trapping_math.
6126         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
6127         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
6128         Change operand 1 predicate to nonimmediate_operand.
6129         (attr "i387_cw"): Remove mask_pm.
6130         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
6131         (enum ix86_entity): Remove I387_MASK_PM.
6132         * config/i386/i386.c (ix86_i387_mode_needed): Do not
6133         handle I387_MASK_PM.
6134         (ix86_mode_needed): Ditto.
6135         (ix86_mode_after): Ditto.
6136         (ix86_mode_entry): Ditto.
6137         (ix86_mode_exit): Ditto.
6138         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
6140 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
6142         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
6143         to_update_switch_stmts to vNULL instead of calling create on them
6144         immediately.
6146 2018-09-25  Richard Biener  <rguenther@suse.de>
6148         PR tree-optimization/87402
6149         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
6150         (visit_phi): Re-instantiate handling of supposed to be VARYING
6151         but non-VARYING backedge value.
6153 2018-09-25  Richard Biener  <rguenther@suse.de>
6155         PR debug/83941
6156         * dwarf2out.c (struct sym_off_pair): New.
6157         (external_die_map): New global.
6158         (lookup_decl_die): When in LTO create DIEs lazily from the
6159         external_die_map.
6160         (lookup_block_die): New function, create DIEs lazily in LTO.
6161         (equate_block_to_die): New function.
6162         (dwarf2out_die_ref_for_decl): During WPA get the association
6163         from the external DIE map.
6164         (dwarf2out_register_external_die): Record mapping into the
6165         external DIE map.
6166         (maybe_create_die_with_external_ref): New function split out from
6167         DIE generation part of old dwarf2out_register_external_die.
6168         (add_abstract_origin_attribute): Do not return the DIE.  When
6169         in LTO reference externals directly.
6170         (dwarf2out_abstract_function): When in LTO ignore calls for
6171         decls with external DIEs (already present abstract instances).
6172         (gen_call_site_die): Adjust.
6173         (add_high_low_attributes): Likewise.
6174         (gen_lexical_block_die): Likewise.
6175         (gen_inlined_subroutine_die): Likewie.
6176         (gen_block_die): Likewise.
6177         (dwarf2out_inline_entry): Likewise.
6178         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
6179         DIEs.
6181 2018-09-25  Martin Liska  <mliska@suse.cz>
6183         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
6184         integers and not by a float value.
6186 2018-09-25  Martin Liska  <mliska@suse.cz>
6188         PR fortran/87394
6189         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
6190         instead of NULL.
6191         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
6192         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
6193         * godump.c (go_format_type): Remove extra parenthesis.
6195 2018-09-25  Martin Liska  <mliska@suse.cz>
6197         * alias.c (set_dest_equal_p): Remove unused function.
6198         * config/i386/i386.c (def_builtin_pure2): Likewise.
6199         * diagnostic-show-locus.c (class layout): Remove
6200         unused field.
6201         (layout::layout): Likewise here.
6202         * dump-context.h (class temp_dump_context): Likewise.
6203         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
6204         (add_AT_loclistsptr): Likewise.
6205         (add_AT_offset): Likewise.
6206         (get_AT_hi_pc): Likewise.
6207         (is_comdat_die): Likewise.
6208         (type_is_enum): Likewise.
6209         (ceiling): Likewise.
6210         (add_AT_vms_delta): Likewise.
6211         (is_class_die): Likewise.
6212         * edit-context.c (class line_event): Remove unused field.
6213         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
6214         unused function.
6215         * ipa-cp.c (ipa_get_vr_lat): Likewise.
6216         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
6217         (ok_for_base_p_nonstrict): Likewise.
6218         * tree-chrec.c (is_not_constant_evolution): Likewise.
6219         (chrec_fold_poly_cst): Likewise.
6220         * tree-if-conv.c (has_pred_critical_p): Likewise.
6221         * tree-ssa-coalesce.c (print_exprs): Likewise.
6222         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
6223         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
6224         * tree-vrp.c (value_ranges_intersect_p): Likewise.
6225         (value_range_nonnegative_p): Likewise.
6227 2018-09-25  Martin Liska  <mliska@suse.cz>
6229         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
6230         Do not handle "GNU Pascal".
6231         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
6232         Likewise.
6233         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
6234         from documentation. Likewise.
6235         * dbxout.c (dbxout_range_type): Likewise.
6236         * doc/cpp.texi: Likewise.
6237         * doc/extend.texi: Likewise.
6238         * doc/frontends.texi: Likewise.
6239         * doc/invoke.texi: Remove Pascal entry.
6240         * tree.def (CLEANUP_POINT_EXPR): Likewise.
6241         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
6243 2018-09-25  Martin Liska  <mliska@suse.cz>
6245         PR middle-end/86078
6246         * doc/invoke.texi: Document all parameters and remove default
6247         of the parameters.
6249 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
6251         PR bootstrap/87417
6252         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
6253         contains HOST_WIDE_INTs when computing its size.
6255 2018-09-24  Jim Wilson  <jimw@sifive.com>
6257         PR target/87391
6258         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
6259         not TARGET_RVE.
6260         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
6262 2018-09-24  Andrew Pinski  <apinski@marvell.com>
6264         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
6265         access prev before checking it for NULLness in the
6266         AARCH64_FUSE_CMP_BRANCH case.
6268 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
6270         PR target/82699
6271         * config/i386/i386.c (rest_of_insert_endbranch): Set
6272         endbr_queued_at_entrance to true and don't insert ENDBR if
6273         x86_function_profiler will be called.
6274         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
6275         is true.
6276         * config/i386/i386.h (machine_function): Add
6277         endbr_queued_at_entrance.
6279 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6281         * genattrtab.c (mk_attr_alt): Use alternative_mask.
6282         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
6283         types.
6284         (check_attr_test): Use alternative_mask.
6285         (get_attr_value): Likewise.
6286         (compute_alternative_mask): Use alternative_mask and XWINT.
6287         (make_alternative_compare): Use alternative_mask.
6288         (attr_alt_subset_p): Use XWINT.
6289         (attr_alt_subset_of_compl_p): Likewise.
6290         (attr_alt_intersection): Use alternative_mask and XWINT.
6291         (attr_alt_union): Likewise.
6292         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
6293         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
6294         (simplify_test_exp): Use alternative_mask and XWINT.
6295         (write_test_expr): Use alternative_mask and XWINT, adjust bit
6296         number calculation to support 64 bits.  Generate code that
6297         checks 64-bit masks.
6298         (main): Use alternative_mask.
6299         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
6301 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
6303         PR target/80080
6304         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
6305         RETURN+USE when returning via %r14.
6307 2018-09-24  Martin Liska  <mliska@suse.cz>
6309         * gcov.c (output_lines): Print colorization legend
6310         for both flag_use_colors and flag_use_hotness_colors.
6311         Reword the help.
6313 2018-09-24  Martin Liska  <mliska@suse.cz>
6315         * coverage.c (get_coverage_counts): Use warning_at
6316         with current_function_decl location. Use %qD in warning
6317         message.
6319 2018-09-24  Martin Liska  <mliska@suse.cz>
6321         * memory-block.h (memory_block_pool::release): Annotate with
6322         valgrind that the memory is not accessible.
6324 2018-09-24  Martin Liska  <mliska@suse.cz>
6326         PR sanitizer/85774
6327         * asan.c: Make asan_handled_variables extern.
6328         * asan.h: Likewise.
6329         * cfgexpand.c (expand_stack_vars): Make sure
6330         a representative is unpoison if another
6331         variable in the partition is handled by
6332         use-after-scope sanitization.
6334 2018-09-24  Richard Biener  <rguenther@suse.de>
6336         PR tree-optimization/63155
6337         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
6338         the worklist when the edge of the respective argument isn't
6339         executable.
6341 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6343         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
6344         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
6345         (MASK_CLASS_P): Update for rename.
6346         (MAYBE_MASK_CLASS_P): Ditto.
6347         (REG_CLASS_NAMES): Update.
6348         (REG_CLASS_CONTENT): Update.
6349         * config/i386/i386.c (regclass_map): Update for MASK_REG
6350         and ALL_MASK_REGS rename.
6351         * config/i386/constraints.md (Yk): Update for rename.
6352         (k): Ditto.
6354 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6356         * config/i386/i386.h (enum reg_class): Remove
6357         EVEX_SSE_REGS and MOD4_SSE_REGS.
6358         (REG_CLASS_NAMES): Update.
6359         (REG_CLASS_CONTENT): Update.
6360         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
6361         registers as ALL_SSE_REGS.
6362         (ix86_additional_allocno_class_p): Remove.
6363         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
6364         (ix86_register_priority): Lower priority of EVEX SSE registers.
6365         Use IN_RANGE macro where appropriate.
6366         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
6367         AVX-5124VNNIW checks.
6368         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
6369         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
6370         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
6371         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
6372         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
6373         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
6374         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
6375         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
6376         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
6377         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
6378         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
6379         * config/i386/constraints.md (Yh): Remove.
6381 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
6383         * config/i386/i386.c (regclass_map): Declare integer REX registers
6384         as GENERAL_REGS.
6386 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
6388         * doc/service.texi (Service): Switch the fsf.org link to https.
6390 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
6392         PR target/86798
6393         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
6394         Define to speculation_safe_value_not_needed.
6396 2018-09-21  Florian Weimer  <fweimer@redhat.com>
6398         PR middle-end/81035
6399         * doc/extend.texi (Common Function Attributes): Mention that
6400         noreturn suppresses tail call optimization.
6402 2018-09-21  Jeff Law  <law@redhat.com>
6404         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
6405         vr_values::cleanup_edges_and_switches.
6406         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
6407         vr_values class.
6408         (identify_jump_threads): Remove EDGE_IGNORE handling.
6409         (execute_vrp): Move handling of to_remove_edges and
6410         to_update_switch_stmts into vr_values class member functions.
6411         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
6412         (to_update_switch_stmts): Likewise.
6413         * vr-values.c: Include cfghooks.h.
6414         (vr_values::vr_values): Initialize to_remove_edges and
6415         to_update_switch_stmts.
6416         (vr_values::~vr_values): Verify to_remove_edges and
6417         to_update_switch_stmts are empty.
6418         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
6419         (vr_values::cleanup_edges_and_switches): New member function.
6420         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
6421         function.  Add new data members.
6423 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
6425         PR tree-optimization/87309
6426         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
6427         calls with pflags and alt_flags.
6428         (selftest::test_capture_of_dump_calls): Add test of interaction of
6429         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
6431 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6433         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
6435 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6437         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
6439 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6441         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
6442         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
6443         (VXWORKS_LIBS_RTP): Minor reordering.
6445 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6447         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
6448         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
6450 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6452         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
6453         (PTRDIFF_TYPE): Likewise.
6455 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6457         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
6458         triplet, similar to support for VxWorks7.
6459         * config/vxworks-dummy.h: Provide a default definition
6460         of TARGET_VXWORKS64 to 0.
6462 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6464         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
6465         * config/vxworks-dummy.h: here.
6467 2018-09-21  Olivier Hainque  <hainque@adacore.com>
6469         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
6471 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
6472             Bo Zhou  <zbo.zhou@hisilicon.com>
6474         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
6475         * config/aarch64/aarch64-tune.md: Regenerated.
6476         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
6477         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
6478         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
6480 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
6481             Julian Brown  <julian@codesourcery.com>
6483         * builtins.c (get_builtin_sync_mem): Handle address spaces.
6485 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6487         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
6488         if the call takes a static chain.
6490 2018-09-21  Martin Liska  <mliska@suse.cz>
6492         * auto-profile.c (autofdo_source_profile::read): Do not
6493         set sum_all.
6494         (read_profile): Do not add working sets.
6495         (read_autofdo_file): Remove sum_all.
6496         (afdo_callsite_hot_enough_for_early_inline): Remove const
6497         qualifier.
6498         * coverage.c (struct counts_entry): Remove gcov_summary.
6499         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
6500         do not support GCOV_TAG_PROGRAM_SUMMARY.
6501         (get_coverage_counts): Remove summary and expected
6502         arguments.
6503         * coverage.h (get_coverage_counts): Likewise.
6504         * doc/gcov-dump.texi: Remove -w option.
6505         * gcov-dump.c (dump_working_sets): Remove.
6506         (main): Do not support '-w' option.
6507         (print_usage): Likewise.
6508         (tag_summary): Likewise.
6509         * gcov-io.c (gcov_write_summary): Do not dump
6510         histogram.
6511         (gcov_read_summary): Likewise.
6512         (gcov_histo_index): Remove.
6513         (gcov_histogram_merge): Likewise.
6514         (compute_working_sets): Likewise.
6515         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
6516         it not obsolete.
6517         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
6518         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
6519         (GCOV_HISTOGRAM_SIZE): Remove.
6520         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
6521         (struct gcov_summary): Simplify rapidly just
6522         to runs and sum_max fields.
6523         (gcov_histo_index): Remove.
6524         (NUM_GCOV_WORKING_SETS): Likewise.
6525         (compute_working_sets): Likewise.
6526         * gcov-tool.c (print_overlap_usage_message): Remove
6527         trailing empty line.
6528         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
6529         (output_lines): Remove program related line.
6530         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
6531         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
6532         histogram.
6533         (input_profile_summary): Do not read it.
6534         (merge_profile_summaries): And do not merge it.
6535         (input_symtab): Do not call removed function.
6536         * modulo-sched.c (sms_schedule): Do not print sum_max.
6537         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
6538         removed when histogram method was invented.
6539         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
6540         mode.
6541         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
6542         GCOV coding style.
6543         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
6544         and dump selected value.
6545         * profile.c (add_working_set): Remove.
6546         (get_working_sets): Likewise.
6547         (find_working_set): Likewise.
6548         (get_exec_counts): Do not work with working sets.
6549         (read_profile_edge_counts): Do not inform as sum_max is removed.
6550         (compute_branch_probabilities): Likewise.
6551         (compute_value_histograms): Remove argument for call of
6552         get_coverage_counts.
6553         * profile.h: Do not make gcov_summary const.
6555 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
6557         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
6559 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
6561         PR tree-optimization/86990
6562         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
6563         Check that the entire merged store group is made of constants only for
6564         overlapping stores.
6566 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
6568         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
6569         (VTABLE_VERIFICATION_SPEC): Likewise.
6570         (SANITIZER_EARLY_SPEC): Likewise.
6571         (SANITIZER_SPEC): Likewise.
6572         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
6573         * doc/invoke.texi (Link Options): Document -r.
6575 2018-09-20  Richard Biener <rguenther@suse.de>
6577         PR middle-end/87054
6578         * gimplify.c (gimplify_expr): Retain alignment of
6579         addressable lvalue in dereference.
6581 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
6583         PR bootstrap/87013
6584         * configure.ac: Check for .loc is_stmt support.
6585         * configure, config.in: Rebuilt.
6586         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
6587         if not supported.
6589 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6591         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
6592         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
6593         -misel=no.
6595 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
6597         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
6598         VECTOR_OTHER.
6599         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
6600         case VECTOR_OTHER.
6602 2018-09-20  Marek Polacek  <polacek@redhat.com>
6604         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
6606 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6608         PR tree-optimization/87288
6609         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
6610         into account when determining PEELING_FOR_NITERS.
6612 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
6614         PR tree-optimization/86877
6615         * tree-vect-loop.c (vect_analyze_loop_2): Call
6616         vect_verify_datarefs_alignment.
6618 2018-09-19  Marek Polacek  <polacek@redhat.com>
6620         * doc/invoke.texi: Document -Wclass-conversion.
6622 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
6624         * config/pa/pa.c (pa_adjust_priority): Delete.
6625         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
6627         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
6628         (atomic_storehi): Likewise.
6629         (atomic_storesi): Likewise.
6630         (atomic_loaddi): Restore compare and swap exchange loop code.
6632 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
6634         PR rtl-optimization/86902
6635         * combine.c (try_combine): When changing the CC mode used, don't change
6636         an unrelated mode in other_insn to that new CC mode.
6638 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
6640         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
6641         with %T in place of calls to dump_generic_expr.
6642         (prune_runtime_alias_test_list): Likewise.
6643         (create_runtime_alias_checks): Likewise.
6644         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
6645         (vect_analyze_data_ref_dependence): Likewise.
6646         (vect_slp_analyze_data_ref_dependence): Likewise.
6647         (vect_record_base_alignment): Likewise.  Use %G in place of call
6648         to dump_gimple_stmt.
6649         (vect_compute_data_ref_alignment): Likewise.
6650         (verify_data_ref_alignment): Likewise.
6651         (vect_find_same_alignment_drs): Likewise.
6652         (vect_analyze_group_access_1): Likewise.
6653         (vect_analyze_data_ref_accesses): Likewise.
6654         (dependence_distance_ge_vf): Likewise.
6655         (dump_lower_bound): Likewise.
6656         (vect_prune_runtime_alias_test_list): Likewise.
6657         (vect_find_stmt_data_reference): Likewise.
6658         (vect_analyze_data_refs): Likewise.
6659         (vect_create_addr_base_for_vector_ref): Likewise.
6660         (vect_create_data_ref_ptr): Likewise.
6661         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
6662         (vect_can_advance_ivs_p): Likewise.
6663         (vect_update_ivs_after_vectorizer): Likewise.
6664         (vect_gen_prolog_loop_niters): Likewise.
6665         (vect_prepare_for_masked_peels): Likewise.
6666         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6667         (vect_determine_vectorization_factor): Likewise.
6668         (vect_is_simple_iv_evolution): Likewise.
6669         (vect_analyze_scalar_cycles_1): Likewise.
6670         (vect_analyze_loop_operations): Likewise.
6671         (report_vect_op): Likewise.
6672         (vect_is_slp_reduction): Likewise.
6673         (check_reduction_path): Likewise.
6674         (vect_is_simple_reduction): Likewise.
6675         (vect_create_epilog_for_reduction): Likewise.
6676         (vect_finalize_reduction:): Likewise.
6677         (vectorizable_induction): Likewise.
6678         (vect_transform_loop_stmt): Likewise.
6679         (vect_transform_loop): Likewise.
6680         (optimize_mask_stores): Likewise.
6681         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
6682         (vect_split_statement): Likewise.
6683         (vect_recog_over_widening_pattern): Likewise.
6684         (vect_recog_average_pattern): Likewise.
6685         (vect_determine_min_output_precision_1): Likewise.
6686         (vect_determine_precisions_from_range): Likewise.
6687         (vect_determine_precisions_from_users): Likewise.
6688         (vect_mark_pattern_stmts): Likewise.
6689         (vect_pattern_recog_1): Likewise.
6690         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6691         (vect_record_max_nunits): Likewise.
6692         (vect_build_slp_tree_1): Likewise.
6693         (vect_build_slp_tree_2): Likewise.
6694         (vect_print_slp_tree): Likewise.
6695         (vect_analyze_slp_instance): Likewise.
6696         (vect_detect_hybrid_slp_stmts): Likewise.
6697         (vect_detect_hybrid_slp_1): Likewise.
6698         (vect_slp_analyze_operations): Likewise.
6699         (vect_slp_analyze_bb_1): Likewise.
6700         (vect_transform_slp_perm_load): Likewise.
6701         (vect_schedule_slp_instance): Likewise.
6702         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6703         (vect_mark_stmts_to_be_vectorized): Likewise.
6704         (vect_init_vector_1): Likewise.
6705         (vect_get_vec_def_for_operand): Likewise.
6706         (vect_finish_stmt_generation_1): Likewise.
6707         (vect_check_load_store_mask): Likewise.
6708         (vectorizable_call): Likewise.
6709         (vectorizable_conversion): Likewise.
6710         (vectorizable_operation): Likewise.
6711         (vectorizable_load): Likewise.
6712         (vect_analyze_stmt): Likewise.
6713         (vect_is_simple_use): Likewise.
6714         (vect_get_vector_types_for_stmt): Likewise.
6715         (vect_get_mask_type_for_stmt): Likewise.
6716         * tree-vectorizer.c (increase_alignment): Likewise.
6718 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
6720         * doc/rtl.texi: Adjust vec_select description.
6721         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
6722         non-constant selectors.
6724 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
6726         * config/aarch64/aarch64-protos.h
6727         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
6728         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
6729         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
6730         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
6731         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
6732         (AARCH64_FL_PROFILE): Move index so flags are ordered.
6733         (AARCH64_ISA_RCPC8_4): New flag.
6734         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
6735         to aarch64_offset_9bit_signed_unscaled_p.
6736         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
6737         and use stlur.
6738         * config/aarch64/constraints.md (Ust): New constraint.
6739         * config/aarch64/predicates.md.
6740         (aarch64_9bit_offset_memory_operand): New predicate.
6741         (aarch64_rcpc_memory_operand): New predicate.
6743 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
6745         PR rtl-optimization/87361
6746         * rtlanal.c (nonzero_bits1): Revert accidental change.
6748 2018-09-19  Richard Biener  <rguenther@suse.de>
6750         PR tree-optimization/87349
6751         PR tree-optimization/87342
6752         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
6754 2018-09-18  Marek Polacek  <polacek@redhat.com>
6756         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
6757         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
6759 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6761         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
6763 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
6765         PR rtl-optimization/86882
6766         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
6768 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
6770         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
6771         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
6773 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
6775         PR other/87353
6776         * doc/invoke.texi (Link Options): Fix formatting and grammar.
6778 2018-09-18  Richard Biener  <rguenther@suse.de>
6780         PR middle-end/63155
6781         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
6782         (compute_samebase_partition_bases): Likewise.
6783         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
6784         (gimple_can_coalesce_p): Simplify.
6786 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
6788         Handle a library implementation of ffs calling __builtin_ffs.
6789         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
6790         (mmix_init_libfuncs): New function: make __builtin_ffs expand
6791         to __ffsdi2.
6793 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6795         * diagnostic-show-locus.c (class layout_range): Add field
6796         "m_original_idx".
6797         (layout_range::layout_range): Add "original_idx" param and use it
6798         to initialize new field.
6799         (make_range): Use 0 for original_idx.
6800         (layout::layout): Pass in index to calls to
6801         maybe_add_location_range.
6802         (layout::maybe_add_location_range): Add param "original_idx" and
6803         pass it on to layout_range.
6804         (layout::print_any_labels): Pass on range->m_original_idx to
6805         get_text call.
6806         (gcc_rich_location::add_location_if_nearby): Use 0 for
6807         original_idx.
6808         * gcc-rich-location.h (text_range_label::get_text): Update for new
6809         param.
6810         (range_label_for_type_mismatch::get_text): Likewise.
6812 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6814         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
6816 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
6818         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
6819         format_string_diagnostic_t.
6820         (fmtwarn_n): Likewise.
6821         * substring-locations.c
6822         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
6823         (format_warning_n_va): Convert to...
6824         (format_string_diagnostic_t::emit_warning_n_va): ...this.
6825         (format_warning_va): Convert to...
6826         (format_string_diagnostic_t::emit_warning_va): ...this.
6827         (format_warning_at_substring): Convert to...
6828         (format_string_diagnostic_t::emit_warning): ...this.
6829         (format_warning_at_substring_n): Convert to...
6830         (format_string_diagnostic_t::emit_warning_n): ...this.
6831         * substring-locations.h (class format_string_diagnostic_t): New
6832         class.
6833         (format_warning_va): Convert to
6834         format_string_diagnostic_t::emit_warning_va.
6835         (format_warning_n_va): Convert to
6836         format_string_diagnostic_t::emit_warning_n_va.
6837         (format_warning_at_substring): Convert to
6838         format_string_diagnostic_t::emit_warning.
6839         (format_warning_at_substring_n): Convert to
6840         format_string_diagnostic_t::emit_warning_n.
6842 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
6843             Bernd Schmidt <bernds_cb1@t-online.de>
6845         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
6846         SImode args.
6848 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
6850         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
6851         operand 0 predicate to nonimmediate operand.
6852         (rint<mode>2_frndint): Remove insn pattern.
6853         (rint<mode>2): Change operand 1 predicate to general_operand.
6854         Extend operand 1 to XFmode and generate rintxf2 insn.
6855         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
6856         Do not use X87MODEF mode macro.
6857         (frndintxf2_<rounding>_i387): Rename from
6858         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
6859         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
6860         to XFmode and generate significandxf3 insn.
6862 2018-09-17  Richard Biener  <rguenther@suse.de>
6864         PR tree-optimization/87328
6865         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
6866         visiting unexecutable backedges when not iterating.
6867         (do_rpo_vn): Mark all edges not executable even when not
6868         iterating.
6870 2018-09-17  Martin Jambor  <mjambor@suse.cz>
6872         PR c/63886
6873         * doc/invoke.texi (Warning Options): Likewise.
6875 2018-09-17  Richard Biener  <rguenther@suse.de>
6877         PR tree-optimization/87301
6878         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
6879         clean EH info from leftover copy assignments.
6881 2018-09-17  Martin Liska  <mliska@suse.cz>
6883         PR gcov-profile/85871
6884         * gcov.c (output_intermediate_file): Fix out of bounds
6885         access.
6887 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
6889         * config/arc/arc.c: Object attributes for core4 not reflected
6890         correctly.
6891         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
6892         core3).
6894 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
6896         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
6898 2018-09-17  Martin Liska  <mliska@suse.cz>
6900         * doc/gcov.texi: Document new option --use-hotness-colors.
6901         * gcov.c (struct source_info): Declare new field.
6902         (source_info::source_info): Set default for maximum_count.
6903         (print_usage): Add new -q option.
6904         (process_args): Process it.
6905         (accumulate_line_info): Save src->maximum_count.
6906         (output_line_beginning): Make color line number if
6907         flag_use_hotness_colors is set.
6908         (output_line_details): Pass default argument value.
6909         (output_lines): Pass src->maximum_count.
6911 2018-09-17  Martin Liska  <mliska@suse.cz>
6913         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
6914         Use processor_names table.
6915         * config/i386/i386.c (ix86_default_align): Use
6916         processor_cost_table for alignment values.
6917         (ix86_option_override_internal): Use processor_names.
6918         (ix86_function_specific_print): Likewise.
6919         * config/i386/i386.h (struct processor_costs):
6920         Add alignment values.
6921         (struct ptt): Remove and replace with const char *.
6922         * config/i386/x86-tune-costs.h (struct processor_costs):
6923         Declare default alignments for all costs.
6925 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
6927         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
6928         symbolics or VR_VARYING ranges for ABS_EXPR.
6929         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
6930         when range will wrap.
6932 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
6934         PR middle-end/86864
6935         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
6936         before and after a JUMP_TABLE_DATA.
6938 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
6940         PR middle-end/87188
6941         * dojump.c (do_compare_and_jump): Canonicalize function pointers
6942         when one operand is a function pointer.  Use POINTER_TYPE_P and
6943         FUNC_OR_METHOD_TYPE_P.
6944         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
6945         * fold-const.c (build_range_check): Likewise.
6946         * match.pd (simple_comparison): Likewise.
6948 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
6950         PR c/82967
6951         * spellcheck.c (get_edit_distance_cutoff): New function.
6952         (selftest::test_edit_distance_unit_test_oneway): Rename to...
6953         (selftest::test_get_edit_distance_one_way): ...this.
6954         (selftest::test_get_edit_distance_unit): Rename to...
6955         (selftest::test_get_edit_distance_both_ways): ...this.
6956         (selftest::test_edit_distances): Move tests to this new function,
6957         and test some more pairs of strings.  Update for above renaming.
6958         (selftest::get_old_cutoff): New function.
6959         (selftest::test_get_edit_distance_cutoff): New function.
6960         (selftest::assert_suggested_for): New function.
6961         (ASSERT_SUGGESTED_FOR): New macro.
6962         (selftest::assert_not_suggested_for): New function.
6963         (ASSERT_NOT_SUGGESTED_FOR): New macro.
6964         (selftest::test_suggestions): New function.
6965         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
6966         tests to selftest::test_edit_distances and call it.  Add calls to
6967         selftest::test_get_edit_distance_cutoff and
6968         selftest::test_suggestions.
6969         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
6970         (best_match::consider): Replace hard-coded cutoff calculation with
6971         a call to...
6972         (best_match::get_cutoff): New declaration.
6973         (best_match::get_best_meaningful_candidate): Likewise.
6975 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6977         * builtins.c (fold_builtin_strlen): Remove TODO comment.
6979 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6981         revert:
6982         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6984         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
6985         terminated string literal.
6987 2018-09-14  Martin Sebor  <msebor@redhat.com>
6989         * builtins.c (unterminated_array): Handle ARRAY_REF.
6990         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
6991         * builtins.h (unterminated_array): Declare extern.
6992         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
6993         arrays.
6994         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
6995         calls.
6997 2018-09-14  Martin Sebor  <msebor@redhat.com>
6998             Jeff Law  <law@redhat.com>
7000         * builtins.c (unterminated_array): New.
7001         (expand_builtin_strcpy): Adjust.
7002         (expand_builtin_strcpy_args): Detect unterminated arrays.
7003         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
7004         unterminated arrays.
7005         * gimple-fold.h (get_maxval_strlen): Add argument.
7006         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
7008         * gimple-fold.c (get_range_strlen): Add argument.
7009         (get_maxval_strlen): Adjust.
7010         * gimple-fold.h (get_range_strlen): Add argument.
7012 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
7014         * config/i386/movdirintrin.h: Fix copyright year.
7016 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
7018         * reg-stack.c: Include regs.h.
7019         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
7020         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
7021         FIRST_STACK_REG, not DFmode.
7022         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
7023         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
7024         (change stack): Default register mode to the reg_raw_mode of
7025         FIRST_STACK_REG, not DFmode.
7026         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
7027         (*swapxf): Rename from swapxf.
7029 2018-09-14  Carl Love  <cel@us.ibm.com>
7031         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
7032         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
7034 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
7036         PR target/87224
7037         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
7038         alternatives.
7040 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
7042         PR target/85628
7043         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
7045 2018-09-14  Jason Merrill  <jason@redhat.com>
7047         Fix --enable-gather-detailed-mem-stats.
7048         * hash-table.c (hash_table_usage): Change from variable to function.
7049         * hash-table.h: Adjust.
7050         * Makefile.in: Add missing dependencies on hash-table.h.
7052 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7054         PR tree-optimization/87259
7055         PR lto/87283
7056         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
7057         execute_cse_reciprocals_1 has tried transforming.
7059 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
7061         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
7062         VR_VARYING for PLUS/MINUS_EXPR.
7064 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
7066         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
7067         formatting.
7069 2018-09-14  Richard Biener  <rguenther@suse.de>
7071         PR middle-end/63155
7072         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
7073         bits for the merged partition.
7075 2018-09-13  Martin Sebor  <msebor@redhat.com>
7076             Bernd Edlinger  <bernd.edlinger@hotmail.de>
7078         * builtins.h (c_srlen): Add argument.
7079         * builtins.c (warn_string_no_nul): New function.
7080         (c_strlen): Add argument and use it.  Update recursive calls.
7081         Pass DECL argument to string_constant to get info on non
7082         terminated strings.  Update *NONSTR as needed.
7083         (fold_builtin_strlen): Add argument to calls to c_strlen.
7084         Warn for unterminated arrays.
7085         (warn_string_no_null): Add prototype.
7086         * expr.c (string_constant): Update arguments.  Update recursive
7087         calls appropriately.  Detect missing NUL terminator and outermost
7088         declaration its missing in.
7089         Improve checks for arrays with nonzero lower bound or elements
7090         that are not a single byte.  Simplify offset computation.
7091         Simplify checks for non-NUL terminated strings.
7092         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
7093         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
7095 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7097         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
7098         correctly.
7099         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
7100         argument.  Fix range checks.
7101         * fold-const.h (c_getstr): Adjust protoype.
7102         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
7103         string is constant but contains no NUL byte.
7105         * expr.c (string_constant): Adjust function comment.
7106         Remove bogus check for zero termination.
7108         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
7110         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
7111         (get_constant_size): Don't make STRING_CSTs larger than they are.
7112         (check_string_literal): New check function for STRING_CSTs.
7113         (output_constant): Use it.
7115 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
7117         PR target/86812
7118         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
7120 2018-09-13  Richard Biener  <rguenther@suse.de>
7122         PR tree-optimization/87263
7123         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
7124         (struct unwind_state): Add max_rpo field.
7125         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
7126         Compute max_rpo, the max RPO number a block can be backwards reached
7127         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
7128         separating it from the iterating mode.
7130 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
7132         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
7133         (rfs_decision): New scheduling decision.
7135 2018-09-13  Richard Biener  <rguenther@suse.de>
7137         PR bootstrap/87134
7138         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
7139         (vn_nary_op_insert_pieces_predicated): Do not write useless
7140         valid_dominated_by_p entry outside of the allocated storage.
7142 2018-09-13  Omar Sandoval  <osandov@osandov.com>
7143             Tom de Vries  <tdevries@suse.de>
7145         PR debug/86985
7146         * dwarf2out.c (is_c): New function.
7147         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
7149 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
7151         PR target/85628
7152         * config/aarch64/aarch64.md (*aarch64_bfxil):
7153         Define.
7154         * config/aarch64/constraints.md (Ulc): Define.
7155         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
7156         Define.
7157         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
7158         New function.
7160 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
7162         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
7163         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
7164         aarch64_layout_frame call.
7165         (aarch64_expand_epilogue): Likewise.
7166         (aarch64_initial_elimination_offset): Likewise.
7167         (aarch64_get_separate_components): Likewise.
7168         (aarch64_use_return_insn_p): Likewise.
7169         (aarch64_layout_frame): Remove unneeded check.
7171 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7173         * configure.ac: Only append
7174         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
7175         gcc_config_arguments if it was never reconfigured or last reconfigure
7176         was with different arguments.
7177         * configure: Regenerated.
7179 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7180             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7182         PR middle-end/87290
7183         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
7184         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
7186 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
7188         PR tree-optimization/87287
7189         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
7190         X % C == 0 to X % (unsigned) C == 0 optimization to ...
7191         * match.pd (X % C == 0): ... here.  New optimization.
7193 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7195         PR middle-end/82853
7196         * expr.h (maybe_optimize_mod_cmp): Declare.
7197         * expr.c (mod_inv): New function.
7198         (maybe_optimize_mod_cmp): New function.
7199         (do_store_flag): Use it.
7200         * cfgexpand.c (expand_gimple_cond): Likewise.
7202 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
7203             Julian Brown  <julian@codesourcery.com>
7205         PR middle-end/86336
7206         * gimplify.c (gimplify_scan_omp_clauses): Set
7207         target_firstprivatize_array_bases in OpenACC parallel and kernels
7208         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
7209         OpenACC data regions.
7211 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
7213         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
7214         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
7216 2018-09-12  Richard Biener  <rguenther@suse.de>
7218         PR tree-optimization/87280
7219         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
7220         edge but unreachable target.
7221         (do_rpo_vn): For conservatively handling a PHI only mark
7222         the backedge executable but not the block reachable.
7224 2018-09-12  Richard Biener  <rguenther@suse.de>
7226         PR tree-optimization/87266
7227         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
7228         visited blocks.
7230 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
7232         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
7233         constants.
7234         ("trunc<BFP:mode><DFP_ALL:mode>2")
7235         ("trunc<DFP_ALL:mode><BFP:mode>2")
7236         ("extend<BFP:mode><DFP_ALL:mode>2")
7237         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
7238         according to the target operand type.
7240 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7241             Andreas Krebbel  <krebbel@linux.ibm.com>
7243         PR tree-optimization/86844
7244         * gimple-ssa-store-merging.c
7245         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
7246         there are any overlapping stores in between them, make sure they are
7247         also coalesced or we give up completely.
7249 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
7251         PR middle-end/87248
7252         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
7253         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
7255 2018-09-12  Tom de Vries  <tdevries@suse.de>
7257         * common.opt (gdescribe-dies): Add option.
7258         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
7259         attribute for artifical and nameless decls.
7260         (dwarf2out_register_external_die): Add description attribute to
7261         external reference die.
7262         (add_desc_attribute): New functions.
7263         (gen_subprogram_die): Add description attribute to
7264         DW_TAG_call_site_parameter.
7265         * tree-pretty-print.c (print_generic_expr_to_str): New function.
7266         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
7267         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
7268         -gno-describe-dies.
7269         (@item -gdescribe-dies): Add.
7271 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
7273         * tree-vrp.c (vrp_shift_undefined_p): Remove.
7274         (extract_range_from_binary_expr_1: Call
7275         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
7276         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
7277         depend on sign.
7279 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
7281         * gimple-ssa-warn-alloca.c
7282         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
7283         field for ALLOCA_BOUND_*_LARGE.
7285 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7287         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
7289 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7291         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
7292         for clobbers.  Remove obsolete comment.
7294 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7296         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
7297         mpxchk, mpxld and mpxst types.
7298         (define_attr length_immediate): Remove all processing of mpx types.
7299         (define_attr prefix_0f): Ditto.
7300         (define_attr memory): Ditto.
7302 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
7304         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
7305         (log<mode>2): Change operand 1 predicate to general_operand.
7306         Extend operand 1 to XFmode and generate logxf3 insn.
7307         (log10<mode>2): Change operand 1 predicate to general_operand.
7308         Extend operand 1 to XFmode and generate log10xf3 insn.
7309         (log2<mode>2): Change operand 1 predicate to general_operand.
7310         Extend operand 1 to XFmode and generate log2xf3 insn.
7311         (fyl2xp1_extend<mode>xf3_i387): Remove.
7312         (log1p<mode>2): Change operand 1 predicate to general_operand.
7313         Extend operand 1 to XFmode and generate log1pxf3 insn.
7314         (fxtract_extend<mode>xf3_i387): Remove.
7315         (logb<mode>2): Change operand 1 predicate to general_operand.
7316         Extend operand 1 to XFmode and generate logbxf3 insn.
7317         (ilogb<mode>2): Change operand 1 predicate to general_operand.
7318         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
7319         (significand<mode>2): Change operand 1 predicate to general_operand.
7320         Extend operand 1 to XFmode and generate significandxf3 insn.
7322 2018-09-11  Nathan Sidwell  <nathan@acm.org>
7324         * gcc.c (perror_with_name, pfatal_with_name): Delete.
7325         (load_specs): Use fatal_error.
7326         (DELETE_IF_ORDINARY, process_command): Use error.
7327         (execute, run_attempt): Use fatal_error.
7329 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
7331         * diagnostic-core.h (sorry_at): New prototype.
7332         * diagnostic.c (sorry_at): New function.
7334 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
7336         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
7337         by zero as VR_UNDEFINED.
7339 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
7341         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
7342         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
7343         (<sincos>mode2): New expander.
7344         (sincos_extend<mode>xf3_i387): Remove insn pattern.
7345         (sincos -> sin, cos splitters): Remove splitter patterns.
7346         (sincos<mode>3): Change operand 2 predicate to general_operand.
7347         Extend operand 2 to XFmode and generate sincosxf3 insn.
7348         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
7349         Change operand 3 predicate to const1_operand.
7350         (fptan_extend<mode>xf4_i387): Remove insn pattern.
7351         (tanxf2): Update operands in the call to fptanxf4_i387.
7352         (tan<mode>2): Change operand 1 predicate to general_operand.
7353         Extend operand 1 to XFmode and generate tanxf3 insn.
7354         (atan2xf3): Rename from *fpatanxf3_i387.
7355         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
7356         (atan2xf3): Remove expander.
7357         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
7358         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
7359         (atan<mode>2): Change operand 1 predicate to general_operand.
7360         Extend operand 1 to XFmode and generate atanxf3 insn.
7362 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
7364         * config/i386/i386.md (x87/SSE constant load splitter): Use
7365         memory_operand instead of nonimmediate_operand for input operand
7366         predicate.
7368 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
7370         * config/i386/i386.md (float partial SSE register stall splitter): Move
7371         splitter near its instruction pattern.
7372         (float_extend partial SSE register stall splitter): Ditto.
7373         (float_truncate partial SSE register stall splitter): Ditto.
7375 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
7377         PR target/86794
7378         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7379         to speculation_safe_value_not_needed.
7381         PR target/85666
7382         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
7383         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
7384         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
7385         leaf_function_p, instead use has_hard_reg_initial_val.
7387 2018-09-09  Nathan Sidwell  <nathan@acm.org>
7389         * gcc.h (pfatal_with_name): Don't declare here.
7390         * gcc.c (pfatal_with_name): Make static.
7392 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
7394         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
7395         earlyclobber.
7397 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
7399         PR rtl-optimization/85458
7400         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
7401         priority hook to reduce the priority of EXPR.
7403 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
7405         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
7406         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
7407         with X87_ENABLE_FLOAT.  Remove preparation code.
7408         (*float<SWI48:mode><MODEF:mode>2): Rename from
7409         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
7410         math using "enabled" attribute.
7411         (*floatdi<MODEF:mode>2_i387): Rename from
7412         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
7413         enable for 32bit targets only.
7414         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
7415         splitter.
7416         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
7417         as operand 1 predicate.  Rewrite as define_insn_and_split.
7418         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
7420 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7422         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
7423         to fallthru to FLOAT case.
7425 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7427         PR target/86731
7428         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
7429         around folding of vec_sl to handle out of range shift values.
7431 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
7433         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
7434         Update callers to gen_fix_trunc<mode>_i387_fisttp
7435         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
7436         nonimmediate_operand.
7437         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
7438         and corresponding splitters.
7439         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
7440         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
7441         (fix_truncdi_i387_with_temp): Remove insn pattern
7442         and corresponding splitters.
7443         (fix_trunc<mode>_i387): Change operand 0 predicate to
7444         nonimmediate_operand.
7445         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
7446         and corresponding splitters.
7447         (*fistdi2_1): Remove.
7448         (fistdi2): Ditto.
7449         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
7450         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
7451         (*fist<mode>2_1): Remove.
7452         (fist<mode>2): Ditto.
7453         (fist<mode>2_with_temp): Remove insn pattern and corresponding
7454         splitters.
7455         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
7456         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
7457         (fistdi2_<rounding>): Change operand 0 predicate to
7458         nonimmediate_operand.
7459         (fistdi2_<rounding>_with_temp): Remove insn pattern
7460         and corresponding splitters.
7461         (fist<mode>2_<rounding>): Change operand 0 predicate to
7462         nonimmediate_operand.
7463         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
7464         and corresponding splitters.
7466         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
7468 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7470         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
7471         the init value.
7473 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7475         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7476         early gimple folding of vec_splat().
7477         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
7478         * gimple-fold.h: Add an extern define for tree_vec_extract().
7480 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
7482         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
7483         wrappers around TREE_TYPE comparisons.
7485 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7487         PR target/80080
7488         * config/s390/predicates.md: Add nonsym_memory_operand.
7489         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
7490         contains a SYMBOL_REF, load it into an intermediate pseudo.
7491         (s390_emit_compare_and_swap): Legitimize operand.
7492         * config/s390/s390.md: Use the new nonsym_memory_operand
7493         with UNSPECV_CAS patterns.
7495 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7497         PR target/80080
7498         * config/s390/s390-passes.def: New file.
7499         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
7500         declaration.
7501         (make_pass_s390_early_mach): Add declaration.
7502         * config/s390/s390.c (make_pass_s390_early_mach):
7503         (s390_option_override): Remove dynamic registration.
7504         * config/s390/t-s390: Add s390-passes.def.
7506 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
7508         * config/s390/s390.c (s390_decompose_constant_pool_ref):
7509         Remove UNSPEC_LTREL_BASE check.
7510         (annotate_constant_pool_refs): Likewise.
7511         (find_constant_pool_ref): Likewise.
7512         (find_ltrel_base): Removed.
7513         (replace_ltrel_base): Removed.
7514         (s390_mainpool_finish): Remove replace_ltrel_base call.
7515         (s390_chunkify_start): Remove pending LTREL_BASE logic.
7516         (s390_chunkify_finish): Remove replace_ltrel_base call.
7517         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
7519 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
7521         PR target/86779
7522         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
7523         to speculation_safe_value_not_needed.
7525 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
7526             Bernd Schmidt  <bernds_cb1@t-online.de>
7528         * config/nvptx/nvptx-opts.h: New file.
7529         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
7530         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
7531         (ASM_SPEC): Define.
7532         (TARGET_SM35): New macro.
7533         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
7534         correct predicate.
7535         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
7536         values.
7537         (misa=): New option.
7538         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
7540 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
7542         * config/i386/i386.md (truncdfsf2): Remove expander.
7543         (truncdfsf2_with_temp): Ditto.
7544         (truncxf<mode>2): Ditto.
7545         (*truncdfsf_fast_mixed): Remove insn pattern.
7546         (*truncdfsf_fast_i387): Ditto.
7547         (*truncdfsf_mixed): Ditto.
7548         (*truncdfsf_i387): Ditto.
7549         (*truncdfsf2_i387_1): Ditto.
7550         (*truncxfsf2_mixed): Ditto.
7551         (*truncxfdf2_mixed): Ditto.
7552         (*truncxf<mode>2_i387_noop): Ditto. Update callers
7553         to call gen_truncxf<mode>2 instead.
7554         (*truncxf<mode>2_i387): Remove.
7555         (reg->reg splitters): Remove splitter pattern.
7556         (reg->mem splitters): Ditto.
7558         (truncdfsf2): New insn pattern.
7559         (truncxf<mode>2): Ditto.
7561 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7563         * tree-ssa-math-opts.c (is_mult_by): New function.
7564         (is_square_of): Use the above.
7565         (optimize_recip_sqrt): New function.
7566         (pass_cse_reciprocals::execute): Use the above.
7568 2018-09-05  Richard Biener  <rguenther@suse.de>
7570         PR bootstrap/87134
7571         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
7572         to zero-init the emplaced vec.
7574 2018-09-05  Martin Liska  <mliska@suse.cz>
7576         PR tree-optimization/87205
7577         * tree-switch-conversion.c (pass_lower_switch::execute):
7578         Group cases for switch statements.
7580 2018-09-05  Richard Biener  <rguenther@suse.de>
7582         PR tree-optimization/87217
7583         * tree-ssa-sccvn.c (vuse_valueize): New.
7584         (vn_reference_lookup_pieces): Use it.
7585         (vn_reference_lookup): Likewise.
7587 2018-09-05  Nathan Sidwell  <nathan@acm.org>
7589         PR c++/87137
7590         * stor-layout.c (place_field): Scan forwards to check last
7591         bitfield when ms_bitfield_placement is in effect.
7593 2018-09-05  Richard Biener  <rguenther@suse.de>
7595         PR bootstrap/87225
7596         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
7597         return.
7599 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
7600             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
7602         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
7603         * config.gcc (extra_objs): Build it.
7604         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
7605         Likewise.
7606         * config/aarch64/aarch64-passes.def
7607         (pass_tag_collision_avoidance): New pass.
7608         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
7609         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
7610         (aarch64_classify_address): Remove static qualifier.
7611         (aarch64_address_info, aarch64_address_type): Move to...
7612         * config/aarch64/aarch64-protos.h: ... here.
7613         (make_pass_tag_collision_avoidance): New function.
7614         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
7615         New tuning flag.
7617 2018-09-05  Martin Liska  <mliska@suse.cz>
7619         * doc/gcov.texi: Update documentation of humar
7620         readable mode.
7621         * gcov.c (format_count): Print one decimal place, it provides
7622         more fine number of situations like '1G' vs. '1.4G'.
7624 2018-09-05  Martin Liska  <mliska@suse.cz>
7626         PR target/87164
7627         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
7628         * optc-gen.awk: Allow 'Var' for Deprecated options in order
7629         to generate a MASK value.
7631 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
7633         PR debug/86593
7634         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
7635         if frame pointer isn't used.
7636         (compute_frame_pointer_to_fb_displacement): Likewise.
7638 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
7640         PR target/87198
7641         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
7642         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
7643         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
7644         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
7645         and OPTION_MASK_ISA_XSAVEC_UNSET.
7647 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
7649         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
7650         XOR operations in NAND case.
7652 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7654         * wide-int-range.cc (wide_int_range_convert): New.
7655         * wide-int-range.h (wide_int_range_convert): New.
7656         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
7657         code into wide_int_range_convert.
7658         (extract_range_into_wide_ints): Do not munge anti range constants
7659         into the entire domain.  Just return the range back.
7661 2018-09-04  Martin Liska  <mliska@suse.cz>
7663         * genmatch.c (output_line_directive): Add new argument
7664         fnargs.
7665         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
7667 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7669         * doc/invoke.texi (Option Summary): Add whitespace.
7671         * doc/invoke.texi (Option Summary): Add -Waligned-new.
7673 2018-09-04  Richard Biener  <rguenther@suse.de>
7675         PR tree-optimization/87211
7676         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
7677         backedge value we're supposed to treat as VARYING also number
7678         the PHI to VARYING in case it got a different value-number already.
7680 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
7682         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
7683         (extract_range_from_binary_expr_1): Do not call
7684         vrp_can_optimize_bit_op.
7685         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
7686         static.
7687         (wide_int_range_get_mask_and_bounds): New.
7688         (wide_int_range_optimize_bit_op): New.
7689         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
7690         (wide_int_range_bit_and): Same.
7691         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
7692         (wide_int_range_optimize_bit_op): New.
7693         (wide_int_range_get_mask_and_bounds): New.
7695 2018-09-04  Richard Biener  <rguenther@suse.de>
7697         PR tree-optimization/87176
7698         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
7699         variable.  When value-numbering a virtual PHI node make sure
7700         to not value-number to the backedge value.
7702 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
7704         * doc/extend.texi (Long Long, Hex Floats): Document support for
7705         long long and hex floats in more recent versions of ISO C++.
7707 2018-09-03  Richard Biener  <rguenther@suse.de>
7709         PR tree-optimization/87177
7710         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
7711         cleanup.
7713 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7715         * bb-reorder.c (edge_order): Convert to C-qsort-style
7716         tri-state comparator.
7717         (reorder_basic_blocks_simple): Change std::stable_sort to
7718         gcc_stablesort.
7720 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7722         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
7723         tri-state comparator.
7724         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
7726 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7728         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
7729         (mergesort): ... here as maximum count for using netsort.
7730         (gcc_qsort): Set nlim to 3 if stable sort is requested.
7731         (gcc_stablesort): New.
7732         * system.h (gcc_stablesort): Declare.
7734 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
7736         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
7737         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
7738         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
7740 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
7742         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
7743         lxsdx and stxsdx alternatives.
7744         (*mov<mode>_hardfloat64): Ditto.
7745         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
7747 2018-09-03  Richard Biener  <rguenther@suse.de>
7749         PR tree-optimization/87200
7750         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
7751         simplify result.
7753 2018-09-03  Martin Liska  <mliska@suse.cz>
7755         PR tree-optimization/87201
7756         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7757         Fix parenthesis in an expression.
7759 2018-09-03  Richard Biener  <rguenther@suse.de>
7761         PR tree-optimization/87197
7762         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
7763         visited.  CSE the VN_INFO hashtable lookup.
7765         PR tree-optimization/87169
7766         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
7767         iterating make sure there's no extra backedges from irreducible
7768         regions feeding the header.  Mark the destination block
7769         executable.
7771 2018-09-03  Martin Liska  <mliska@suse.cz>
7773         PR driver/83193
7774         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
7775         * common/common-targhooks.c (default_get_valid_option_values):
7776         New function.
7777         * common/common-targhooks.h (default_get_valid_option_values):
7778         Likewise.
7779         * common/config/i386/i386-common.c: Move processor_target_table
7780         from i386.c.
7781         (ix86_get_valid_option_values): New function.
7782         (TARGET_GET_VALID_OPTION_VALUES): New macro.
7783         * config/i386/i386.c (struct ptt): Move to i386-common.c.
7784         (PTA_*): Move all defined masks into i386-common.c.
7785         (ix86_function_specific_restore): Use new processor_cost_table.
7786         * config/i386/i386.h (struct ptt): Moved from i386.c.
7787         (struct pta): Likewise.
7788         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
7789         * doc/tm.texi.in: Likewise.
7790         * opt-suggestions.c (option_proposer::suggest_option):
7791         Pass prefix to build_option_suggestions.
7792         (option_proposer::get_completions): Likewise.
7793         (option_proposer::build_option_suggestions): Use the new target
7794         hook.
7795         * opts.c (struct option_help_tuple): New struct.
7796         (print_filtered_help): Use the new target hook.
7798 2018-09-03  Martin Liska  <mliska@suse.cz>
7800         PR middle-end/59521
7801         * predict.c (set_even_probabilities): Add likely_edges
7802         argument and handle cases where we have precisely one
7803         likely edge.
7804         (combine_predictions_for_bb): Catch also likely_edges.
7805         (tree_predict_by_opcode): Handle gswitch statements.
7806         * tree-cfg.h (find_case_label_for_value): New declaration.
7807         (find_taken_edge_switch_expr): Likewise.
7808         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
7809         Find pivot in decision tree based on probabily, not by number of
7810         nodes.
7812 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
7814         * doc/standards.texi (Standards): Update Objective-C reference.
7816 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7818         * doc/install.texi (Prerequisites): Update link for MPC.
7820 2018-09-01  Michael Matz  <matz@suse.de>
7822         PR tree-optimization/87074
7823         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
7824         PHIs for outer-loop uses.
7826 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7828         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
7829         * doc/invoke.texi (C Dialect Options): Ditto.
7831 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
7833         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
7835 2018-08-31  Richard Biener  <rguenther@suse.de>
7837         PR tree-optimization/87168
7838         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
7839         (rpo_elim::eliminate_avail): When OP was not visited it must
7840         be available.
7842 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
7844         * tree-vrp.c (copy_value_range): Convert param "from" from
7845         "value_range *" to "const value_range *".
7846         (range_is_null): Likewise for param "vr".
7847         (range_int_cst_p): Likewise.
7848         (range_int_cst_singleton_p): Likewise.
7849         (symbolic_range_p): Likewise.
7850         (value_ranges_intersect_p): Likewise for both params.
7851         (value_range_nonnegative_p): Likewise for param "vr".
7852         (value_range_constant_singleton): Likewise.
7853         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
7854         (extract_range_into_wide_ints): Likewise for param "vr".
7855         (extract_range_from_multiplicative_op): Likewise for params "vr0"
7856         and "vr1".
7857         (vrp_can_optimize_bit_op): Likewise.
7858         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
7859         "vr1_".
7860         (extract_range_from_unary_expr): Likewise.
7861         (debug_value_range): Likewise for param "vr".
7862         (value_range::dump): Add "const" qualifier.
7863         (vrp_prop::check_array_ref): Convert local "vr" from
7864         "value_range *" to "const value_range *".
7865         (vrp_prop::check_mem_ref): Likewise.
7866         (vrp_prop::visit_stmt): Likewise for local "old_vr".
7867         (vrp_intersect_ranges_1): Likewise for param "vr_1".
7868         (vrp_intersect_ranges): Likewise.
7869         (simplify_stmt_for_jump_threading): Likewise for local "vr".
7870         (vrp_prop::vrp_finalize): Likewise.
7871         * tree-vrp.h (value_range::dump): Add "const" qualifier.
7872         (vrp_intersect_ranges): Add "const" qualifier to params as above.
7873         (extract_range_from_unary_expr): Likewise.
7874         (value_range_constant_singleton): Likewise.
7875         (symbolic_range_p): Likewise.
7876         (copy_value_range): Likewise.
7877         (extract_range_from_binary_expr_1): Likewise.
7878         (range_int_cst_p): Likewise.
7879         (vrp_set_zero_nonzero_bits): Likewise.
7880         (range_int_cst_singleton_p): Likewise.
7882 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
7884         * config/aarch64/arm_neon.h (vabsd_s64): New.
7885         (vnegd_s64): Likewise.
7887 2018-08-31  Martin Jambor  <mjambor@suse.cz>
7889         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
7891 2018-08-31  Martin Liska  <mliska@suse.cz>
7893         * ipa-icf.c (sem_item::add_type): Use
7894         sem_item::m_type_hash_cache.
7895         * ipa-icf.h: Move the cache from sem_item_optimizer
7896         to sem_item.
7898 2018-08-31  Nathan Sidwell  <nathan@acm.org>
7900         * doc/extend.texi (Backwards Compatibility): Remove implicit
7901         extern C leeway of () being (...).
7903 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
7905         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
7907 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
7909         PR target/86684
7910         PR target/87149
7911         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
7913 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
7915         PR middle-end/87138
7916         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
7917         gen_int_mode.  Formatting fixes.
7919 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
7921         * target.def (custom_function_descriptors): Improve documentation.
7922         * doc/tm.texi.in (Trampolines): Expand discussion of function
7923         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
7924         beginning of the section.
7925         * doc/tm.texi: Regenerated.
7927 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
7929         * cfg.h (class auto_edge_flag): Spell out the template-id of the
7930         base class in the initializer list.  This is a workaround for
7931         building with older GCC.
7932         (class auto_bb_flag): Likewise.
7934 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
7936         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
7937         (altivec_vcmpequ<VI_char>_p): Remove star.
7938         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
7939         vector load modes.
7940         (expand_strncmp_vec_sequence): New function.
7941         (emit_final_str_compare_vec): New function.
7942         (expand_strn_compare): Add support for vector strncmp.
7943         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
7944         length specification to bytes.
7945         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
7946         (vcmpnezb_p): New pattern.
7947         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
7948         for option -mstring-compare-inline-limit.
7950 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
7952         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
7953         (PTA_SKYLAKE): Add PTA_AES.
7954         (PTA_GOLDMONT): Likewise.
7956 2018-08-29  Jan Hubicka  <jh@suse.cz>
7958         PR lto/86517
7959         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
7960         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
7962 2018-08-29  Jan Hubicka  <jh@suse.cz>
7964         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
7965         TYPE_STUB_DECL.
7966         (hash_tree): Do not visit TYPE_STUB_DECL.
7967         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
7968         stream TYPE_STUB_DECL.
7969         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
7970         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
7971         after free_lang_data.
7972         (type_in_anonymous_namespace_p): Likewise.
7974 2018-08-29  Jan Hubicka  <jh@suse.cz>
7976         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
7977         comment that it has to be even number.
7978         (class sreal): Change m_sig type to int32_t.
7979         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
7980         int64_t for temporary calculations.
7981         (sreal_verify_basics): Drop one bit from minimum and maximum.
7983 2018-08-30  Richard Biener  <rguenther@suse.de>
7985         PR tree-optimization/87147
7986         * tree-ssa-sccvn.c (SSA_VISITED): New function.
7987         (visit_phi): When the degenerate result is from the backedge and
7988         we didn't visit its definition yet drop to VARYING.
7989         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
7991 2018-08-29  Jan Hubicka  <jh@suse.cz>
7993         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
7994         DECL_VINDEX.
7995         (hash_tree): Likewise.
7997 2018-08-29  Jan Hubicka  <jh@suse.cz>
7999         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
8000         and TYPE_NEXT_REF_TO.
8002 2018-08-29  Jan Hubicka  <jh@suse.cz>
8004         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
8005         comment that it has to be even number.
8006         (class sreal): Change m_sig type to int32_t.
8007         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
8008         int64_t for temporary calculations.
8009         (sreal_verify_basics): Drop one bit from minimum and maximum.
8011 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
8013         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
8015 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
8017         PR middle-end/86995
8018         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
8019         if to_add is negative.
8021 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8023         PR middle-end/87053
8024         * builtins.c (c_strlen): Improve range checks.
8026 2018-08-29  Martin Sebor  <msebor@redhat.com>
8027             Jeff Law  <law@redhat.com>
8029         PR tree-optimization/86714
8030         PR tree-optimization/86711
8031         * builtins.c (c_strlen): Add arguments to call to string_constant.
8032         * expr.c (string_constant): Add argument.  Detect missing nul
8033         terminator and outermost declaration it's missing in.
8034         * expr.h (string_constant): Add argument.
8035         * fold-const.c (read_from_constant_string): Add arguments to call to
8036         string_constant.
8037         (c_getstr): Likewise.
8038         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
8039         to string_constant.
8040         * tree-ssa-strlen.c (get_stridx): Likewise.
8042 2018-08-29  Jan Hubicka  <jh@suse.cz>
8044         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
8045         Do not stream DECL_VINDEX.
8046         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
8047         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
8048         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
8050 2018-08-29  Richard Biener  <rguenther@suse.de>
8052         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
8053         virtual operands that are not default defs to honor region
8054         boundaries.
8055         (rpo_vn_valueize): Remove ineffective code here.
8057 2018-08-29  Richard Biener  <rguenther@suse.de>
8059         PR tree-optimization/87132
8060         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
8061         when skipping defs reachable over backedges.
8063 2018-08-29  Richard Biener  <rguenther@suse.de>
8065         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
8066         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
8067         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
8068         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
8069         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
8070         (vn_reference_lookup_3): Remove use of const_parms.
8071         (free_rpo_vn): Do not free const_parms.
8072         (do_rpo_vn): Do not call init_const_parms.
8073         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
8074         SSA_NAME_POINTS_TO_READONLY_MEMORY.
8075         (call_may_clobber_ref_p_1): Likewise.
8077 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
8079         PR other/86726
8080         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
8081         (-O): Ditto.
8082         (-ftree-scev-cprop): Document.
8084 2018-08-29  Jan Hubicka  <jh@suse.cz>
8086         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
8087         parameters.
8088         (sreal constructor): Update.
8089         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
8090         sreal:operator/): Update.
8092 2018-08-29  Martin Liska  <mliska@suse.cz>
8094         * tree-switch-conversion.c (switch_conversion::expand):
8095         Strenghten assumption about gswitch statements.
8097 2018-08-29  Richard Biener  <rguenther@suse.de>
8099         PR tree-optimization/87117
8100         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
8101         re-value-number released SSA VDEFs.
8103 2018-08-29  Richard Biener  <rguenther@suse.de>
8105         PR tree-optimization/87126
8106         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
8108 2018-08-28  Jim Wilson  <jimw@sifive.com>
8110         * config/riscv/pic.md: Rewrite.
8111         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
8112         invalid address.
8113         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
8114         (SOFTF, default_load, softload, softstore): New.
8116 2018-08-28  Jeff Law  <law@redhat.com>
8118         * fold-const.c (fold_binary_loc): Remove recently added assert.
8120 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
8122         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
8123         to OP parmeter of generated function.
8125 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
8127         PR tree-optimization/87009
8128         * match.pd: Add boolean optimizations.
8130 2018-08-28  Martin Sebor  <msebor@redhat.com>
8132         PR middle-end/86631
8133         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
8134         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
8135         (pass_walloca::gate): Use it.
8136         (alloca_call_type): Same.
8137         (pass_walloca::execute): Same.
8138         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
8140 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
8142         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
8143         GCC_VERSION for usage of "__gcc_dump_printf__" format from
8144         >= 3005 to >= 9000.
8146 2018-08-28  Richard Biener  <rguenther@suse.de>
8148         PR tree-optimization/87124
8149         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
8150         constants before looking up avail.
8152 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
8154         PR middle-end/87099
8155         * calls.c (maybe_warn_nonstring_arg): Punt early if
8156         warn_stringop_overflow is zero.  Don't call get_range_strlen
8157         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
8158         Swap comparison operands to have constants on rhs.  Only use
8159         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
8160         increment lenrng[0].
8162 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
8164         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
8165         use of tree_to_shwi.  Remove duplicated test for the size being
8166         a whole number of bytes.
8168 2018-08-28  Richard Biener  <rguenther@suse.de>
8170         PR tree-optimization/87117
8171         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
8172         Handle removed stmt without LHS (GIMPLE_NOP).
8174 2018-08-28  Richard Biener  <rguenther@suse.de>
8176         PR tree-optimization/87117
8177         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
8178         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
8180 2018-08-28  Richard Biener  <rguenther@suse.de>
8182         PR tree-optimization/87117
8183         * tree-ssa-pre.c (compute_avail): Do not make expressions
8184         with predicated values available.
8185         (get_expr_value_id): Assert we do not run into predicated value
8186         expressions.
8188 2018-08-28  Richard Biener  <rguenther@suse.de>
8190         PR tree-optimization/87117
8191         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
8192         get virtual operands.
8193         (get_expr_operands): Handle STRING_CST like other decls.
8195 2018-08-28  Martin Liska  <mliska@suse.cz>
8197         * tree.h: Update documentation of fndecl_built_in_p
8198         functions.
8201 2018-08-27  Jeff Law  <law@redhat.com>
8202         PR tree-optimization/87110
8203         * tree-ssa-dse.c (compute_trims): Handle non-constant
8204         TYPE_SIZE_UNIT.
8206 2018-08-27  Martin Sebor  <msebor@redhat.com>
8208         PR tree-optimization/86914
8209         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
8211 2018-08-27  Martin Sebor  <msebor@redhat.com>
8213         PR tree-optimization/87112
8214         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
8215         the type of the bound argument.
8217 2018-08-27  Jeff Law  <law@redhat.com>
8219         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
8220         type does not have a TYPE_SIZE_UNIT.
8222 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
8224         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
8225         with include of backend.h.
8227 2018-08-27  Richard Biener  <rguenther@suse.de>
8229         PR tree-optimization/86927
8230         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
8231         use const cond reduction code.
8233 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
8235         PR tree-optimization/85758
8236         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
8238 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
8240         PR c++/87091
8241         * diagnostic-show-locus.c (class layout_range): Update for
8242         conversion of show_caret_p to a tri-state.
8243         (layout_range::layout_range): Likewise.
8244         (make_range): Likewise.
8245         (layout::maybe_add_location_range): Likewise.
8246         (layout::should_print_annotation_line_p): Don't show annotation
8247         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
8248         (layout::get_state_at_point): Update for conversion of
8249         show_caret_p to a tri-state.  Bail out early for
8250         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
8251         underlining or source colorization.
8252         (gcc_rich_location::add_location_if_nearby): Update for conversion
8253         of show_caret_p to a tri-state.
8254         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
8255         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
8256         Likewise.
8257         (selftest::test_one_liner_labels): Likewise.
8258         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
8259         conversion of show_caret_p to a tri-state.
8260         * pretty-print.c (text_info::set_location): Likewise.
8261         * pretty-print.h (text_info::set_location): Likewise.
8262         * substring-locations.c (format_warning_n_va): Likewise.
8263         * tree-diagnostic.c (default_tree_printer): Likewise.
8264         * tree-pretty-print.c (newline_and_indent): Likewise.
8266 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
8268         PR c++/87091
8269         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
8270         line above for line-insertion fix-it hints.
8271         (selftest::test_fixit_insert_containing_newline): Update the
8272         expected results, and add a test with line-numbering enabled.
8274 2018-08-27  Martin Liska  <mliska@suse.cz>
8276         PR sanitizer/86962
8277         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
8278         params with DECL_HAS_VALUE_EXPR_P.
8280 2018-08-27  Martin Liska  <mliska@suse.cz>
8282         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
8283         selected expansion strategy.
8285 2018-08-27  Martin Liska  <mliska@suse.cz>
8287         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
8288         * builtins.c (is_builtin_fn): Likewise.
8289         * attribs.c (diag_attr_exclusions): Use new function
8290         fndecl_built_in_p and remove check for FUNCTION_DECL if
8291         possible.
8292         (builtin_mathfn_code): Likewise.
8293         (fold_builtin_expect): Likewise.
8294         (fold_call_expr): Likewise.
8295         (fold_builtin_call_array): Likewise.
8296         (fold_call_stmt): Likewise.
8297         (set_builtin_user_assembler_name): Likewise.
8298         (is_simple_builtin): Likewise.
8299         * calls.c (gimple_alloca_call_p): Likewise.
8300         (maybe_warn_nonstring_arg): Likewise.
8301         * cfgexpand.c (expand_call_stmt): Likewise.
8302         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
8303         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
8304         (cgraph_node::verify_node): Likewise.
8305         * cgraphclones.c (build_function_decl_skip_args): Likewise.
8306         (cgraph_node::create_clone): Likewise.
8307         * config/arm/arm.c (arm_insert_attributes): Likewise.
8308         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
8309         * dse.c (scan_insn): Likewise.
8310         * expr.c (expand_expr_real_1): Likewise.
8311         * fold-const.c (operand_equal_p): Likewise.
8312         (fold_binary_loc): Likewise.
8313         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
8314         * gimple-low.c (lower_stmt): Likewise.
8315         * gimple-pretty-print.c (dump_gimple_call): Likewise.
8316         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
8317         Likewise.
8318         * gimple.c (gimple_build_call_from_tree): Likewise.
8319         (gimple_call_builtin_p): Likewise.
8320         (gimple_call_combined_fn): Likewise.
8321         * gimplify.c (gimplify_call_expr): Likewise.
8322         (gimple_boolify): Likewise.
8323         (gimplify_modify_expr): Likewise.
8324         (gimplify_addr_expr): Likewise.
8325         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
8326         * ipa-cp.c (determine_versionability): Likewise.
8327         * ipa-fnsummary.c (compute_fn_summary): Likewise.
8328         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
8329         * ipa-split.c (visit_bb): Likewise.
8330         (split_function): Likewise.
8331         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
8332         * lto-cgraph.c (input_node): Likewise.
8333         * lto-streamer-out.c (write_symbol): Likewise.
8334         * omp-low.c (setjmp_or_longjmp_p): Likewise.
8335         (lower_omp_1): Likewise.
8336         * predict.c (strip_predict_hints): Likewise.
8337         * print-tree.c (print_node): Likewise.
8338         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
8339         * trans-mem.c (is_tm_irrevocable): Likewise.
8340         (is_tm_load): Likewise.
8341         (is_tm_simple_load): Likewise.
8342         (is_tm_store): Likewise.
8343         (is_tm_simple_store): Likewise.
8344         (is_tm_abort): Likewise.
8345         (tm_region_init_1): Likewise.
8346         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
8347         * tree-cfg.c (verify_gimple_call): Likewise.
8348         (move_stmt_r): Likewise.
8349         (stmt_can_terminate_bb_p): Likewise.
8350         * tree-eh.c (lower_eh_constructs_2): Likewise.
8351         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
8352         * tree-inline.c (remap_gimple_stmt): Likewise.
8353         (copy_bb): Likewise.
8354         (estimate_num_insns): Likewise.
8355         (fold_marked_statements): Likewise.
8356         * tree-sra.c (scan_function): Likewise.
8357         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
8358         (optimize_stack_restore): Likewise.
8359         (pass_fold_builtins::execute): Likewise.
8360         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
8361         (mark_all_reaching_defs_necessary_1): Likewise.
8362         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
8363         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
8364         (pass_forwprop::execute): Likewise.
8365         * tree-ssa-loop-im.c (stmt_cost): Likewise.
8366         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
8367         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
8368         * tree-ssa-strlen.c (get_string_length): Likewise.
8369         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
8370         (find_func_aliases_for_call): Likewise.
8371         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
8372         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
8373         * tree-tailcall.c (find_tail_calls): Likewise.
8374         * tree.c (need_assembler_name_p): Likewise.
8375         (free_lang_data_in_decl): Likewise.
8376         (get_call_combined_fn): Likewise.
8377         * ubsan.c (is_ubsan_builtin_p): Likewise.
8378         * varasm.c (incorporeal_function_p): Likewise.
8379         * tree.h (DECL_BUILT_IN): Remove and replace with
8380         fndecl_built_in_p.
8381         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
8382         (fndecl_built_in_p): New.
8384 2018-08-27  Martin Liska  <mliska@suse.cz>
8386         PR tree-optimization/86847
8387         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
8388         Dump also subtree probability.
8389         (switch_decision_tree::do_jump_if_equal): New function.
8390         (switch_decision_tree::emit_case_nodes): Handle special
8391         situations in balanced tree that can be emitted much simpler.
8392         Fix calculation of probabilities that happen in tree expansion.
8393         * tree-switch-conversion.h (struct cluster): Add
8394         is_single_value_p.
8395         (struct simple_cluster): Likewise.
8396         (struct case_tree_node): Add new function has_child.
8397         (do_jump_if_equal): New.
8399 2018-08-27  Martin Liska  <mliska@suse.cz>
8401         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
8402         Add new argument to bit_test_cluster constructor.
8403         (bit_test_cluster::emit): Set bits really number of values
8404         handlel by a test.
8405         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
8406         probability argument.
8407         * tree-switch-conversion.h (struct bit_test_cluster):
8408         Add m_handles_entire_switch.
8410 2018-08-27  Martin Liska  <mliska@suse.cz>
8412         PR tree-optimization/86702
8413         * tree-switch-conversion.c (jump_table_cluster::emit):
8414         Make probabilities even for values in jump table
8415         according to number of cases handled.
8416         (switch_decision_tree::compute_cases_per_edge): Pass
8417         argument to reset_out_edges_aux function.
8418         (switch_decision_tree::analyze_switch_statement): Likewise.
8419         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
8420         Make it static.
8422 2018-08-27  Martin Liska  <mliska@suse.cz>
8424         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
8425         cfun argument explicitly.
8426         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
8427         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
8428         function gimple_switch_default_bb.
8429         (convert_switch_statements):
8430         (expand_builtins):
8431         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
8432         * stmt.c (label_to_block_fn): Use label_to_block and pass
8433         cfun argument explicitly and use gimple_switch_label_bb.
8434         (expand_case): Likewise.
8435         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
8436         cfun argument explicitly. Likewise.
8437         (make_edges_bb): Likewise.
8438         (make_cond_expr_edges): Likewise.
8439         (get_cases_for_edge): Likewise.
8440         (make_gimple_switch_edges): Likewise.
8441         (label_to_block_fn): Likewise.
8442         (label_to_block): Likewise.
8443         (make_goto_expr_edges): Likewise.
8444         (make_gimple_asm_edges): Likewise.
8445         (main_block_label): Likewise.
8446         (group_case_labels_stmt): Likewise.
8447         (find_taken_edge_computed_goto): Likewise.
8448         (find_taken_edge_switch_expr): Likewise.
8449         (gimple_verify_flow_info): Likewise.
8450         (gimple_redirect_edge_and_branch): Likewise.
8451         (gimple_switch_label_bb): New function.
8452         (gimple_switch_default_bb): Likewise.
8453         (gimple_switch_edge): Likewise.
8454         (gimple_switch_default_edge): Likewise.
8455         * tree-cfg.h (label_to_block_fn): Remove and replace ...
8456         (label_to_block): ... with this.
8457         (gimple_switch_label_bb): New.
8458         (gimple_switch_default_bb): Likewise.
8459         (gimple_switch_edge): Likewise.
8460         (gimple_switch_default_edge): Likewise.
8461         * tree-cfgcleanup.c (convert_single_case_switch): Use
8462         new gimple functions and pass new argument to label_to_block.
8463         (cleanup_control_flow_bb):
8464         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
8465         cfun argument explicitly.
8466         (make_eh_edges): Likewise.
8467         (redirect_eh_dispatch_edge): Likewise.
8468         (lower_resx): Likewise.
8469         (lower_eh_dispatch): Likewise.
8470         (maybe_remove_unreachable_handlers): Likewise.
8471         (unsplit_eh): Likewise.
8472         (cleanup_empty_eh): Likewise.
8473         (verify_eh_edges): Likewise.
8474         (verify_eh_dispatch_edge): Likewise.
8475         * tree-ssa-dom.c (record_edge_info): Likewise.
8476         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
8477         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
8478         (thread_through_normal_block): Likewise.
8479         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
8480         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
8481         * tree-switch-conversion.c (switch_conversion::collect): Use new
8482         gimple functions.
8483         (switch_conversion::check_final_bb): Likewise.
8484         (switch_conversion::gather_default_values): Pass new argument
8485         to label_to_block.
8486         (switch_conversion::build_constructors): Likewise.
8487         (switch_decision_tree::compute_cases_per_edge): Use new
8488         gimple_switch_edge function.
8489         (switch_decision_tree::analyze_switch_statement): Pass new argument
8490         to label_to_block.
8491         (switch_decision_tree::try_switch_expansion): Use
8492         gimple_switch_default_edge.
8493         * tree-vrp.c (find_switch_asserts): Pass new argument
8494         to label_to_block.
8495         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
8496         (vr_values::simplify_switch_using_ranges): Likewise.
8498 2018-08-27  Richard Biener  <rguenther@suse.de>
8500         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
8501         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
8503         * tree-ssa-sccvn.h (struct vn_pval): New structure.
8504         (struct vn_nary_op_s): Add unwind_to member.  Add
8505         predicated_values flag and put result into a union together
8506         with a linked list of vn_pval.
8507         (struct vn_ssa_aux): Add name member to make maintaining
8508         a map of SSA name to vn_ssa_aux possible.  Remove no longer
8509         needed info, dfsnum, low, visited, on_sccstack, use_processed
8510         and range_info_anti_range_p members.
8511         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
8512         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
8513         New functions.
8514         (vn_valueize): New global.
8515         (vn_context_bb): Likewise.
8516         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
8517         VN_INFO_PTR_INFO): Remove.
8518         * tree-ssa-sccvn.c: ... (rewrite)
8519         (pass_fre::execute): For -O2+ initialize loops and run
8520         RPO VN in optimistic mode (iterating).  For -O1 and -Og
8521         run RPO VN in non-optimistic mode.
8522         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
8523         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
8524         * doc/invoke.texi (sccvn-max-scc-size): Remove.
8525         (rpo-vn-max-loop-depth): Document.
8526         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
8527         when valuezing the VUSE signals we walked out of the region.
8528         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
8529         (phi_translate): Set VN context block to use for availability
8530         lookup.
8531         (compute_avail): Likewise.
8532         (pre_valueize): New function.
8533         (pass_pre::execute): Adjust to the RPO VN API.
8535         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
8536         (propagate_constants_for_unrolling): Remove.
8537         (tree_unroll_loops_completely): Perform value-numbering
8538         on the unrolled bodies loop parent.
8540 2018-08-27  Richard Biener  <rguenther@suse.de>
8542         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
8543         for partial antic compute.
8545 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
8547         PR rtl-optimization/87065
8548         * combine.c (simplify_if_then_else): Formatting fix.
8549         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
8550         check.
8551         (known_cond): Don't return const_true_rtx for vector modes.  Use
8552         CONST0_RTX instead of const0_rtx.  Formatting fixes.
8554 2018-08-27  Martin Liska  <mliska@suse.cz>
8556         PR gcov-profile/87069
8557         * gcov.c (process_file): Record files already processed
8558         and warn about a file being processed multiple times.
8560 2018-08-27  Martin Liska  <mliska@suse.cz>
8562         PR driver/83193
8563         * config/aarch64/aarch64.c (aarch64_override_options_internal):
8564         Set default values for x_aarch64_*_string strings.
8565         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
8566         prefix.  For -mabi do not print '=ABI' in help and use
8567         <option_value> format for -msve-vector-bits and -moverride
8568         options.
8570 2018-08-26  Jeff Law  <law@redhat.com>
8572         * config/mips/frame-header-opt.c: Include "backend.h" rather than
8573         "cfg.h"
8575 2018-08-26  Marek Polacek  <polacek@redhat.com>
8577         PR c++/87029, Implement -Wredundant-move.
8578         * doc/invoke.texi: Document -Wredundant-move.
8580 2018-08-25  Martin Sebor  <msebor@redhat.com>
8582         PR tree-optimization/87059
8583         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
8584         to the same type as the other.
8585         * fold-const.c (fold_binary_loc): Assert expectation.
8587 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
8589         * config/darwin.c (machopic_legitimize_pic_address): Clean up
8590         extraneous parentheses, dead code section and formatting.
8592 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
8594         PR c++/87091
8595         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
8596         wide enough for jumps in the line-numbering to be visible.
8597         (layout::print_gap_in_line_numbering): New member function.
8598         (layout::calculate_line_spans): When using line numbering, merge
8599         line spans that are only 1 line apart.
8600         (diagnostic_show_locus): When printing line numbers, show gaps in
8601         line numbering directly, rather than printing headers.
8602         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
8603         line-numbering with multiple line spans.
8604         (selftest::test_fixit_insert_containing_newline_2): Add test of
8605         line-numbering, in which the spans are close enough to be merged.
8607 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8609         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
8610         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
8611         * tree-vrp.c (range_is_nonnull): Remove.
8612         (range_includes_zero_p): Accept value_range instead of min/max.
8613         (extract_range_from_binary_expr_1): Do not early bail on
8614         POINTER_PLUS_EXPR.
8615         Use range_includes_zero_p instead of range_is_nonnull.
8616         (extract_range_from_unary_expr): Use range_includes_zero_p instead
8617         of range_is_nonnull.
8618         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
8619         special case VR_ANTI_RANGE.
8620         (vrp_finalize): Same.
8621         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
8622         instead of min/max.
8623         (range_is_nonnull): Remove.
8624         * vr-values.c (vrp_stmt_computes_nonzero): Use
8625         range_includes_zero_p instead of range_is_nonnull.
8626         (extract_range_basic): Pass value_range to range_includes_zero_p
8627         instead of range_is_nonnull.
8629 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
8631         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
8632         * emit-rtl.h (rtl_data): Remove return_bnd.
8633         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
8634         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
8635         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
8636         (POINTER_BOUNDS_MODE): Remove definition.
8637         (make_pointer_bounds_mode): Remove.
8638         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
8639         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
8640         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
8641         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
8642         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
8643         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
8644         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
8646         * config/i386/i386-modes.def (BND32, BND64): Remove.
8647         * config/i386/i386.c (dbx_register_map): Remove bound registers.
8648         (dbx64_register_map): Ditto.
8649         (svr4_dbx_register_map): Ditto.
8650         (indirect_thunk_bnd_needed): Remove.
8651         (indirect_thunks_bnd_used): Ditto.
8652         (indirect_return_bnd_needed): Ditto.
8653         (indirect_return_via_cx_bnd): Ditto.
8654         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
8655         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
8656         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
8657         (output_indirect_thunk_function): Remove handling of
8658         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
8659         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
8660         (ix86_save_reg): Remove handling of crtl->return_bnd.
8661         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
8662         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
8663         and UNSPEC_BNDLX_ADDR.
8664         (ix86_output_indirect_branch_via_reg): Remove handling of
8665         indirect_thunk_prefix_bnd.
8666         (ix86_output_indirect_branch_via_push): Ditto.
8667         (ix86_output_function_return): Ditto.
8668         (ix86_output_indirect_function_return): Ditto.
8669         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
8670         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
8671         (CALL_USED_REGISTERS): Ditto.
8672         (REG_ALLOC_ORDER): Update for removal of bound registers.
8673         (HI_REGISTER_NAMES): Ditto.
8674         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
8675         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
8676         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
8677         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
8678         (FIRST_PSEUDO_REG): Update.
8679         (BND): Remove mode iterator.
8680         * config/i386/predicates.md (bnd_mem_operator): Remove.
8682 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8684         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
8685         vectors.
8687 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
8689         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
8690         the case in which the permute needs only a single element and
8691         repeats for every vector of the result.  Extend that case to
8692         handle variable-length vectors.
8693         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
8695 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
8697         PR debug/79342
8698         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
8699         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
8701 2018-08-24  Richard Biener  <rguenther@suse.de>
8703         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
8704         bb_flags_allocated members.
8705         (auto_flag): New RAII class for allocating flags.
8706         (auto_edge_flag): New RAII class for allocating edge flags.
8707         (auto_bb_flag): New RAII class for allocating bb flags.
8708         * cfgloop.c (verify_loop_structure): Allocate temporary edge
8709         flag dynamically.
8710         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
8711         in favor of temporarily allocated BB flag.
8712         * hsa-brig.c: Re-order includes.
8713         * hsa-dump.c: Likewise.
8714         * hsa-regalloc.c: Likewise.
8715         * print-rtl.c: Likewise.
8716         * profile-count.c: Likewise.
8718 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
8720         PR target/86989
8721         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
8722         the TOC register.
8724 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
8726         PR 87073/bootstrap
8727         * wide-int-range.cc (wide_int_range_div): Do not ignore result
8728         from wide_int_range_multiplicative_op.
8730 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
8732         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
8733         "permutaion".
8735 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
8737         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
8738         to 'expanded'.
8740 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
8742         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
8743         full GENERIC expression used for replacement.
8745 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
8747         * tree-vrp.c (abs_extent_range): Remove.
8748         (extract_range_into_wide_ints): Pass wide ints by reference.
8749         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
8750         Pass wide ints by reference in all calls to
8751         extract_range_into_wide_ints.
8752         * wide-int-range.cc (wide_int_range_div): New.
8753         * wide-int-range.h (wide_int_range_div): New.
8754         (wide_int_range_includes_zero_p): New.
8755         (wide_int_range_zero_p): New.
8757 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
8759         * config/aarch64/aarch64.md (arches): New enum.
8760         (arch): New enum attr.
8761         (arch_enabled): New attr.
8762         (enabled): Now uses arch_enabled only.
8763         (simd, sve, fp16): Removed attribute.
8764         (fp): Attr now defined in terms of 'arch'.
8765         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
8766         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
8767         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
8768         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
8769         attributes into 'arch'.
8770         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
8771         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
8772         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
8773         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
8774         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
8775         'simd' attribute into 'arch'.
8776         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
8777         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
8778         Convert use of 'fp' attribute to 'arch'.
8779         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
8780         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
8781         into 'arch'.
8782         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
8783         (different modes) Merge 'fp' and 'simd' into 'arch'.
8784         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
8785         'simd' into 'arch'.
8787 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
8789         PR rtl-optimization/87026
8790         * expmed.c (canonicalize_comparison): If we can no longer create
8791         pseudoregisters, don't.
8793 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
8795         PR target/86951
8796         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
8797         prototype.
8798         * config/arm/arm.c (speculation_barrier_libfunc): New static
8799         variable.
8800         (arm_init_libfuncs): Initialize it.
8801         (arm_emit_speculation_barrier): New function.
8802         * config/arm/arm.md (speculation_barrier): Call
8803         arm_emit_speculation_barrier for architectures that do not have
8804         DSB or ISB.
8805         (speculation_barrier_insn): Only match on Armv7 or later.
8807 2018-08-23  Richard Biener  <rguenther@suse.de>
8809         PR middle-end/87024
8810         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
8811         calls.
8813 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8815         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
8816         of single-vector TBLs.
8817         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
8818         one input is given.
8820 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
8822         PR target/85910
8823         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
8824         aarch64_evpc_tbl guard.
8826 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8828         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
8829         behaviour.
8831 2018-08-22  Martin Sebor  <msebor@redhat.com>
8833         PR middle-end/87052
8834         * tree-pretty-print.c (pretty_print_string): Add argument.
8835         (dump_generic_node): Call to pretty_print_string with string size.
8837 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
8839         PR rtl-optimization/86771
8840         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
8841         of two SETs into those two SETs, one to be placed at i2, if that SETs
8842         destination is modified between i2 and i3.
8844 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8846         PR tree-optimization/86725
8847         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
8848         function.
8849         (vect_analyze_scalar_cycles_1): Check it.
8851 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8853         PR tree-optimization/86725
8854         * tree-vect-loop.c (vect_is_simple_reduction): When treating
8855         an outer loop phi as a double reduction, make sure that the
8856         single user of the phi result is an inner loop phi.
8858 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8860         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
8861         grouped stores with gaps to a strided group.
8863 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
8865         * tree-vect-stmts.c (get_group_load_store_type)
8866         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
8867         first statement in a group.
8869 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8871         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
8872         the sequence used in gcc/gcc.c.
8874 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8876         PR other/704
8877         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
8878         building it.
8880 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8882         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
8883         Darwin10-specific unwinder-shim.
8884         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
8885         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
8886         New to cater for Darwin10 Rosetta.
8888 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8890         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
8891         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
8893 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
8895         PR bootstrap/81033
8896         PR target/81733
8897         PR target/52795
8898         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
8899         (dwarf2out_switch_text_section): Generate a local label for the
8900         second function sub-section and apply it as the second FDE start
8901         label.
8902         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
8903         second sub-section start.
8905 2018-08-22  Richard Biener  <rguenther@suse.de>
8907         PR tree-optimization/86988
8908         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
8910 2018-08-22  Richard Biener  <rguenther@suse.de>
8912         PR tree-optimization/86945
8913         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
8915 2018-08-22  Alexandre Oliva <oliva@adacore.com>
8917         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
8918         a comment about how uses of r2 for .sdata2 come about.
8920 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
8922         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
8924 2018-08-21  Marek Polacek  <polacek@redhat.com>
8926         PR c++/86981, Implement -Wpessimizing-move.
8927         * doc/invoke.texi: Document -Wpessimizing-move.
8929 2018-08-21  Jan Hubicka  <jh@suse.cz>
8931         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
8932         * tree.h (is_redundant_typedef): Remove.
8933         * dwarf2out.c (is_redundant_typedef): Turn into static function.
8935 2018-08-21  Jan Hubicka  <jh@suse.cz>
8937         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
8938         when possible.
8940 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
8942         * expmed.c (extract_low_bits): Reject invalid subregs early.
8944 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8946         PR middle-end/86121
8947         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
8948         behaviour.
8950 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8952         * config/vxworks.h: Guard vxworks_asm_out_constructor and
8953         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
8954         * config/vxworks.c: Likewise.
8956 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8958         * config/vxworks.c: Set targetm.have_ctors_dtors
8959         if HAVE_INITFINI_ARRAY_SUPPORT.
8960         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
8961         if HAVE_INITFINI_ARRAY_SUPPORT.
8963 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
8965         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
8966         default search path for VxWorks < 7.
8968 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8970         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
8971         (get_format_string): Refer to c_getstr.
8973 2018-08-21  Tom de Vries  <tdevries@suse.de>
8975         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8976         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8977         (debuginfo_early_stop): Declare.
8978         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
8979         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
8980         (debuginfo_early_stop): New function.
8981         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
8982         and debuginfo_early_stop.
8983         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
8984         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
8985         (general_init): Call debuginfo_early_init.
8986         (finalize): Call debuginfo_fini.
8987         (do_compile): Call debuginfo_init.
8988         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
8989         -fdump-early-debug.
8990         (@item -fdump-debug, @item -fdump-earlydebug): Add.
8992 2018-08-21  Tom de Vries  <tdevries@suse.de>
8994         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
8995         flag_dump_noaddr and flag_dump_unnumbered.
8997 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
8999         * wide-int-range.cc (wide_int_range_abs): New.
9000         (wide_int_range_order_set): Rename from wide_int_range_min_max.
9001         * wide-int-range.h (wide_int_range_abs): New.
9002         (wide_int_range_min_max): New.
9003         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
9004         case to call wide_int_range_abs.
9005         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
9006         (extract_range_from_abs_expr): Delete.
9008 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
9010         PR target/87033
9011         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
9012         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
9013         for indexed loads.
9015 2018-08-20  Nathan Sidwell  <nathan@acm.org>
9016             Jeff Law <law@redhat.com>
9018         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
9019         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
9021 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
9023         PR other/84889
9024         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
9025         (decl_attributes): Likewise.
9026         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
9027         instance.
9028         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
9029         * diagnostic-core.h (class auto_diagnostic_group): New class.
9030         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
9031         (diagnostic_report_diagnostic): Handle the first diagnostics within
9032         a group.
9033         (emit_diagnostic): Add auto_diagnostic_group instance.
9034         (inform): Likewise.
9035         (inform_n): Likewise.
9036         (warning): Likewise.
9037         (warning_at): Likewise.
9038         (warning_n): Likewise.
9039         (pedwarn): Likewise.
9040         (permerror): Likewise.
9041         (error): Likewise.
9042         (error_n): Likewise.
9043         (error_at): Likewise.
9044         (sorry): Likewise.
9045         (fatal_error): Likewise.
9046         (internal_error): Likewise.
9047         (internal_error_no_backtrace): Likewise.
9048         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
9049         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
9050         * diagnostic.h (struct diagnostic_context): Add fields
9051         "diagnostic_group_nesting_depth",
9052         "diagnostic_group_emission_count", "begin_group_cb",
9053         "end_group_cb".
9054         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
9055         Add auto_diagnostic_group instance(s).
9056         (find_explicit_erroneous_behavior): Likewise.
9057         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
9058         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
9059         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
9060         (gimplify_va_arg_expr): Likewise.
9061         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
9062         (HSA_SORRY_AT): Likewise.
9063         * ipa-devirt.c (compare_virtual_tables): Likewise.
9064         (warn_odr): Likewise.
9065         * multiple_target.c (expand_target_clones): Likewise.
9066         * opts-common.c (cmdline_handle_error): Likewise.
9067         * reginfo.c (globalize_reg): Likewise.
9068         * substring-locations.c (format_warning_n_va): Likewise.
9069         * tree-inline.c (expand_call_inline): Likewise.
9070         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
9071         * tree-ssa-loop-niter.c
9072         (do_warn_aggressive_loop_optimizations): Likewise.
9073         * tree-ssa-uninit.c (warn_uninit): Likewise.
9074         * tree.c (warn_deprecated_use): Likewise.
9076 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
9078         PR target/87014
9079         * config/i386/i386.md (eh_return): Always update EH return
9080         address in word_mode.
9082 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
9084         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
9085         TARGET_SPLIT_COMPLEX_ARG is defined.
9087 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9089         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
9091 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9093         PR target/86984
9094         * expr.c (expand_assignment): Assert that bitpos is positive.
9095         (store_field): Likewise
9096         (expand_expr_real_1): Make sure that bitpos is positive.
9097         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
9098         integer overflow.
9100 2018-08-20  Nathan Sidwell  <nathan@acm.org>
9102         * Makefile.in (CPP_ID_DATA_H): Delete.
9103         (CPP_INTERNAL_H): Don't add it.
9104         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
9105         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
9107 2018-08-20  Richard Biener  <rguenther@suse.de>
9109         PR tree-optimization/78655
9110         * tree-vrp.c (extract_range_from_binary_expr_1): Make
9111         pointer + offset nonnull if either operand is nonnull work.
9113 2018-08-20  Tom de Vries  <tdevries@suse.de>
9115         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
9116         unless the referenced die describes the added property using
9117         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
9118         Otherwise, add a DW_AT_location to the referenced die.
9120 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
9122         PR target/86994
9123         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
9124         register_operand when calling ix86_set_reg_reg_cost.
9125         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
9126         Set *total to 0 for operands that satisfy x86_64_immediate_operand
9127         predicate and to 1 otherwise.
9129 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
9131         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
9132         emit a diagnostic that it is not supported and reset the option.
9133         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
9134         supported and consume the option.  (ASM_FINAL_SPEC): New.
9136 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
9138         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
9139         a sentence.
9141 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
9143         C-SKY port: Documentation
9145         * doc/extend.texi (C-SKY Function Attributes): New section.
9146         * doc/invoke.texi (Option Summary): Add C-SKY options.
9147         (C-SKY Options): New section.
9148         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
9150 2018-08-17  Jojo  <jijie_rong@c-sky.com>
9151             Huibin Wang  <huibin_wang@c-sky.com>
9152             Sandra Loosemore  <sandra@codesourcery.com>
9153             Chung-Lin Tang  <cltang@codesourcery.com>
9155         C-SKY port: Backend implementation
9157         * config/csky/*: New.
9158         * common/config/csky/*: New.
9160 2018-08-17  Jojo  <jijie_rong@c-sky.com>
9161             Huibin Wang  <huibin_wang@c-sky.com>
9162             Sandra Loosemore  <sandra@codesourcery.com>
9163             Chung-Lin Tang  <cltang@codesourcery.com>
9164             Andrew Jenner  <andrew@codesourcery.com>
9166         C-SKY port: Configury
9168         * config.gcc (csky-*-*): New.
9169         * configure.ac: Add csky to targets for dwarf2 debug_line support.
9170         * configure: Regenerated.
9172 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
9174         * dump-context.h: Include "dumpfile.h".
9175         (dump_context::dump_printf_va): Convert final param from va_list
9176         to va_list *.  Convert from ATTRIBUTE_PRINTF to
9177         ATTRIBUTE_GCC_DUMP_PRINTF.
9178         (dump_context::dump_printf_loc_va): Likewise.
9179         * dumpfile.c: Include "stringpool.h".
9180         (make_item_for_dump_printf_va): Delete.
9181         (make_item_for_dump_printf): Delete.
9182         (class dump_pretty_printer): New class.
9183         (dump_pretty_printer::dump_pretty_printer): New ctor.
9184         (dump_pretty_printer::emit_items): New member function.
9185         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
9186         function.
9187         (dump_pretty_printer::emit_item): New member function.
9188         (dump_pretty_printer::stash_item): New member function.
9189         (dump_pretty_printer::format_decoder_cb): New member function.
9190         (dump_pretty_printer::decode_format): New member function.
9191         (dump_context::dump_printf_va): Reimplement in terms of
9192         dump_pretty_printer.
9193         (dump_context::dump_printf_loc_va): Convert final param from va_list
9194         to va_list *.
9195         (dump_context::begin_scope): Reimplement call to
9196         make_item_for_dump_printf.
9197         (dump_printf): Update for change to dump_printf_va.
9198         (dump_printf_loc): Likewise.
9199         (selftest::test_capture_of_dump_calls): Convert "stmt" from
9200         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
9201         with %T, %E, and %G.
9202         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
9203         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
9204         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
9205         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
9206         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
9207         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
9208         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
9209         within a dump_printf_loc call to "%wu".
9210         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
9211         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
9212         missing space after "=".
9213         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
9214         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
9215         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
9216         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
9217         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
9218         duplicate "vectorized" from message.
9220 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
9222         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
9223         polyNxK_t element's TYPE_STRING_FLAG.
9225 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
9227         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
9228         (they were unnamed before).  Fix comments.
9230 2018-08-17  Nathan Sidwell  <nathan@acm.org>
9232         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
9234 2018-08-17  Richard Biener  <rguenther@suse.de>
9236         PR tree-optimization/86841
9237         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
9239 2018-08-17  Martin Liska  <mliska@suse.cz>
9241         * common.opt: Remove Warn, Init and Report for options with
9242         Ignore/Deprecated flag. Warning is done automatically for
9243         Deprecated flags.
9244         * config/i386/i386.opt: Likewise.
9245         * config/ia64/ia64.opt: Likewise.
9246         * config/rs6000/rs6000.opt: Likewise.
9247         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
9248         Remove usage of flag_check_pointer_bounds.
9249         * lto-wrapper.c (merge_and_complain): Do not handle
9250         OPT_fcheck_pointer_bounds.
9251         (append_compiler_options): Likewise.
9252         * opt-functions.awk: Do not handle Deprecated.
9253         * optc-gen.awk: Check that Var, Report and Init are not
9254         used for an option with Ignore/Deprecated flag.
9255         * opts-common.c (decode_cmdline_option): Do not report
9256         CL_ERR_DEPRECATED.
9257         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
9258         options.
9259         * opts.h (struct cl_option): Remove cl_deprecated flag.
9260         (CL_ERR_DEPRECATED): Remove error enum value.
9262 2018-08-17  Richard Biener  <rguenther@suse.de>
9264         PR middle-end/86505
9265         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
9266         across a va-arg-pack using call adjust its return value accordingly.
9268 2018-08-16  Martin Sebor  <msebor@redhat.com>
9270         PR tree-optimization/86853
9271         * gimple-ssa-sprintf.c (struct format_result): Rename member.
9272         (struct fmtresult): Add member and initialize it in ctors.
9273         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
9274         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
9275         (format_directive): Set POSUNDER4K when MAYFAIL is set.
9276         (parse_directive): Handle %C same as %c.
9277         (sprintf_dom_walker::compute_format_length): Adjust.
9278         (is_call_safe): Adjust.
9280 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
9282         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
9283         for determining how to count the elements.
9284         * builtins.h (c_strlen): Adjust prototype.
9285         * expr.c (string_constant): Add new parameter mem_size.
9286         Set *mem_size appropriately.
9287         * expr.h (string_constant): Adjust protoype.
9288         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
9289         * gimple-fold.h (get_range_strlen): Adjust prototype.
9290         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
9291         (format_string): Call get_string_length with eltsize.
9293 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9295         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
9296         to emit the span, rather than setting it as the prefix.
9298 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
9300         * diagnostic-show-locus.c (layout::start_annotation_line): Add
9301         "margin_char" parameter, defaulting to space.  Use it in place
9302         of pp_space for the initial part of the margin.
9303         (layout::print_leading_fixits): Use '+' when filling the margin
9304         of line-insertion fix-it hints.
9306 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9308         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
9309         Delete.
9311 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9313         * config/rs6000/altivec.md: Don't set length attribute to the default
9314         value.
9315         * config/rs6000/darwin.md: Ditto.
9316         * config/rs6000/dfp.md: Ditto.
9317         * config/rs6000/htm.md: Ditto.
9318         * config/rs6000/rs6000.md: Ditto.
9319         * config/rs6000/sync.md: Ditto.
9320         * config/rs6000/vsx.md: Ditto.
9322 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9324         * config/rs6000/altivec.md: Don't set length attribute to the default
9325         value, for branch instructions.
9326         * config/rs6000/darwin.md: Ditto.
9327         * config/rs6000/rs6000.md: Ditto.
9329 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
9331         * config/rs6000/rs6000.md (length): Always define as const_int 4.
9332         (unnamed conditional branch define_insn): Set length to 4 or 8
9333         depending on offset.
9334         (<bd>_<mode>): Similar, for alternative 0.
9335         (<bd>tf_<mode>): Ditto.
9337 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9339         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
9341 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
9343         * doc/rtl.texi: Replace old RTX class names with new names.
9346 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
9348         * expmed.h (canonicalize_comparison): New declaration.
9349         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
9350         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
9351         * optabs.c (prepare_cmp_insn): Likewise.
9352         * rtl.h (unsigned_condition_p): New function which checks if a
9353         comparison operator is unsigned.
9355 2018-08-16  Nathan Sidwell  <nathan@acm.org>
9357         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
9358         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
9360 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9362         PR target/84711
9363         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
9364         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
9365         (mov<mov>): ..this and enable unconditionally.
9367 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
9369         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
9371 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
9373         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
9374         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
9375         "Common" with "Target".
9377 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
9379         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
9380         * doc/invoke.texi (mmitigate-rop): Remove.
9381         * config/i386/i386.c: Do not include "regrename.h".
9382         (ix86_rop_should_change_byte_p, reg_encoded_number)
9383         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
9384         Remove.
9385         (ix86_reorg): Remove call to ix86_mitigate_rop.
9386         * config/i386/i386.md (attr "modrm_class"): Remove.
9387         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
9388         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
9389         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
9391 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
9393         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
9394         allow folding of mergeh() and mergel() for the float and double types.
9395         (fold_mergehl_helper): Rework to handle building a permute tree
9396         for float vectors.
9398 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
9400         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
9401         for TARGET_SSE.
9403 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
9405         * common.opt (fdiagnostics-show-labels): New option.
9406         * diagnostic-show-locus.c (class layout_range): Add field
9407         "m_label".
9408         (class layout): Add field "m_show_labels_p".
9409         (layout_range::layout_range): Add param "label" and use it to
9410         initialize m_label.
9411         (make_range): Pass in NULL for new "label" param of layout_range's
9412         ctor.
9413         (layout::layout): Initialize m_show_labels_p.
9414         (layout::maybe_add_location_range): Pass in loc_range->m_label
9415         when constructing layout_range instances.
9416         (struct line_label): New struct.
9417         (layout::print_any_labels): New member function.
9418         (layout::print_line): Call it if label-printing is enabled.
9419         (selftest::test_one_liner_labels): New test.
9420         (selftest::test_diagnostic_show_locus_one_liner): Call it.
9421         * diagnostic.c (diagnostic_initialize): Initialize
9422         context->show_labels_p.
9423         * diagnostic.h (struct diagnostic_context): Add field
9424         "show_labels_p".
9425         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9426         -fno-diagnostics-show-labels.
9427         * dwarf2out.c (gen_producer_string): Add
9428         OPT_fdiagnostics_show_labels to the ignored options.
9429         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
9430         param.
9431         (gcc_rich_location::maybe_add_expr): Likewise.
9432         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
9433         label" param, defaulting to NULL.
9434         (gcc_rich_location::add_expr): Add "label" param.
9435         (gcc_rich_location::maybe_add_expr): Likewise.
9436         (class text_range_label): New class.
9437         (class range_label_for_type_mismatch): New class.
9438         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
9439         of format_warning_va.
9440         (fmtwarn_n): Likewise for new params of format_warning_n_va.
9441         * lto-wrapper.c (merge_and_complain): Add
9442         OPT_fdiagnostics_show_labels to the "pick one setting" options.
9443         (append_compiler_options): Likewise to the dropped options.
9444         (append_diag_options): Likewise to the passed-on options.
9445         * opts.c (common_handle_option): Handle the new option.
9446         * selftest-diagnostic.c
9447         (test_diagnostic_context::test_diagnostic_context): Enable
9448         show_labels_p.
9449         * substring-locations.c: Include "gcc-rich-location.h".
9450         (format_warning_n_va): Add "fmt_label" and "param_label" params
9451         and use them as appropriate.
9452         (format_warning_va): Add "fmt_label" and "param_label" params,
9453         passing them on to format_warning_n_va.
9454         (format_warning_at_substring): Likewise.
9455         (format_warning_at_substring_n): Likewise.
9456         * substring-locations.h (format_warning_va): Add "fmt_label" and
9457         "param_label" params.
9458         (format_warning_n_va): Likewise.
9459         (format_warning_at_substring): Likewise.
9460         (format_warning_at_substring_n): Likewise.
9461         * toplev.c (general_init): Initialize global_dc->show_labels_p.
9463 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
9465         PR testsuite/86519
9466         * builtins.c (expand_builtin_memcmp): Do not expand the call
9467         when overflow is detected.
9469 2018-08-15  Martin Sebor  <msebor@redhat.com>
9471         PR tree-optimization/71625
9472         * config/aarch64/aarch64-builtins.c
9473         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
9475 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
9477         * config/s390/s390.c (s390_reorg): Remove loop.
9479 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9481         * config/darwin.c
9482          (darwin_function_switched_text_sections): Delete.
9483         * gcc/config/darwin.h
9484          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
9486 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
9488         PR target/81685
9489         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
9490         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
9491         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
9493 2018-08-15  Martin Liska  <mliska@suse.cz>
9495         PR tree-optimization/86925
9496         * predict.c (expr_expected_value_1): When taking
9497         later predictor, assign also probability.
9498         Use fold_build2_initializer_loc in order to fold
9499         the expression in -frounding-math.
9501 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
9503         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
9504         patterns.
9505         (expand_vec_perm_1): Try the new method.
9507 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
9509         PR target/86547
9510         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
9511         Check whether lra_live_max_point is 0 before dividing.
9513 2018-08-14  Martin Sebor  <msebor@redhat.com>
9515         PR tree-optimization/86650
9516         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
9517         (vrp_prop::check_mem_ref): Same.
9519 2018-08-13  Liu Hao <lh_mouse@126.com>
9521         * pretty-print.c (eat_esc_sequence): Swap the foreground and
9522         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
9523         and clear it thereafter, as it only works for DBCS.
9525 2018-08-13  Liu Hao <lh_mouse@126.com>
9527         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
9528         handle returned by _get_osf_handle().
9530 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9532         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
9533         for folding vec_perm.
9535 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
9537         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
9538         Add support for gimple-folding of vec_pack() and vec_unpack()
9539         intrinsics.
9541 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
9543         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
9544         vec_xst variants to the list.
9545         (rs6000_gimple_fold_builtin): Add support for folding unaligned
9546         vector loads and stores.
9548 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
9550         * config.gcc (rs6000-ibm-aix4.x): Delete.
9551         (rs6000-ibm-aix5.1): Delete.
9552         (rs6000-ibm-aix5.2): Delete.
9553         (rs6000-ibm-aix5.3): Delete.
9554         * config/rs6000/aix43.h: Delete.
9555         * config/rs6000/aix51.h: Delete.
9556         * config/rs6000/aix52.h: Delete.
9557         * config/rs6000/t-aix43: Delete.
9559 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
9561         * config/s390/s390.c (s390_decompose_constant_pool_ref):
9562         New function.
9563         (s390_decompose_address): Factor out constant pool ref
9564         decomposition.
9566 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9568         * config/nds32/nds32-predicates.c
9569         (nds32_can_use_bclr_p): Change return type as bool.
9570         (nds32_can_use_bset_p): Ditto.
9571         (nds32_can_use_btgl_p): Ditto.
9572         (nds32_can_use_bitci_p): Ditto.
9573         * config/nds32/nds32-protos.h
9574         (nds32_can_use_bclr_p): Change declaration.
9575         (nds32_can_use_bset_p): Ditto.
9576         (nds32_can_use_btgl_p): Ditto.
9577         (nds32_can_use_bitci_p): Ditto.
9579 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9581         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
9582         Support -msched-prolog-epilog option.
9583         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
9585 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9587         * common/config/nds32/nds32-common.c
9588         (nds32_option_optimization_table): Enalbe -malways-align.
9590 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
9592         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
9593         extra_headers.
9594         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
9595         OPT_misr_secure_ case.
9596         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
9597         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
9598         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
9599         secure attribute.
9600         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
9601         (nds32_isr_info): New field security_level.
9602         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
9603         * config/nds32/nds32.md (return_internal): Consider critical attribute.
9604         * config/nds32/nds32.opt (misr-secure): New option.
9605         * config/nds32/nds32_init.inc: New file.
9606         * config/nds32/nds32_isr.h: New file.
9608 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
9610         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
9611         Update comment for atomic instructions.
9612         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
9613         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
9614         Remove.
9615         (atomic_loaddi): Revise fence expansion to only emit fence prior to
9616         load for __ATOMIC_SEQ_CST model.
9617         (atomic_loaddi_1): Remove float register target.
9618         (atomic_storedi): Handle CONST_INT values.
9619         (atomic_storedi_1): Remove float register source.  Add special case
9620         for zero value.
9621         (memory_barrier): New expander and insn.
9623 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
9625         PR tree-optimization/86835
9626         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
9627         new_stmt after def_gsi, make sure to insert new_square_stmt after
9628         that stmt, not 2 stmts before it.
9630 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
9632         PR target/82418
9633         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
9634         instead of SWI48.
9636 2018-08-10  Martin Liska  <mliska@suse.cz>
9638         PR target/83610
9639         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
9640         function type.
9641         * builtins.c (expand_builtin_expect_with_probability):
9642         New function.
9643         (expand_builtin_expect_with_probability): New function.
9644         (build_builtin_expect_predicate): Add new argumnet probability
9645         for BUILT_IN_EXPECT_WITH_PROBABILITY.
9646         (fold_builtin_expect):
9647         (fold_builtin_2):
9648         (fold_builtin_3):
9649         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
9650         * builtins.h (fold_builtin_expect): Set new argument.
9651         * doc/extend.texi: Document __builtin_expect_with_probability.
9652         * doc/invoke.texi: Likewise.
9653         * gimple-fold.c (gimple_fold_call): Pass new argument.
9654         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
9655         also BUILT_IN_EXPECT_WITH_PROBABILITY.
9656         * predict.c (get_predictor_value): New function.
9657         (expr_expected_value): Add new argument probability. Assume
9658         that predictor and probability are always non-null.
9659         (expr_expected_value_1): Likewise.  For __builtin_expect and
9660         __builtin_expect_with_probability set probability.  Handle
9661         combination in binary expressions.
9662         (tree_predict_by_opcode): Simplify code by simply calling
9663         get_predictor_value.
9664         (pass_strip_predict_hints::execute): Add handling of
9665         BUILT_IN_EXPECT_WITH_PROBABILITY.
9666         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
9667         new predictor.
9668         * tree.h (DECL_BUILT_IN_P): New function.
9670 2018-08-10  Martin Liska  <mliska@suse.cz>
9672         PR tree-optimization/85799
9673         * passes.def: Add argument for pass_strip_predict_hints.
9674         * predict.c (class pass_strip_predict_hints): Add new argument
9675         early_p.
9676         (strip_predictor_early): New function.
9677         (pass_strip_predict_hints::execute): Call the function to
9678         strip predictors.
9679         (strip_predict_hints): New function.
9680         * predict.def: Fix comment.
9682 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
9684         * Makefile.in: Clarify which tm.texi to copy over to assert the
9685         right to grant a GFDL license for all.
9687 2018-08-09  Jeff Law  <law@redhat.com>
9689         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
9690         unused variable.
9692 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
9694         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
9695         prototype.
9697 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9699         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
9700         reductions for variable-length vectors.
9702 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
9704         PR other/84889
9705         * common.opt (fdiagnostics-show-line-numbers): New option.
9706         * diagnostic-show-locus.c (class layout): Add fields
9707         "m_show_line_numbers_p" and "m_linenum_width";
9708         (num_digits): New function.
9709         (test_num_digits): New function.
9710         (layout::layout): Initialize new fields.  Update m_x_offset
9711         logic to handle any left margin.
9712         (layout::print_source_line): Print line number when requested.
9713         (layout::start_annotation_line): New member function.
9714         (layout::print_annotation_line): Call it.
9715         (layout::print_leading_fixits): Likewise.
9716         (layout::print_trailing_fixits): Likewise.  Update calls to
9717         move_to_column for new parameter.
9718         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
9719         it to potentially call start_annotation_line.
9720         (layout::show_ruler): Call start_annotation_line.
9721         (selftest::test_line_numbers_multiline_range): New selftest.
9722         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
9723         and selftest::test_line_numbers_multiline_range.
9724         * diagnostic.c (diagnostic_initialize): Initialize
9725         show_line_numbers_p.
9726         * diagnostic.h (struct diagnostic_context): Add field
9727         "show_line_numbers_p".
9728         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
9729         -fno-diagnostics-show-line-numbers.
9730         * dwarf2out.c (gen_producer_string): Add
9731         OPT_fdiagnostics_show_line_numbers to the ignored options.
9732         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
9733         one setting" options.
9734         (append_compiler_options): Likewise to the dropped options.
9735         (append_diag_options): Likewise to the passed-on options.
9736         * opts.c (common_handle_option): Handle the new option.
9737         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
9739 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9741         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
9742         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
9743         third argument of type "const signed char" to descriptions of
9744         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
9745         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
9746         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
9747         __builtin_bcdsub_ov functions.
9749 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9751         PR tree-optimization/86858
9752         * tree-vect-loop.c (vect_is_simple_reduction): Restore
9753         flow_bb_inside_loop_p calls.
9755 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
9757         PR tree-optimization/86871
9758         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
9759         instead of gimple_assign_lhs.
9761 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
9763         PR target/86887
9764         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
9765         register constraint to operand 0.
9766         (add<mode>3_carryinC): Likewise.
9767         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
9769 2018-08-09  Martin Liska  <mliska@suse.cz>
9771         PR c/86895
9772         * common.opt: Remove extra line.
9774 2018-08-09  Martin Liska  <mliska@suse.cz>
9776         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
9777         at the end of a line, make first letter capital and end up
9778         a sentence with a dot.
9779         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
9780         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
9781         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
9782         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
9783         (PARAM_MAX_ISL_OPERATIONS): Likewise.
9784         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
9785         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
9786         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
9787         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
9788         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
9789         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
9790         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
9791         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
9792         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
9793         (PARAM_TREE_REASSOC_WIDTH): Likewise.
9794         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
9795         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
9796         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
9798 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
9800         PR target/84332
9801         * config/s390/s390.c (s390_option_override_internal): Reduce the
9802         stack-clash-protection-probe-interval param if it would be too big
9803         for z900.
9805 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
9807         PR target/46179
9808         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
9809         * config/m68k/m68k.c (handle_move_double): Don't call
9810         m68k_final_prescan_insn.
9811         (m68k_adjust_decorated_operand): Renamed from
9812         m68k_final_prescan_insn, remove first and third operand and
9813         simplify.
9814         (print_operand): Call it.
9815         (print_operand_address): Call it.
9817 2018-08-08  Nathan Sidwell  <nathan@acm.org>
9819         * diagnostic.c (diagnostic_report_current_module): Use
9820         linemap_included_from & linemap_included_from_linemap.
9822 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
9824         * config/aarch64/aarch64-cores.def: Add phecda core.
9825         * config/aarch64/aarch64-tune.md: Regenerate.
9826         * doc/invoke.texi: Add phecda core.
9828 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9830         PR target/85295
9831         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
9832         definitions.
9833         * config/s390/s390.md ("movti"): Add more alternatives for
9834         constant to GPR copies.
9836 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
9838         * config/s390/s390.c: Fix whitespace damage throughout the file.
9839         * config/s390/s390.h: Likewise.
9840         * config/s390/tpf.h: Likewise.
9842 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9844         * config/s390/s390.c (s390_loadrelative_operand_p):
9845         Remove TARGET_CPU_ZARCH usages.
9846         (s390_rtx_costs): Likewise.
9847         (s390_legitimate_constant_p): Likewise.
9848         (s390_cannot_force_const_mem): Likewise.
9849         (legitimate_reload_constant_p): Likewise.
9850         (s390_preferred_reload_class): Likewise.
9851         (legitimize_pic_address): Likewise.
9852         (legitimize_tls_address): Likewise.
9853         (s390_split_branches): Removed.
9854         (s390_add_execute): Removed.
9855         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
9856         (s390_mainpool_start): Likewise.
9857         (s390_mainpool_finish): Likewise.
9858         (s390_mainpool_cancel): Removed.
9859         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
9860         (s390_chunkify_cancel): Likewise.
9861         (s390_return_addr_rtx): Likewise.
9862         (s390_register_info): Remove split_branches_pending_p uages.
9863         (s390_optimize_register_info): Likewise.
9864         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
9865         split_branches_pending_p usages.
9866         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
9867         (s390_load_got): Likewise.
9868         (s390_expand_split_stack_prologue): Likewise.
9869         (output_asm_nops): Likewise.
9870         (s390_function_profiler): Likewise.
9871         (s390_emit_call): Likewise.
9872         (s390_conditional_register_usage): Likewise.
9873         (s390_optimize_prologue): Likewise.
9874         (s390_reorg): Remove TARGET_CPU_ZARCH and
9875         split_branches_pending_p usages.
9876         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
9877         usages.
9878         (s390_output_indirect_thunk_function): Likewise.
9879         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
9880         (TARGET_CPU_ZARCH_P): Removed.
9881         (struct machine_function): Remove split_branches_pending_p.
9882         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
9884 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
9886         * common/config/s390/s390-common.c (processor_flags_table):
9887         Remove flags.
9888         * config.gcc: Remove with_arch/with_tune support.
9889         * config/s390/2064.md: Remove cpu attribute comparisons.
9890         * config/s390/driver-native.c (s390_host_detect_local_cpu):
9891         Remove MTN.
9892         * config/s390/linux.h (ASM_SPEC):
9893         Remove -march support.
9894         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
9895         Use a table to get an arch level.
9896         * config/s390/s390-opts.h (enum processor_type):
9897         Remove enum values.
9898         * config/s390/s390.c
9899         (processor_table): Remove entries, add arch_level values.
9900         (s390_issue_rate): Remove cases.
9901         (s390_option_override): Adjust
9902         s390_option_override_internal() call.
9903         (s390_option_override_internal): Remove deprecation warning.
9904         (s390_valid_target_attribute_tree): Adjust
9905         s390_option_override_internal() call.
9906         * config/s390/s390.h (struct s390_processor):
9907         Share with s390-c.c, add arch_level field.
9908         * config/s390/s390.md:
9909         Remove occurrences in cpu attribute.
9910         * config/s390/s390.opt: Remove -march/-mtune support.
9911         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
9912         * doc/invoke.texi: Remove deprecation warning.
9914 2018-08-08  Luis Machado  <luis.machado@linaro.org>
9916         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
9917         global.
9918         (qdf24xx_tunings): Set vector cost structure to
9919         qdf24xx_vector_cost.
9921         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
9922         <register_sextend>: Set to 3.
9924 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
9926         PR target/86838
9927         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
9928         * config/aarch64/aarch64-simd.md
9929         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
9930         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
9931         (aarch64_frecpx<mode>): ...this new pattern.
9932         * config/aarch64/aarch64-simd-builtins.def: Remove comment
9933         about aarch64_frecp<FRECP:frecp_suffix><mode>.
9935 2018-08-07  Martin Liska  <mliska@suse.cz>
9937         PR middle-end/83023
9938         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
9939         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
9940         * predict.def (PRED_MALLOC_NONNULL): New predictor.
9941         * doc/extend.texi: Document that malloc attribute adds
9942         hit to compiler.
9944 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
9946         PR target/86785
9947         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9948         Define to speculation_safe_value_not_needed.
9950 2018-08-06  Jeff Law  <law@redhat.com>
9952         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
9953         the vr_values instance to cprop_into_stmt.
9954         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
9955         (cprop_operand): Also query EVRP to determine if OP is a constant.
9957 2018-08-06  Nathan Sidwell  <nathan@acm.org>
9959         * diagnostic.c (diagnostic_report_current_module): Reroll
9960         included-at loop.  Translate text.
9962 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
9964         * function-tests.c (selftest::test_expansion_to_rtl): Call
9965         free_after_compilation.
9967 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9969         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
9971 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
9973         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
9974         loops with memory block operations from getting unrolled.
9976 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
9978         PR target/86807
9979         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
9980         Define to speculation_safe_value_not_needed.
9982 2018-08-06  Jeff Law  <law@redhat.com>
9984         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
9985         assert.
9987 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
9989         PR target/86662
9990         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
9991         with all enabled __intN types.
9993         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
9995 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
9997         * alias.c (record_set): Check for clobber high.
9998         * cfgexpand.c (expand_gimple_stmt): Likewise.
9999         * combine-stack-adj.c (single_set_for_csa): Likewise.
10000         * combine.c (find_single_use_1): Likewise.
10001         (set_nonzero_bits_and_sign_copies): Likewise.
10002         (get_combine_src_dest): Likewise.
10003         (is_parallel_of_n_reg_sets): Likewise.
10004         (try_combine): Likewise.
10005         (record_dead_and_set_regs_1): Likewise.
10006         (reg_dead_at_p_1): Likewise.
10007         (reg_dead_at_p): Likewise.
10008         * dce.c (deletable_insn_p): Likewise.
10009         (mark_nonreg_stores_1): Likewise.
10010         (mark_nonreg_stores_2): Likewise.
10011         * df-scan.c (df_find_hard_reg_defs): Likewise.
10012         (df_uses_record): Likewise.
10013         (df_get_call_refs): Likewise.
10014         * dwarf2out.c (mem_loc_descriptor): Likewise.
10015         * haifa-sched.c (haifa_classify_rtx): Likewise.
10016         * ira-build.c (create_insn_allocnos): Likewise.
10017         * ira-costs.c (scan_one_insn): Likewise.
10018         * ira.c (equiv_init_movable_p): Likewise.
10019         (rtx_moveable_p): Likewise.
10020         (interesting_dest_for_shprep): Likewise.
10021         * jump.c (mark_jump_label_1): Likewise.
10022         * postreload-gcse.c (record_opr_changes): Likewise.
10023         * postreload.c (reload_cse_simplify): Likewise.
10024         (struct reg_use): Add source expr.
10025         (reload_combine): Check for clobber high.
10026         (reload_combine_note_use): Likewise.
10027         (reload_cse_move2add): Likewise.
10028         (move2add_note_store): Likewise.
10029         * print-rtl.c (print_pattern): Likewise.
10030         * recog.c (decode_asm_operands): Likewise.
10031         (store_data_bypass_p): Likewise.
10032         (if_test_bypass_p): Likewise.
10033         * regcprop.c (kill_clobbered_value): Likewise.
10034         (kill_set_value): Likewise.
10035         * reginfo.c (reg_scan_mark_refs): Likewise.
10036         * reload1.c (maybe_fix_stack_asms): Likewise.
10037         (eliminate_regs_1): Likewise.
10038         (elimination_effects): Likewise.
10039         (mark_not_eliminable): Likewise.
10040         (scan_paradoxical_subregs): Likewise.
10041         (forget_old_reloads_1): Likewise.
10042         * reorg.c (find_end_label): Likewise.
10043         (try_merge_delay_insns): Likewise.
10044         (redundant_insn): Likewise.
10045         (own_thread_p): Likewise.
10046         (fill_simple_delay_slots): Likewise.
10047         (fill_slots_from_thread): Likewise.
10048         (dbr_schedule): Likewise.
10049         * resource.c (update_live_status): Likewise.
10050         (mark_referenced_resources): Likewise.
10051         (mark_set_resources): Likewise.
10052         * rtl.c (copy_rtx): Likewise.
10053         * rtlanal.c (reg_referenced_p): Likewise.
10054         (single_set_2): Likewise.
10055         (noop_move_p): Likewise.
10056         (note_stores): Likewise.
10057         * sched-deps.c (sched_analyze_reg): Likewise.
10058         (sched_analyze_insn): Likewise.
10060 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10062         * cse.c (invalidate_reg): New function extracted from...
10063         (invalidate): ...here.
10064         (canonicalize_insn): Check for clobber high.
10065         (invalidate_from_clobbers): invalidate clobber highs.
10066         (invalidate_from_sets_and_clobbers): Likewise.
10067         (count_reg_usage): Check for clobber high.
10068         (insn_live_p): Likewise.
10069         * cselib.c (cselib_expand_value_rtx_1):Likewise.
10070         (cselib_invalidate_regno): Check for clobber in setter.
10071         (cselib_invalidate_rtx): Pass through setter.
10072         (cselib_invalidate_rtx_note_stores):
10073         (cselib_process_insn): Check for clobber high.
10074         * cselib.h (cselib_invalidate_rtx): Add operand.
10076 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10078         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
10079         (mark_not_eliminable): Likewise.
10080         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
10081         * lra-lives.c (process_bb_lives): Check for clobber high.
10082         * lra.c (new_insn_reg): Remember clobber highs.
10083         (collect_non_operand_hard_regs): Check for clobber high.
10084         (lra_set_insn_recog_data): Likewise.
10085         (add_regs_to_insn_regno_info): Likewise.
10086         (lra_update_insn_regno_info): Likewise.
10088 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10090         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
10091         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
10093 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10095         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
10096         (copy_insn_1): Likewise.
10097         (gen_hard_reg_clobber_high): New gen function.
10098         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
10099         * genemit.c (gen_exp): Likewise.
10100         (gen_emit_seq): Pass through info.
10101         (gen_insn): Check for CLOBBER_HIGH.
10102         (gen_expand): Pass through info.
10103         (gen_split): Likewise.
10104         (output_add_clobbers): Likewise.
10105         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
10106         (remove_clobbers): Likewise.
10107         * rtl.h (gen_hard_reg_clobber_high): New declaration.
10109 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
10111         * doc/rtl.texi (clobber_high): Add.
10112         (parallel): Add in clobber high
10113         * rtl.c (rtl_check_failed_code3): Add function.
10114         * rtl.def (CLOBBER_HIGH): Add expression.
10115         * rtl.h (RTL_CHECKC3): Add macro.
10116         (rtl_check_failed_code3): Add declaration.
10117         (XC3EXP): Add macro.
10119 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
10121         PR target/86386
10122         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
10123         cfun->machine->max_used_stack_alignment if needed.
10125 2018-08-04  Martin Sebor  <msebor@redhat.com>
10127         PR tree-optimization/86571
10128         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
10129         NaN output to 4.
10131 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
10133         PR target/86799
10134         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10135         Define.
10137 2018-08-03  Jeff Law  <law@redhat.com>
10139         PR target/86795
10140         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10141         Define to speculation_safe_value_not_needed.
10143 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
10145         * doc/gcov.texi (-x): Remove duplicate "to".
10146         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
10147         (-Wif-not-aligned): Remove duplicate "is".
10148         (-flto): Remove duplicate "the".
10149         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
10150         duplicate "v5.00.b".
10151         (MSP430 Options): Remove duplicate "and" from the description
10152         of "-mgprel-sec=regexp".
10153         (x86 Options): Remove duplicate copies of "vmldLog102" and
10154         vmlsLog104 from description of "-mveclibabi=type".
10156 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
10158         * internal-fn.h (first_commutative_argument): Declare.
10159         * internal-fn.c (first_commutative_argument): New function.
10160         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
10161         restrictions for pattern statements.  Use first_commutative_argument
10162         to look for commutative operands in calls to internal functions.
10164 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
10166         * Makefile.in (wide-int-range.o): New.
10167         * tree-vrp.c: Move all the wide_int_* functions to...
10168         * wide-int-range.cc: ...here.
10169         * tree-vrp.h: Move all the wide_int_* prototypes to...
10170         * wide-int-range.h: ...here.
10172 2018-08-03  Tom de Vries  <tdevries@suse.de>
10174         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
10175         UI_NONE.
10176         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
10177         * except.c (output_function_exception_table): Do early exit if
10178         targetm_common.except_unwind_info (&global_options) == UI_NONE.
10180 2018-08-03  Martin Liska  <mliska@suse.cz>
10182         * predict.c (dump_prediction): Change to 2 digits
10183         in fraction part.
10185 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
10187         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
10188         neon_dup_q to...
10189         (falkor_am_1_gtov_gtov): ... a new insn reservation.
10191 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
10193         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
10194         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
10195         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
10196         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
10197         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
10199 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
10201         * diagnostic-show-locus.c (diagnostic_show_locus): Use
10202         pp_take_prefix when saving the existing prefix.
10203         * diagnostic.c (diagnostic_append_note): Likewise.
10204         * langhooks.c (lhd_print_error_function): Likewise.
10205         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
10206         param's type.  Free the existing prefix.
10207         (pp_take_prefix): New function.
10208         (pretty_printer::pretty_printer): Drop the prefix parameter.
10209         Rename the length parameter to match the comment.
10210         (pretty_printer::~pretty_printer): Free the prefix.
10211         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
10212         parameter.
10213         (struct pretty_printer): Drop the "const" from "prefix" field's
10214         type and clarify memory management.
10215         (pp_set_prefix): Drop the "const" from the 2nd param.
10216         (pp_take_prefix): New decl.
10218 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
10220         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
10221         for word_mode_ok here instead of passing as argument.
10222         (expand_block_compare): Change select_block_compare_mode() call.
10223         (expand_strncmp_gpr_sequence): New function.
10224         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
10226 2018-08-02  Jeff Law  <law@redhat.com>
10228         PR target/86790
10229         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10230         Define to speculation_safe_value_not_needed.
10232         PR target/86784
10233         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10234         Define to speculation_safe_value_not_needed.
10236 2018-08-02  Tom de Vries  <tdevries@suse.de>
10238         PR target/86660
10239         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
10240         function.  Return UI_TARGET unconditionally.
10241         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
10242         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
10244 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10246         * genemit.c (print_overload_test): Fix typo.
10248 2018-08-02  Richard Biener  <rguenther@suse.de>
10250         PR tree-optimization/86816
10251         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
10252         which checks for value availability before querying it.
10253         (gvn_uses_equal): Use it.
10254         (same_succ_hash): Likewise.
10255         (gimple_equal_p): Likewise.
10257 2018-08-02  Nick Clifton  <nickc@redhat.com>
10259         PR target/86813
10260         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10261         Define to speculation_safe_value_not_needed.
10263         PR target/86810
10264         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10265         Define to speculation_safe_value_not_needed.
10267         PR target/86810
10268         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10269         Define to speculation_safe_value_not_needed.
10271         PR target/86803
10272         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10273         Define to speculation_safe_value_not_needed.
10275         PR target/86797
10276         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10277         Define to speculation_safe_value_not_needed.
10279         PR target/86791
10280         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10281         Define to speculation_safe_value_not_needed.
10283         PR target/86789
10284         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10285         Define to speculation_safe_value_not_needed.
10287         PR target/86787
10288         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
10289         Define to speculation_safe_value_not_needed.
10291         PR target/86782
10292         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10293         speculation_safe_value_not_needed.
10295         PR target/86781
10296         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
10297         to speculation_safe_value_not_needed.
10299 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10301         * doc/md.texi: Expand the documentation of instruction names
10302         to mention port-local uses.  Document '@' in pattern names.
10303         * read-md.h (overloaded_instance, overloaded_name): New structs.
10304         (mapping): Declare.
10305         (md_reader::handle_overloaded_name): New member function.
10306         (md_reader::get_overloads): Likewise.
10307         (md_reader::m_first_overload): New member variable.
10308         (md_reader::m_next_overload_ptr): Likewise.
10309         (md_reader::m_overloads_htab): Likewise.
10310         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
10311         m_next_overload_ptr and m_overloads_htab.
10312         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
10313         (get_mode_token, get_code_token, get_int_token): New functions.
10314         (map_attr_string): Add an optional argument that passes back
10315         the associated iterator.
10316         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
10317         (md_reader::handle_overloaded_name, add_overload_instance): New
10318         functions.
10319         (apply_iterators): Handle '@' names.  Report an error if '@'
10320         is used without iterators.
10321         (initialize_iterators): Initialize the new iterator_group fields.
10322         * genopinit.c (handle_overloaded_code_for)
10323         (handle_overloaded_gen): New functions.
10324         (main): Use them to print declarations of maybe_code_for_* and
10325         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
10326         * genemit.c (print_overload_arguments, print_overload_test)
10327         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
10328         (main): Use it to print definitions of maybe_code_for_* and
10329         maybe_gen_* functions.
10330         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
10331         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
10332         instead of explicit mode checks.
10333         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
10334         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
10335         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
10336         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
10337         (aarch64_expand_compare_and_swap): Likewise
10338         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
10339         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
10340         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
10341         (aarch64_constant_pool_reload_icode): Delete.
10342         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
10343         instead of aarch64_constant_pool_reload_icode.  Use
10344         code_for_aarch64_reload_mov instead of explicit mode checks.
10345         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
10346         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
10347         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
10348         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
10349         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
10350         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
10351         (aarch64_atomic_load_op_code): Delete.
10352         (aarch64_emit_atomic_load_op): Likewise.
10353         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
10354         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
10355         instead of aarch64_emit_atomic_load_op.
10356         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
10357         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
10358         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
10359         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
10360         character before the pattern name.
10361         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
10362         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
10363         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
10364         (aarch64_frecps<mode>): Likewise.
10365         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
10366         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
10367         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
10368         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
10369         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
10371 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
10373         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
10374         Allow HFmode constants if TARGET_FP_F16INST.
10376 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
10378         PR target/86014
10379         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
10380         No longer check last store for clobber of address register.
10382 2018-08-02  Martin Liska  <mliska@suse.cz>
10384         PR gcov-profile/86817
10385         * gcov.c (process_all_functions): New function.
10386         (main): Call it.
10387         (process_file): Move functions processing to
10388         process_all_functions.
10390 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
10392         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
10393         "const" to the "gimple *" and "rtx_insn *" parameters.
10394         * dumpfile.h (dump_user_location_t::dump_user_location_t):
10395         Likewise.
10396         (dump_location_t::dump_location_t): Likewise.
10398 2018-08-01  Martin Sebor  <msebor@redhat.com>
10400         PR tree-optimization/86650
10401         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
10402         rather than a "gcall *".  Directly pass the data of interest
10403         to percent_K_format, rather than building a temporary CALL_EXPR
10404         to hold it.
10405         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
10406         (gimple_fold_builtin_strncat): Adjust.
10407         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
10408         gcall* argument with gimple*.
10409         * gimple-ssa-warn-restrict.c (check_call): Same.
10410         (wrestrict_dom_walker::before_dom_children): Same.
10411         (builtin_access::builtin_access): Same.
10412         (check_bounds_or_overlap): Same
10413         (maybe_diag_overlap): Same.
10414         (maybe_diag_offset_bounds): Same.
10415         * tree-diagnostic.c (default_tree_printer): Move usage of
10416         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
10417         to this callsite.
10418         * tree-pretty-print.c (percent_K_format): Add argument.
10419         * tree-pretty-print.h: Add argument.
10420         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
10421         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
10422         (maybe_diag_stxncpy_trunc): Same.
10423         (handle_builtin_stxncpy): Same.
10424         (handle_builtin_strcat): Same.
10426 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10428         * match.pd: Optimise pointer range checks.
10430 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10432         PR tree-optimization/86758
10433         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
10434         to remove pattern statements.
10436 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10438         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
10439         result of dfs_enumerate_from when constructing stmt_vec_infos,
10440         instead of additionally calling get_loop_body.
10442 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10444         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
10445         parameter.
10446         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
10447         When creating an iv, assert that the step is not known to be zero.
10448         (vect_setup_realignment): Update call accordingly.
10449         * tree-vect-stmts.c (vectorizable_store): Likewise.
10450         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
10452 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10454         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
10455         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
10456         (vectorizable_reduction): Likewise.
10457         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
10458         (vect_detect_hybrid_slp_stmts): Likewise.
10459         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
10461 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
10463         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
10464         (wide_int_set_zero_nonzero_bits): ...this.
10465         (zero_nonzero_bits_from_vr): Rename to...
10466         (vrp_set_zero_nonzero_bits): ...this.
10467         (extract_range_from_multiplicative_op_1): Abstract wide int
10468         code...
10469         (wide_int_range_multiplicative_op): ...here.
10470         (extract_range_from_binary_expr_1): Extract wide int binary
10471         operations into their own functions.
10472         (wide_int_range_lshift): New.
10473         (wide_int_range_can_optimize_bit_op): New.
10474         (wide_int_range_shift_undefined_p): New.
10475         (wide_int_range_bit_xor): New.
10476         (wide_int_range_bit_ior): New.
10477         (wide_int_range_bit_and): New.
10478         (wide_int_range_trunc_mod): New.
10479         (extract_range_into_wide_ints): New.
10480         (vrp_shift_undefined_p): New.
10481         (extract_range_from_multiplicative_op): New.
10482         (vrp_can_optimize_bit_op): New.
10483         * tree-vrp.h (value_range::dump): New.
10484         (wide_int_range_multiplicative_op): New.
10485         (wide_int_range_lshift):New.
10486         (wide_int_range_shift_undefined_p): New.
10487         (wide_int_range_bit_xor): New.
10488         (wide_int_range_bit_ior): New.
10489         (wide_int_range_bit_and): New.
10490         (wide_int_range_trunc_mod): New.
10491         (zero_nonzero_bits_from_bounds): Rename to...
10492         (wide_int_set_zero_nonzero_bits): ...this.
10493         (zero_nonzero_bits_from_vr): Rename to...
10494         (vrp_set_zero_nonzero_bits): ...this.
10495         (range_easy_mask_min_max): Rename to...
10496         (wide_int_range_can_optimize_bit_op): this.
10497         * vr-values.c (simplify_bit_ops_using_ranges): Rename
10498         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
10500 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10502         * tree-vectorizer.h (vect_orig_stmt): New function.
10503         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
10504         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
10505         (vect_create_epilog_for_reduction): Likewise.
10506         (vectorizable_live_operation): Likewise.
10507         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
10508         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
10509         * tree-vect-stmts.c (vectorizable_call): Likewise.
10510         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
10512 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10514         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
10515         argument.
10516         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
10517         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
10518         (vect_transform_loop): Likewise.
10519         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10521 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10523         * tree-vectorizer.h (vect_schedule_slp): Return void.
10524         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
10525         (vect_schedule_slp): Likewise.
10527 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10529         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
10530         argument.
10531         (vect_transform_loop): Update calls accordingly.  Schedule SLP
10532         instances before the main loop, if any exist.
10534 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
10536         PR tree-optimization/86749
10537         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
10538         If the lhs is used in a COND_EXPR, check that it is being used
10539         as the "then" or "else" value.
10541 2018-08-01  Tom de Vries  <tdevries@suse.de>
10543         PR target/86800
10544         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
10545         speculation_safe_value_not_needed.
10547 2018-08-01  Richard Biener  <rguenther@suse.de>
10549         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
10550         as base and offset.
10552 2018-08-01  Martin Liska  <mliska@suse.cz>
10554         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
10555         format how successful transformation is dumped.
10556         (gimple_mod_pow2_value_transform): Likewise.
10557         (gimple_mod_subtract_transform): Likewise.
10558         (gimple_stringops_transform): Likewise.
10560 2018-08-01  Martin Liska  <mliska@suse.cz>
10562         PR value-prof/35543
10563         * value-prof.c (interesting_stringop_to_profile_p):
10564         Simplify the code and add BUILT_IN_MEMMOVE.
10565         (gimple_stringops_transform): Likewise.
10567 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
10569         * config/aarch64/aarch64-simd.md
10570         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
10571         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
10572         use GPI iterator instead of SI mode.
10574 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10576         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
10577         rs6000_speculation_barrier.
10578         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
10579         new barrier pattern name.
10581 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
10583         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
10584         (speculation_barrier): New insn.
10586 2018-08-01  Richard Biener  <rguenther@suse.de>
10588         PR bootstrap/86724
10589         * graphite.h: Include isl/id.h and isl/space.h to allow build
10590         with ISL 0.20.
10592 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
10594         PR target/86651
10595         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
10596         mode for COFF targets.
10597         * defaults.h (TARGET_COFF): Define.
10598         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
10599         TARGET_COFF): Define.
10600         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
10601         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
10602         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
10604 2018-07-31  Alexandre Oliva <oliva@adacore.com>
10606         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
10607         * gimple-streamer-out.c (output_bb): Save it.
10608         * lto-streamer-in.c (input_struct_function_base): Restore
10609         instance discriminator if available.  Create map on demand.
10610         * lto-streamer-out.c (output_struct_function_base): Save it if
10611         available.
10612         * final.c (decl_to_instance_map): Document LTO strategy.
10614 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
10615             Olivier Hainque  <hainque@adacore.com>
10617         * debug.h (decl_to_instance_map_t): New type.
10618         (decl_to_instance_map): Declare.
10619         (maybe_create_decl_to_instance_map): New inline function.
10620         * final.c (bb_discriminator, last_bb_discriminator): New statics,
10621         to track basic block discriminators.
10622         (final_start_function_1): Initialize them.
10623         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
10624         bb_discriminator.
10625         (decl_to_instance_map): New variable.
10626         (map_decl_to_instance, maybe_set_discriminator): New functions.
10627         (notice_source_line): Set discriminator.
10629 2018-07-31  Ian Lance Taylor  <iant@golang.org>
10631         * targhooks.c (default_have_speculation_safe_value): Add
10632         ATTRIBUTE_UNUSED.
10634 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10636         * dump-context.h: Include "pretty-print.h".
10637         (dump_context::refresh_dumps_are_enabled): New decl.
10638         (dump_context::emit_item): New decl.
10639         (class dump_context): Add fields "m_test_pp" and
10640         "m_test_pp_flags".
10641         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10642         (temp_dump_context::get_dumped_text): New decl.
10643         (class temp_dump_context): Add field "m_pp".
10644         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
10645         (dump_context::refresh_dumps_are_enabled): ...and add a test for
10646         m_test_pp.
10647         (set_dump_file): Update for above change.
10648         (set_alt_dump_file): Likewise.
10649         (dump_loc): New overload, taking a pretty_printer *.
10650         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
10651         to any test pretty-printer.
10652         (make_item_for_dump_gimple_stmt): New function, adapted from
10653         optinfo::add_gimple_stmt.
10654         (dump_context::dump_gimple_stmt): Call it, and use the result,
10655         eliminating the direct usage of dump_file and alt_dump_file in
10656         favor of indirectly using them via emit_item.
10657         (make_item_for_dump_gimple_expr): New function, adapted from
10658         optinfo::add_gimple_expr.
10659         (dump_context::dump_gimple_expr): Call it, and use the result,
10660         eliminating the direct usage of dump_file and alt_dump_file in
10661         favor of indirectly using them via emit_item.
10662         (make_item_for_dump_generic_expr): New function, adapted from
10663         optinfo::add_tree.
10664         (dump_context::dump_generic_expr): Call it, and use the result,
10665         eliminating the direct usage of dump_file and alt_dump_file in
10666         favor of indirectly using them via emit_item.
10667         (make_item_for_dump_printf_va): New function, adapted from
10668         optinfo::add_printf_va.
10669         (make_item_for_dump_printf): New function.
10670         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
10671         and use the result, eliminating the direct usage of dump_file and
10672         alt_dump_file in favor of indirectly using them via emit_item.
10673         (make_item_for_dump_dec): New function.
10674         (dump_context::dump_dec): Call it, and use the result,
10675         eliminating the direct usage of dump_file and alt_dump_file in
10676         favor of indirectly using them via emit_item.
10677         (make_item_for_dump_symtab_node): New function, adapted from
10678         optinfo::add_symtab_node.
10679         (dump_context::dump_symtab_node): Call it, and use the result,
10680         eliminating the direct usage of dump_file and alt_dump_file in
10681         favor of indirectly using them via emit_item.
10682         (dump_context::begin_scope): Reimplement, avoiding direct usage
10683         of dump_file and alt_dump_file in favor of indirectly using them
10684         via emit_item.
10685         (dump_context::emit_item): New member function.
10686         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
10687         Set up test pretty-printer on the underlying context.  Call
10688         refresh_dumps_are_enabled.
10689         (temp_dump_context::~temp_dump_context): Call
10690         refresh_dumps_are_enabled.
10691         (temp_dump_context::get_dumped_text): New member function.
10692         (selftest::verify_dumped_text): New function.
10693         (ASSERT_DUMPED_TEXT_EQ): New macro.
10694         (selftest::test_capture_of_dump_calls): Run all tests twice, with
10695         and then without optinfo enabled.  Add uses of
10696         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
10697         * dumpfile.h: Update comment for the dump_* API.
10698         * optinfo-emit-json.cc
10699         (selftest::test_building_json_from_dump_calls): Update for new
10700         param for temp_dump_context ctor.
10701         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
10702         and "m_owned" field.
10703         (optinfo_item::~optinfo_item): Likewise.
10704         (optinfo::add_item): New member function.
10705         (optinfo::emit): Update comment.
10706         (optinfo::add_string): Delete.
10707         (optinfo::add_printf): Delete.
10708         (optinfo::add_printf_va): Delete.
10709         (optinfo::add_gimple_stmt): Delete.
10710         (optinfo::add_gimple_expr): Delete.
10711         (optinfo::add_tree): Delete.
10712         (optinfo::add_symtab_node): Delete.
10713         (optinfo::add_dec): Delete.
10714         * optinfo.h (class dump_context): New forward decl.
10715         (optinfo::add_item): New decl.
10716         (optinfo::add_string): Delete.
10717         (optinfo::add_printf): Delete.
10718         (optinfo::add_printf_va): Delete.
10719         (optinfo::add_gimple_stmt): Delete.
10720         (optinfo::add_gimple_expr): Delete.
10721         (optinfo::add_tree): Delete.
10722         (optinfo::add_symtab_node): Delete.
10723         (optinfo::add_dec): Delete.
10724         (optinfo::add_poly_int): Delete.
10725         (optinfo_item::optinfo_item): Remove "owned" param.
10726         (class optinfo_item): Remove field "m_owned".
10728 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10730         PR middle-end/86705
10731         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
10732         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
10733         requested variable alignment.
10734         (expand_one_ssa_partition): Likewise.
10735         (expand_one_var): Likewise.
10737 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10739         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
10740         to speculation_safe_value_not_needed.
10742 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10744         * targhooks.h (speculation_safe_value_not_needed): New prototype.
10745         * targhooks.c (speculation_safe_value_not_needed): New function.
10746         * target.def (have_speculation_safe_value): Update documentation.
10747         * doc/tm.texi: Regenerated.
10749 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10751         * config/aarch64/iterators.md (ALLI_TI): New iterator.
10752         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
10753         expand.
10754         (despeculate_copy<ALLI:mode>_insn): New insn.
10755         (despeculate_copyti_insn): New insn.
10756         (despeculate_simple<ALLI:mode>): New insn
10757         (despeculate_simpleti): New insn.
10758         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
10759         function.
10760         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
10761         aarch64_speculation_safe_value.
10762         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
10764 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10766         * config/aarch64/aarch64-speculation.cc: New file.
10767         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
10768         before pass_reorder_blocks.
10769         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
10770         prototype.
10771         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
10772         X14 and X15 when tracking speculation.
10773         * config/aarch64/aarch64.md (register name constants): Add
10774         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
10775         (unspec): Add UNSPEC_SPECULATION_TRACKER.
10776         (speculation_barrier): New insn attribute.
10777         (cmp<mode>): Allow SP in comparisons.
10778         (speculation_tracker): New insn.
10779         (speculation_barrier): Add speculation_barrier attribute.
10780         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
10781         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
10782         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
10784 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10786         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
10787         aarch64_track_speculation is true.
10788         (tb<optab><mode>1): Likewise.
10789         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
10790         generate CB[N]Z when tracking speculation.
10791         (aarch64_split_compare_and_swap): Likewise.
10792         (aarch64_split_atomic_op): Likewise.
10794 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10796         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
10798 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10800         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
10801         (speculation_barrier): New insn.
10803 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10805         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
10806         * config/arm/arm.md (speculation_barrier): New expand.
10807         (speculation_barrier_insn): New pattern.
10809 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
10811         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
10812         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
10813         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
10814         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
10815         list.
10816         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
10817         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
10818         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
10819         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
10820         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
10821         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
10822         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
10823         * builtins.c (expand_speculation_safe_value): New function.
10824         (expand_builtin): Call it.
10825         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
10826         * doc/extend.texi: Document __builtin_speculation_safe_value.
10827         * doc/md.texi: Document "speculation_barrier" pattern.
10828         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
10829         TARGET_HAVE_SPECULATION_SAFE_VALUE.
10830         * doc/tm.texi: Regenerated.
10831         * target.def (have_speculation_safe_value, speculation_safe_value): New
10832         hooks.
10833         * targhooks.c (default_have_speculation_safe_value): New function.
10834         (default_speculation_safe_value): New function.
10835         * targhooks.h (default_have_speculation_safe_value): Add prototype.
10836         (default_speculation_safe_value): Add prototype.
10838 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
10840         * dump-context.h (dump_context::dump_loc): New decl.
10841         * dumpfile.c (dump_context::dump_loc): New member function.
10842         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
10843         and dump_gimple_stmt.
10844         (dump_context::dump_gimple_expr_loc): Likewise, using
10845         dump_gimple_expr.
10846         (dump_context::dump_generic_expr_loc): Likewise, using
10847         dump_generic_expr.
10848         (dump_context::dump_printf_loc_va): Likewise, using
10849         dump_printf_va.
10850         (dump_context::begin_scope): Explicitly using the global function
10851         "dump_loc", rather than the member function.
10853 2018-07-31  Martin Sebor  <msebor@redhat.com>
10855         PR tree-optimization/86741
10856         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
10858 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
10860         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
10861         depend on whether prefetch instructions will be emitted or not.
10862         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
10863         will be emitted or not.
10864         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
10865         (TARGET_SETMEM_PFD): New macros.
10867 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10869         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
10870         (NULL_STMT_VEC_INFO): Delete.
10871         (stmt_vec_info::operator*): Likewise.
10872         (stmt_vec_info::operator gimple *): Likewise.
10873         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
10874         of NULL_STMT_VEC_INFO.
10875         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
10876         (vect_reassociating_reduction_p): Likewise.
10877         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
10878         (vectorizable_store): Likewise.
10879         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
10880         (vec_info::free_stmt_vec_infos): Likewise.
10882 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10884         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
10885         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
10887 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10889         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
10890         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10891         (vec_info::free_stmt_vec_info): New private member functions.
10892         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
10893         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
10894         * tree-parloops.c (gather_scalar_reductions): Remove calls to
10895         set_stmt_vec_info_vec and free_stmt_vec_infos.
10896         * tree-vect-loop.c (_loop_vec_info): Remove call to
10897         set_stmt_vec_info_vec.
10898         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
10899         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
10900         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
10901         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
10902         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
10903         assignments in {vec_info::,}new_stmt_vec_info that are redundant
10904         with the clearing in the xcalloc.
10905         (stmt_vec_info_vec): Delete.
10906         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
10907         (vectorize_loops): Likewise.
10908         (vec_info::~vec_info): Remove argument from call to
10909         free_stmt_vec_infos.
10910         (vec_info::add_stmt): Remove vinfo argument from call to
10911         new_stmt_vec_info.
10913 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10915         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
10916         rather than a gimple stmt.
10917         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
10918         information for pattern statements when passed the original
10919         statement; instead wait to be passed the pattern statement itself.
10920         Don't call set_vinfo_for_stmt here.
10921         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
10922         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
10923         stmt_vec_infos here.
10924         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
10925         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
10926         stmt_vec_infos entry.
10928 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10930         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
10931         * tree-vectorizer.c (vec_info::replace_stmt): New function.
10932         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
10933         * tree-vect-stmts.c (vectorizable_call): Likewise.
10934         (vectorizable_simd_clone_call): Likewise.
10936 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10938         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
10939         * tree-vectorizer.c (vec_info::remove_stmt): New function.
10940         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
10941         * tree-vect-loop.c (vect_transform_loop): Likewise.
10942         * tree-vect-slp.c (vect_schedule_slp): Likewise.
10943         * tree-vect-stmts.c (vect_remove_stores): Likewise.
10945 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10947         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
10948         (vect_dr_stmt): Delete.
10949         * tree-vectorizer.c (vec_info::lookup_dr): New function.
10950         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
10951         of DR_VECT_AUX.
10952         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
10953         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
10954         (vect_verify_datarefs_alignment, vect_peeling_supportable)
10955         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
10956         (vect_analyze_data_refs): Likewise.
10957         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
10958         argument.
10959         (vect_find_same_alignment_drs): Likewise.
10960         (vect_slp_analyze_node_dependences): Update calls accordingly.
10961         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10962         instead of DR_VECT_AUX.
10963         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
10964         of a vector data references.  Use vec_info::lookup_dr instead of
10965         DR_VECT_AUX.
10966         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
10967         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
10968         instead of DR_VECT_AUX.
10970 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10972         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
10973         dr_vec_info.
10974         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
10975         accordingly.
10976         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
10977         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
10978         (vect_gen_prolog_loop_niters): Likewise.
10980 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
10982         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
10983         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
10984         (vect_known_alignment_in_bytes, vect_dr_behavior)
10985         (vect_get_scalar_dr_size): Take references as dr_vec_infos
10986         instead of data_references.  Update calls to other routines for
10987         which the same change has been made.
10988         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
10989         dr_vec_infos instead of stmt_vec_infos.
10990         (vect_analyze_data_ref_dependence): Update call accordingly.
10991         (vect_slp_analyze_data_ref_dependence)
10992         (vect_record_base_alignments): Use DR_VECT_AUX.
10993         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
10994         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
10995         (vector_alignment_reachable_p, vect_get_data_access_cost)
10996         (vect_peeling_supportable, vect_analyze_group_access_1)
10997         (vect_analyze_group_access, vect_analyze_data_ref_access)
10998         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
10999         (vect_compile_time_alias, vect_small_gap_p)
11000         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
11001         (vect_supportable_dr_alignment): Take references as dr_vec_infos
11002         instead of data_references.  Update calls to other routines for
11003         which the same change has been made.
11004         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
11005         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
11006         (vect_slp_analyze_and_verify_node_alignment)
11007         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
11008         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
11009         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
11010         above changes.
11011         (_vect_peel_info::dr): Replace with...
11012         (_vect_peel_info::dr_info): ...this new field.
11013         (vect_peeling_hash_get_most_frequent)
11014         (vect_peeling_hash_choose_best_peeling): Update accordingly.
11015         (vect_peeling_hash_get_lowest_cost):
11016         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
11017         routines for which the same change has been made.
11018         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
11019         data_reference.
11020         * tree-vect-loop-manip.c (get_misalign_in_elems)
11021         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
11022         above changes.
11023         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11024         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
11025         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
11026         (get_group_load_store_type, get_negative_load_store_type)
11027         (vect_get_data_ptr_increment, vectorizable_store)
11028         (vectorizable_load): Likewise.
11029         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
11030         Update calls to other routines for which the same change has been made.
11032 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11034         * tree-vectorizer.h (vec_info::move_dr): New member function.
11035         (dataref_aux): Rename to...
11036         (dr_vec_info): ...this and add "dr" and "stmt" fields.
11037         (_stmt_vec_info::dr_aux): Update accordingly.
11038         (_stmt_vec_info::data_ref_info): Delete.
11039         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
11040         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
11041         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
11042         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
11043         of data_ref.
11044         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
11045         (STMT_VINFO_DR_INFO): New macro.
11046         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
11047         (set_dr_misalignment): Update after rename of dataref_aux.
11048         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
11049         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
11050         initialization of STMT_VINFO_DATA_REF.
11051         * tree-vectorizer.c (vec_info::move_dr): New function.
11052         * tree-vect-patterns.c (vect_recog_bool_pattern)
11053         (vect_recog_mask_conversion_pattern)
11054         (vect_recog_gather_scatter_pattern): Use it.
11055         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
11056         the "dr" and "stmt" fields of dr_vec_info instead of
11057         STMT_VINFO_DATA_REF.
11059 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11061         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
11062         (is_pattern_stmt_p): Use it.
11063         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
11064         on pattern statements.
11066 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11068         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
11069         original stmt as a stmt_vec_info rather than a gimple stmt.
11070         (vect_pattern_recog_1): Take the statement directly as a
11071         stmt_vec_info, rather than via a gimple_stmt_iterator.
11072         Update call to vect_mark_pattern_stmts.
11073         (vect_pattern_recog): Update calls accordingly.
11075 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11077         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
11078         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
11079         a vect_def_type for the first argument.
11080         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
11081         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
11082         operand if it isn't defined by a vectorized statement.
11083         (vect_build_gather_load_calls): Remove the mask_dt argument and
11084         update calls to vect_get_vec_def_for_stmt_copy.
11085         (vectorizable_bswap): Likewise the dt argument.
11086         (vectorizable_call): Update calls to vectorizable_bswap and
11087         vect_get_vec_def_for_stmt_copy.
11088         (vectorizable_simd_clone_call, vectorizable_assignment)
11089         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
11090         (vectorizable_comparison): Update calls to
11091         vect_get_vec_def_for_stmt_copy.
11092         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
11093         vect_is_simple_use.
11094         (vect_get_loop_based_defs): Remove dt argument and update call
11095         to vect_get_vec_def_for_stmt_copy.
11096         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
11097         and vect_get_vec_def_for_stmt_copy.
11098         (vectorizable_load): Update calls to vect_build_gather_load_calls
11099         and vect_get_vec_def_for_stmt_copy.
11100         * tree-vect-loop.c (vect_create_epilog_for_reduction)
11101         (vectorizable_reduction, vectorizable_live_operation): Update calls
11102         to vect_get_vec_def_for_stmt_copy.
11104 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11106         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
11107         and gimple stmt arguments with a stmt_vec_info.
11108         (vect_record_base_alignments): Update calls accordingly.
11109         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
11110         and gimple stmt arguments with a stmt_vec_info.
11111         (vect_build_slp_tree_1): Remove vinfo argument and update call
11112         to vect_record_max_nunits.
11113         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
11114         and vect_record_max_nunits.
11116 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11118         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
11119         file and take a stmt_vec_info instead of a gimple stmt.
11120         (supportable_widening_operation, vect_finish_replace_stmt)
11121         (vect_finish_stmt_generation, vect_get_store_rhs)
11122         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
11123         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
11124         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
11125         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
11126         (vect_create_data_ref_ptr, bump_vector_ptr)
11127         (vect_permute_store_chain, vect_setup_realignment)
11128         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
11129         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
11130         (vectorizable_reduction, vectorizable_induction)
11131         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
11132         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
11133         than gimple stmts as arguments.
11134         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
11135         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
11136         (can_group_stmts_p, vect_check_gather_scatter)
11137         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
11138         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
11139         (vect_permute_load_chain, vect_shift_permute_load_chain)
11140         (vect_transform_grouped_load)
11141         (vect_record_grouped_load_vectors): Likewise.
11142         * tree-vect-loop.c (vect_fixup_reduc_chain)
11143         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
11144         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
11145         (vectorizable_reduction, vectorizable_induction)
11146         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
11147         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
11148         (vect_get_load_store_mask): Likewise.
11149         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
11150         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
11151         * tree-vect-stmts.c (vect_mark_relevant)
11152         (is_simple_and_all_uses_invariant)
11153         (exist_non_indexing_operands_for_use_p, process_use)
11154         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
11155         (vect_get_vec_def_for_operand, vect_get_vec_defs)
11156         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
11157         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
11158         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
11159         (get_negative_load_store_type, get_load_store_type)
11160         (vect_check_load_store_mask, vect_check_store_rhs)
11161         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
11162         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11163         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11164         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11165         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
11166         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
11167         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
11168         (supportable_widening_operation): Likewise.
11170 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11172         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
11173         a stmt_vec_info instead of a gcall.
11174         (vect_check_gather_scatter): Update call accordingly.
11175         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
11176         of a gphi.
11177         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
11178         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
11179         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
11180         instead of a gimple stmt.
11181         (vect_transform_loop): Update calls accordingly.
11182         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
11183         stmt_vec_infos instead of gimple stmts.
11184         (vect_analyze_slp_instance): Update use accordingly.
11185         * tree-vect-stmts.c (read_vector_array, write_vector_array)
11186         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
11187         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
11188         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
11189         (vect_gen_widened_results_half, vect_get_loop_based_defs)
11190         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
11191         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
11192         down to subroutines.
11194 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11196         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
11197         of the worklist from a vector of gimple stmts to a vector of
11198         stmt_vec_infos.
11199         * tree-vect-stmts.c (vect_mark_relevant, process_use)
11200         (vect_mark_stmts_to_be_vectorized): Likewise
11202 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11204         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
11205         statement before passing it to vect_analyze_stmt.
11206         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
11207         the chain of phi vector definitions.  Track the exit phi via its
11208         stmt_vec_info.
11209         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
11210         STMT_VINFO_REDUC_DEF.
11211         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
11212         stmt_vec_infos to handle the statement chains.
11213         (vect_get_slp_defs): Record the first statement in the node
11214         using a stmt_vec_info.
11215         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
11216         statements here and pass their stmt_vec_info down to subroutines.
11217         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
11218         down to vect_finish_stmt_generation.
11219         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
11220         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
11221         stmt_vec_infos to subroutines.
11222         (vect_remove_stores): Use stmt_vec_infos to handle the statement
11223         chains.
11225 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11227         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
11228         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
11229         (vect_permute_store_chain, vect_setup_realignment)
11230         (vect_permute_load_chain, vect_shift_permute_load_chain)
11231         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
11232         stmts internally, and when passing values to other vectorizer routines.
11233         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11234         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
11235         (vect_analyze_loop_operations, get_initial_def_for_reduction)
11236         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
11237         (vectorizable_reduction, vectorizable_induction)
11238         (vectorizable_live_operation, vect_transform_loop_stmt)
11239         (vect_transform_loop): Likewise.
11240         * tree-vect-patterns.c (vect_reassociating_reduction_p)
11241         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
11242         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
11243         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11244         (vect_slp_analyze_node_operations_1): Likewise.
11245         * tree-vect-stmts.c (vect_mark_relevant, process_use)
11246         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
11247         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
11248         (vect_finish_stmt_generation_1, get_group_load_store_type)
11249         (get_load_store_type, vect_build_gather_load_calls)
11250         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11251         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11252         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11253         (vectorizable_store, vectorizable_load, vectorizable_condition)
11254         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
11255         (supportable_widening_operation): Likewise.
11256         (vect_get_vector_types_for_stmt): Likewise.
11257         * tree-vectorizer.h (vect_dr_behavior): Likewise.
11259 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11261         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11262         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
11263         (vect_permute_store_chain, vect_permute_load_chain)
11264         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
11265         repeated stmt_vec_info lookups.
11266         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
11267         (vect_update_ivs_after_vectorizer): Likewise.
11268         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
11269         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11270         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
11271         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11272         (vect_bb_slp_scalar_cost): Likewise.
11273         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
11275 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11277         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
11278         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
11279         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
11280         to get gassigns and gcalls, rather than operating on generc gimple
11281         stmts.
11282         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
11283         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
11284         (vectorizable_load, vect_analyze_stmt): Likewise.
11285         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
11287 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11289         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
11290         return stmt_vec_infos rather than gimple stmts.  Do not accept
11291         null arguments.
11292         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
11293         of a gimple stmt.
11294         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
11295         Update use of get_later_stmt.
11296         (vect_get_constant_vectors): Update call accordingly.
11297         (vect_schedule_slp_instance): Likewise
11298         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
11299         (vect_slp_analyze_instance_dependence): Likewise.
11300         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
11302 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11304         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
11305         (stmt_info_for_cost::stmt_info): ...this new field.
11306         (add_stmt_costs): Update accordingly.
11307         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11308         (vect_get_known_peeling_cost): Likewise.
11309         (vect_estimate_min_profitable_iters): Likewise.
11310         * tree-vect-stmts.c (record_stmt_cost): Likewise.
11312 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11314         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
11315         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11316         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
11317         accordingly.
11318         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
11320 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11322         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
11323         a gimple stmt to a stmt_vec_info.
11324         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
11326 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11328         * tree-vectorizer.h (vec_info::grouped_stores): Change from
11329         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11330         (_loop_vec_info::reduction_chains): Likewise.
11331         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
11332         accordingly.
11333         * tree-vect-slp.c (vect_analyze_slp): Likewise.
11335 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11337         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
11338         a gimple stmt to a stmt_vec_info.
11339         (_stmt_vec_info::next_element): Likewise.
11340         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
11341         (vect_slp_analyze_and_verify_node_alignment)
11342         (vect_analyze_group_access_1, vect_analyze_group_access)
11343         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
11344         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
11345         (vect_supportable_dr_alignment): Update accordingly.
11346         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
11347         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
11348         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
11349         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
11350         * tree-vect-slp.c (vect_build_slp_tree_1)
11351         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
11352         (vect_split_slp_store_group, vect_analyze_slp_instance)
11353         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
11354         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
11355         (get_group_load_store_type, get_load_store_type)
11356         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
11357         (vect_transform_stmt, vect_remove_stores): Likewise.
11359 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11361         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
11362         than a gimple stmt.
11363         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
11364         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
11365         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
11366         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
11367         (vector_alignment_reachable_p, vect_get_data_access_cost)
11368         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
11369         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
11370         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
11371         (vect_analyze_group_access_1, vect_analyze_group_access)
11372         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
11373         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
11374         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
11375         result of vect_dr_stmt and use the stmt_vec_info instead of
11376         the associated gimple stmt.
11377         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
11378         (vect_gen_prolog_loop_niters): Likewise.
11379         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
11381 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11383         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
11384         to a vec<stmt_vec_info>.
11385         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
11386         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
11387         vec<stmt_vec_info>.
11388         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
11389         to a vec<stmt_vec_info>.
11390         (bst_traits::value_type, bst_traits::value_type): Likewise.
11391         (bst_traits::hash): Update accordingly.
11392         (vect_get_and_check_slp_defs): Change the stmts parameter from
11393         a vec<gimple *> to a vec<stmt_vec_info>.
11394         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
11395         (vect_build_slp_tree): Likewise.
11396         (vect_build_slp_tree_2): Likewise.  Update uses of
11397         SLP_TREE_SCALAR_STMTS.
11398         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
11399         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
11400         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
11401         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
11402         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
11403         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
11404         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
11405         (vect_get_constant_vectors, vect_get_slp_defs)
11406         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
11407         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
11408         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
11409         instead of gimple stmts.
11410         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
11411         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
11412         (vect_slp_analyze_instance_dependence): Update uses of
11413         SLP_TREE_SCALAR_STMTS.
11414         (vect_slp_analyze_and_verify_node_alignment): Likewise.
11415         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
11416         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
11417         (get_initial_defs_for_reduction): Likewise.
11418         (vect_create_epilog_for_reduction): Likewise.
11419         (vectorize_fold_left_reduction): Likewise.
11420         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
11421         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
11422         (can_vectorize_live_stmts): Likewise.
11424 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11426         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
11427         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
11428         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
11429         than gimple stmts.
11430         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
11431         of a gimple stmt.
11432         (gather_scalar_reductions): Update after above interface changes.
11433         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
11434         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
11435         than gimple stmts.
11436         (vect_force_simple_reduction): Likewise.
11437         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
11438         LOOP_VINFO_REDUCTIONS.
11439         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
11441 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11443         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
11444         a gimple stmt to a stmt_vec_info.
11445         * tree-vect-loop.c (vect_active_double_reduction_p)
11446         (vect_force_simple_reduction, vectorizable_reduction): Update
11447         accordingly.
11449 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11451         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
11452         vec<gimple *> to a vec<stmt_vec_info>.
11453         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
11454         the reduction_phis argument from a vec<gimple *> to a
11455         vec<stmt_vec_info>.
11456         (vectorizable_reduction): Likewise the phis local variable that
11457         is passed to vect_create_epilog_for_reduction.  Update for new type
11458         of SLP_TREE_VEC_STMTS.
11459         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
11460         (vectorizable_live_operation): Likewise.
11461         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
11462         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
11464 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11466         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
11467         a gimple stmt to a stmt_vec_info.
11468         (vectorizable_condition, vectorizable_live_operation)
11469         (vectorizable_reduction, vectorizable_induction): Pass back the
11470         vectorized statement as a stmt_vec_info.
11471         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
11472         use of STMT_VINFO_VEC_STMT.
11473         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
11474         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
11475         as stmt_vec_infos rather than gimple stmts.
11476         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
11477         to a stmt_vec_info.
11478         (vectorizable_live_operation): Likewise.
11479         (vectorizable_reduction, vectorizable_induction): Likewise,
11480         updating use of STMT_VINFO_VEC_STMT.
11481         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
11482         of STMT_VINFO_VEC_STMT.
11483         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
11484         (vectorizable_simd_clone_call, vectorizable_conversion)
11485         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11486         (vectorizable_store, vectorizable_load, vectorizable_condition)
11487         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
11488         from a gimple stmt to a stmt_vec_info.
11489         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
11490         pointer to a stmt_vec_info to the vectorizable_* routines.
11492 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11494         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
11495         a gimple stmt to a stmt_vec_info.
11496         (is_pattern_stmt_p): Update accordingly.
11497         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
11498         (vect_record_grouped_load_vectors): Likewise.
11499         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
11500         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
11501         (vect_model_reduction_cost): Likewise.
11502         (vect_create_epilog_for_reduction): Likewise.
11503         (vectorizable_reduction, vectorizable_induction): Likewise.
11504         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11505         Return the stmt_vec_info for the pattern statement.
11506         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
11507         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
11508         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11509         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
11510         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
11511         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
11512         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
11513         (free_stmt_vec_info, vect_is_simple_use): Likewise.
11515 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11517         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
11518         (vect_finish_stmt_generation): Likewise.
11519         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11520         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
11521         (vect_build_gather_load_calls): Use the return value of the above
11522         functions instead of a separate call to vinfo_for_stmt.  Use narrow
11523         scopes for the input gimple stmt and wider scopes for the associated
11524         stmt_vec_info.  Use vec_info::lookup_def when setting these
11525         stmt_vec_infos from an SSA_NAME definition.
11526         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
11527         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
11528         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
11529         (vectorizable_store, vectorizable_load, vectorizable_condition)
11530         (vectorizable_comparison): Likewise.
11531         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
11532         (vectorizable_reduction): Likewise.
11534 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11536         * tree-vectorizer.h (vect_is_simple_use): Add an optional
11537         stmt_vec_info * parameter before the optional gimple **.
11538         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11539         (process_use, vect_get_vec_def_for_operand_1): Update callers.
11540         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
11541         * tree-vect-loop.c (vectorizable_reduction): Likewise.
11542         (vectorizable_live_operation): Likewise.
11543         * tree-vect-patterns.c (type_conversion_p): Likewise.
11544         (vect_look_through_possible_promotion): Likewise.
11545         (vect_recog_rotate_pattern): Likewise.
11546         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
11548 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11550         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
11551         a typedef to a wrapper class.
11552         (NULL_STMT_VEC_INFO): New macro.
11553         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
11554         (stmt_vec_info::operator*): New function.
11555         (stmt_vec_info::operator gimple *): Likewise.
11556         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
11557         (add_stmt_costs): Likewise.
11558         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
11559         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
11560         (vect_get_known_peeling_cost): Likewise.
11561         (vect_estimate_min_profitable_iters): Likewise.
11562         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11563         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
11564         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
11565         (vectorizable_store, free_stmt_vec_infos): Likewise.
11566         (new_stmt_vec_info): Change return type of xcalloc to
11567         _stmt_vec_info *.
11569 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11571         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
11572         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
11573         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
11574         a single_imm_use-based sequence.
11575         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
11577 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11579         * tree-vectorizer.h (vec_info::lookup_def): Declare.
11580         * tree-vectorizer.c (vec_info::lookup_def): New function.
11581         * tree-vect-patterns.c (vect_get_internal_def): Use it.
11582         (vect_widened_op_tree): Likewise.
11583         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
11584         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
11585         (vectorizable_reduction): Likewise.
11586         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
11587         of a gimple *.
11588         (vect_is_slp_reduction): Update calls accordingly.  Use
11589         vec_info::lookup_def.
11590         (vect_is_simple_reduction): Likewise
11591         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
11593 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11595         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
11596         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
11597         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
11598         of vinfo_for_stmt.
11599         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
11600         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
11601         (vect_update_vf_for_slp, vect_analyze_loop_operations)
11602         (vect_is_slp_reduction, vectorizable_induction)
11603         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
11604         * tree-vect-patterns.c (vect_init_pattern_stmt):
11605         (vect_determine_min_output_precision_1, vect_determine_precisions)
11606         (vect_pattern_recog): Likewise.
11607         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
11608         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
11609         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
11610         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
11611         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
11612         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
11613         info field from a loop to a loop_vec_info.
11615 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11617         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
11618         (vec_info::add_stmt): Declare.
11619         * tree-vectorizer.c (vec_info::add_stmt): New function.
11620         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
11621         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
11622         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
11623         (vectorizable_induction): Likewise.
11624         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
11625         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
11626         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
11627         (vectorizable_load): Likewise.
11628         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
11629         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
11630         (vect_recog_gather_scatter_pattern): Likewise.
11631         (append_pattern_def_seq): Likewise.  Remove a check that is
11632         performed by add_stmt itself.
11634 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11636         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
11637         which make_ssa_name was called with new_stmt before new_stmt
11638         had been created.
11640 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11642         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
11643         split out from...
11644         (vect_is_slp_reduction): ...here...
11645         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
11646         that are already known to be false.
11648 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11650         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
11651         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
11652         STMT_VINFO_NUM_SLP_USES when it's true.
11653         (vect_free_slp_instance): Add a final_p parameter and pass it to
11654         vect_free_slp_tree.
11655         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
11656         (vect_analyze_slp_instance): Likewise.
11657         (vect_slp_analyze_operations): Likewise.
11658         (vect_slp_analyze_bb_1): Likewise.
11659         * tree-vectorizer.c (vec_info): Likewise.
11660         * tree-vect-loop.c (vect_transform_loop): Likewise.
11662 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11664         * tree-vect-loop.c (vectorizable_reduction): Assert that the
11665         function is not called for second and subsequent members of
11666         a reduction group.
11668 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
11670         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
11671         cases for nested loops from here to ...
11672         (vect_create_epilog_for_reduction): ...here.  Only call
11673         vect_is_simple_use for inner-loop reductions.
11675 2018-07-31  Martin Liska  <mliska@suse.cz>
11677         PR gcov-profile/85338
11678         PR gcov-profile/85350
11679         PR gcov-profile/85372
11680         * profile.c (struct location_triplet): New.
11681         (struct location_triplet_hash): Likewise.
11682         (output_location): Do not output a BB that
11683         is already recorded for a line.
11684         (branch_prob): Use streamed_locations.
11686 2018-07-31  Martin Liska  <mliska@suse.cz>
11688         PR gcov-profile/85370
11689         * coverage.c (coverage_begin_function): Do not mark target
11690         clones as artificial functions.
11692 2018-07-31  Martin Liska  <mliska@suse.cz>
11694         PR gcov-profile/83813
11695         PR gcov-profile/84758
11696         PR gcov-profile/85217
11697         PR gcov-profile/85332
11698         * profile.c (branch_prob): Do not record GOTO expressions
11699         for GIMPLE statements which locations are already streamed.
11701 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11703         * gcc.c (handle_spec_function): Accept a soft_matched_part
11704         argument, as do_spec_1.  Pass it down to ...
11705         (eval_spec_function): Accept a soft_matched_part argument,
11706         and pass it down to ...
11707         (do_spec_2): Accept a soft_matched_part argument, and pass
11708         it down to do_spec_1.
11709         (do_spec_1): Pass soft_matched_part to handle_spec_function.
11710         (handle_braces): Update call to handle_spec_function.
11711         (driver::set_up_specs): Update calls to do_spec_2.
11712         (compare_debug_dump_opt_spec_function): Likewise.
11713         (compare_debug_self_opt_spec_function): Likewise.
11715 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11717         * common.opt (nolibc): New option.
11718         * doc/invoke.texi (Link Options): Document it.
11719         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
11720         * config/alpha/linux.h: Likewise.
11721         * config/arc/elf.h: Likewise.
11722         * config/arm/uclinux-elf.h: Likewise.
11723         * config/arm/unknown-elf.h: Likewise.
11724         * config/avr/avrlibc.h: Likewise.
11725         * config/bfin/bfin.h: Likewise.
11726         * config/bfin/linux.h: Likewise.
11727         * config/bfin/uclinux.h: Likewise.
11728         * config/darwin.h: Likewise.
11729         * config/darwin10.h: Likewise.
11730         * config/darwin12.h: Likewise.
11731         * config/gnu-user.h: Likewise.
11732         * config/lm32/uclinux-elf.h: Likewise.
11733         * config/pa/pa-hpux11.h: Likewise.
11734         * config/pa/pa64-hpux.h: Likewise.
11735         * config/sparc/sparc.h: Likewise.
11737 2018-07-31  Olivier Hainque  <hainque@adacore.com>
11739         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
11740         undefined variables.
11742 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11744         PR target/86640
11745         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
11746         instead of GEN_INT.
11748 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
11750         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
11751         terminated string literal.
11753 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
11755         PR rtl-optimization/85160
11756         * combine.c (is_just_move): New function.
11757         (try_combine): Allow combining two instructions into two if neither of
11758         the original instructions was a move.
11760 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
11762         PR target/86673
11763         * doc/extend.texi (Global Register Variables): Discourage use of type
11764         qualifiers.
11765         (Local Register Variables): Likewise.
11767 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
11769         PR tree-optimization/86506
11770         * hwint.h (ceil_log2): Resync with hwint.c implementation.
11772 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
11774         PR target/86547
11775         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
11776         hard_regno, make sure no insn between `from` and `to` clobbers it.
11778 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
11779             Tom de Vries  <tdevries@suse.de>
11781         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
11782         (PTX_DEFAULT_RUNTIME_DIM): ... this.
11783         (nvptx_goacc_validate_dims): Set default worker and gang dims to
11784         PTX_DEFAULT_RUNTIME_DIM.
11785         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
11787 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
11789         * config/pa/pa.c (pa_output_addr_vec): Align address table.
11790         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
11791         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
11793 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
11795         * config/rs6000/constraints.md (wG constraint): Delete, no longer
11796         used.
11797         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
11798         predicate to reflect toc fusion has been deleted.
11799         (toc_fusion_mem_raw): Delete, no longer used.
11800         (toc_fusion_mem_wrapped): Likewise.
11801         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
11802         fusion mask bit.
11803         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
11804         Delete, no longer used.
11805         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
11806         meant to be used for toc fusion.
11807         (rs6000_debug_print_mode): Delete toc fusion debugging.
11808         (rs6000_debug_reg_global): Likewise.
11809         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
11810         fusion and secondary reload support that were never used.
11811         (rs6000_option_override_internal): Delete TOC fusion, that was only
11812         partially defined, and it did not work unless you also used the
11813         -mcmodel= switch.
11814         (rs6000_legitimate_address_p): Delete TOC fusion support.
11815         (rs6000_opt_masks): Likewise.
11816         (fusion_wrap_memory_address): Delete function, no longer used.
11817         (fusion_split_address); Delete TOC fusion support.
11818         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
11819         longer used with toc fusion being deleted.
11820         (TARGET_TOC_FUSION_FP): Likewise.
11821         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
11822         UNSPEC.
11823         (toc fusion spliter): Delete TOC fusion support.
11824         (toc_fusionload_<mode>): Likewise.
11825         (toc_fusionload_di): Likewise.
11826         (fusion_gpr_load_<mode>): Delete generator function, this insn no
11827         longer needs to be named.  Rename predicate to delete TOC fusion.
11828         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11829         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11830         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
11831         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
11832         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
11834 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11836         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
11837         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
11838         __int128 in built-in function prototypes.
11839         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
11840         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
11842 2018-07-27  Martin Sebor  <msebor@redhat.com>
11844         PR tree-optimization/86696
11845         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
11846         types, including enums.
11847         (handle_char_store): Be prepared for the above function to fail.
11849 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
11851         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
11852         when optimization level is lower than 2 or optimize for size.
11854 2018-07-26  Martin Sebor  <msebor@redhat.com>
11856         PR tree-optimization/86043
11857         PR tree-optimization/86042
11858         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
11859         (get_string_cst_length): Rename...
11860         (get_min_string_length): ...to this.  Add argument.
11861         (handle_char_store): Extend to handle multi-character stores by
11862         MEM_REF.
11863         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
11864         * tree.h (initializer_zerop): Add argument.
11866 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
11868         PR middle-end/86660
11869         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
11870         declare target to variables if they have always,{to,from,tofrom} map
11871         kinds.
11873 2018-07-26  Martin Liska  <mliska@suse.cz>
11875         PR lto/86548
11876         * lto-wrapper.c: Add linker_output as prefix
11877         for ltrans_output_file.
11879 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
11881         PR rtl-optimization/85805
11882         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
11883         value for hard registers if that was written in the same mode.
11885 2018-07-26  Martin Liska  <mliska@suse.cz>
11887         PR gcov-profile/86536
11888         * gcov.c (format_gcov): Use printf format %.*f directly
11889         and do not handle special values.
11891 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11893         * common/config/arc/arc-common.c (arc_option_optimization_table):
11894         Update default optimizations for size.
11896 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11898         * config/arc/arc.md (movsf_insn): Add short instruction selection.
11899         * config/arc/constraints.md (CfZ): New constraint.
11900         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
11901         (subsf3_fpu): Likewise.
11902         (cmpsf_fpu): Likewise.
11903         (cmpsf_fpu_uneq): Likewise.
11905 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11907         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
11908         accesses as well.
11909         (arc_is_uncached_mem_p): uncached applies to both the variable and
11910         the pointer.
11912 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
11914         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
11915         register names.
11917 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
11919         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
11920         field "m_scopes" from vec to auto_vec.
11922 2018-07-25  Martin Liska  <mliska@suse.cz>
11924         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
11925         return type.
11927 2018-07-25  Richard Biener  <rguenther@suse.de>
11929         PR debug/86654
11930         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
11931         special wrt context_die late.
11932         (gen_subprogram_die): Re-use DIEs in local scope.
11934 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
11936         PR tree-optimization/86644
11937         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
11939 2018-07-25  Martin Liska  <mliska@suse.cz>
11941         PR middle-end/86645
11942         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
11943         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
11945 2018-07-25  Martin Liska  <mliska@suse.cz>
11947         PR sanitizer/79635
11948         * params.def: Explain ASan abbreviation and provide
11949         a documentation link.
11951 2018-07-24  Martin Sebor  <msebor@redhat.com>
11953         PR tree-optimization/86622
11954         PR tree-optimization/86532
11955         * builtins.h (string_length): Declare.
11956         * builtins.c (c_strlen): Correct handling of non-constant offsets.
11957         (check_access): Be prepared for non-constant length ranges.
11958         (string_length): Make extern.
11959         * expr.c (string_constant): Only handle the minor non-constant
11960         array index.  Use string_constant to compute the length of
11961         a generic string constant.
11963 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
11965         PR tree-optimization/86618
11966         * tree-vect-stmts.c (vectorizable_call): Don't take the address
11967         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
11969 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
11971         PR tree-optimization/86636
11972         * json.cc (json::object::set): Fix comment.  Add assertions.
11973         (json::array::append): Move here from json.h.  Add comment and an
11974         assertion.
11975         (json::string::string): Likewise.
11976         * json.h (json::array::append): Move to json.cc.
11977         (json::string::string): Likewise.
11978         * optinfo-emit-json.cc
11979         (optrecord_json_writer::impl_location_to_json): Assert that we
11980         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
11981         wrapper around it.  Expand the location once, rather than three
11982         times.
11983         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
11984         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
11985         wrappers.
11986         (optrecord_json_writer::optinfo_to_json): Likewise, in four
11987         places.  Fix some overlong lines.
11989 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
11991         * config/aarch64/aarch64-simd.md
11992         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
11993         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
11994         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
11995         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
11996         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
11997         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
11998         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
11999         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
12000         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
12002 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
12004         PR middle-end/86627
12005         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
12006         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
12007         and abs_d == d, do the power of two handling if profitable.
12009 2018-07-24  Richard Biener  <rguenther@suse.de>
12011         * match.pd: Add BIT_FIELD_REF canonicalizations.
12013 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12015         PR c/86617
12016         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
12018 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12020         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
12021         terminated STRING_CST object.
12023 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12025         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
12027 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
12029         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
12030         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
12031         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
12032         the elements into a register.
12033         (rs6000_split_v4si_init_di_reg): Delete.
12034         (rs6000_split_v4si_init): Delete.
12035         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
12036         (vsx_init_v4si): Rewrite as a define_expand.
12038 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
12040         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
12041         zero_extend argument from memory): New.
12043 2018-07-22  Martin Sebor  <msebor@redhat.com>
12045         PR bootstrap/86621
12046         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
12047         diagnosing calls with unknown arguments unless -Walloca-larger-than
12048         is restricted to less than PTRDIFF_MAX bytes.
12050 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
12052         * doc/gcov.texi (Invoking Gcov): Editorial changes.
12054 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
12056         * pretty-print.c (text_info::set_location): Remove redundant
12057         "line_table" parameter from call to rich_location::set_range.
12059 2018-07-20  Martin Sebor  <msebor@redhat.com>
12061         PR middle-end/82063
12062         * builtins.c (expand_builtin_alloca): Adjust.
12063         * calls.c (alloc_max_size): Simplify.
12064         * cgraphunit.c (cgraph_node::expand): Adjust.
12065         * common.opt (larger_than_size, warn_frame_larger_than): Remove
12066         variables.
12067         (frame_larger_than_size): Same.
12068         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
12069         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
12070         Initialize.
12071         * doc/invoke.texi (GCC Command Options): Document option arguments.
12072         Explain byte-size arguments and suffixes.
12073         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
12074         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
12075         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
12076         * doc/options.texi (UInteger): Expand.
12077         (Host_Wide_Int, ByteSize): Document new properties.
12078         * final.c (final_start_function_1): Include sizes in an error message.
12079         * function.c (frame_offset_overflow): Same.
12080         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
12081         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
12082         Diagnose unbounded alloca calls only for limits of less than
12083         PTRDIFF_MAX.
12084         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
12085         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
12086         for alloca(0).
12087         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
12088         only for limits of less than PTRDIFF_MAX.
12089         * langhooks-def.h (lhd_handle_option): Change function argument
12090         to HOST_WIDE_INT.
12091         * langhooks.c (lhd_handle_option): Same.
12092         * langhooks.h (handle_option): Same.
12093         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
12094         ByteSize flags.
12095         (var_type, var_type_struct): Same.
12096         (var_set): Handle ByteSize flag.
12097         * optc-gen.awk: Add comments to output to ease debugging.  Make
12098         use of HOST_WIDE_INT where appropriate.
12099         * opts-gen-save.awk:  Use %lx to format unsigned long.
12100         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
12101         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
12102         arguments.  Parse bytes-size suffixes.
12103         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
12104         (enum_value_to_arg): Same.
12105         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
12106         (handle_option): Adjust.
12107         (generate_option): Change function argument to HOST_WIDE_INT.
12108         (cmdline_handle_error): Adjust.
12109         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
12110         (set_option): Change function argument to HOST_WIDE_INT.
12111         (option_enabled): Handle cl_host_wide_int.
12112         (get_option_state): Handle CLVC_SIZE.
12113         (control_warning_option): Same.
12114         * opts.c (common_handle_option): Change function argument to
12115         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
12116         OPT_Wvla_larger_than_.
12117         * opts.h (enum cl_var_type): Add an enumerator.
12118         * stor-layout.c (layout_decl): Print a more meaningful warning.
12119         * toplev.c (output_stack_usage): Adjust.
12121 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
12123         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
12124         call to inline_expand_builtin_string_cmp.
12125         (expand_builtin_strcmp): Likewise.
12126         (expand_builtin_strncmp): Likewise.
12127         (inline_string_cmp): Delete the last parameter, change char_type_node
12128         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
12129         two operands.
12130         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
12131         the inlining expansion on target where the type of the call has same or
12132         narrower precision than unsigned char.
12134 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
12136         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
12137         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
12138         * common.opt (fsave-optimization-record): New option.
12139         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
12140         * doc/invoke.texi (-fsave-optimization-record): New option.
12141         * dumpfile.c: Include "optinfo-emit-json.h".
12142         (struct kv_pair): Move to coretypes.h.
12143         (optgroup_options): Make non-static.
12144         (dump_context::end_scope): Call
12145         optimization_records_maybe_pop_dump_scope.
12146         * dumpfile.h (optgroup_options): New decl.
12147         * json.cc: New file.
12148         * json.h: New file.
12149         * optinfo-emit-json.cc: New file.
12150         * optinfo-emit-json.h: New file.
12151         * optinfo.cc: Include "optinfo-emit-json.h".
12152         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
12153         (optinfo_enabled_p): Check optimization_records_enabled_p.
12154         (optinfo_wants_inlining_info_p): Likewise.
12155         * optinfo.h: Update comment.
12156         * profile-count.c (profile_quality_as_string): New function.
12157         * profile-count.h (profile_quality_as_string): New decl.
12158         (profile_count::quality): New accessor.
12159         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
12160         and optinfo_emit_json_cc_tests.
12161         * selftest.h (selftest::json_cc_tests): New decl.
12162         (selftest::optinfo_emit_json_cc_tests): New decl.
12163         * toplev.c: Include "optinfo-emit-json.h".
12164         (compile_file): Call optimization_records_finish.
12165         (do_compile): Call optimization_records_start.
12166         * tree-ssa-live.c: Include optinfo.h.
12167         (remove_unused_scope_block_p): Retain inlining information if
12168         optinfo_wants_inlining_info_p returns true.
12170 2018-07-20  Richard Biener  <rguenther@suse.de>
12172         PR debug/86585
12173         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
12174         to cover -flto-partition=none.
12176 2018-07-20  Martin Liska  <mliska@suse.cz>
12178         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
12179         (get_decl_source_range): Remove unused function.
12181 2018-07-20  Richard Biener  <rguenther@suse.de>
12183         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
12184         (struct vn_phi_s): Likewise.
12185         (struct vn_reference_s): Likewise.
12186         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
12187         for searching the slot of an entry known to be in the hash itself.
12188         (vn_phi_hasher::equal): Likewise.
12189         (vn_reference_hasher::equal): Likewise.
12190         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
12191         globals.
12192         (optimistic_info, current_info): Remove, keeping only valid_info.
12193         (vn_reference_lookup_1): Remove fallback lookup.
12194         (vn_reference_lookup_2): Likewise.
12195         (vn_nary_op_lookup_1): Likewise.
12196         (vn_phi_lookup): Likewise.
12197         (vn_nary_build_or_lookup_1): Make sure to not chain the built
12198         hash element.
12199         (vn_reference_insert): Adjust, chain the inserted hash element
12200         at last_inserted_ref.
12201         (vn_reference_insert_pieces): Likewise.
12202         (visit_reference_op_call): Likewise.
12203         (vn_nary_op_insert_into): Chain the inserted hash element at
12204         last_inserted_nary.
12205         (vn_nary_op_insert_pieces): Adjust.
12206         (vn_nary_op_insert): Likewise.
12207         (vn_nary_op_insert_stmt): Likewise.
12208         (vn_phi_insert): Adjust, chain the inserted hash element at
12209         last_inserted_phi.
12210         (process_scc): Remove clearing and copying the optimistic
12211         table.  Instead remove elements inserted during an optimistic
12212         iteration from the single table we maintain.
12213         (init_scc_vn): Adjust.
12214         (free_scc_vn): Likewise.
12215         (sccvn_dom_walker::record_cond): Likewise.
12216         (sccvn_dom_walker::after_dom_children): Likewise.
12218 2018-07-19  Martin Sebor  <msebor@redhat.com>
12220         PR tree-optimization/84047
12221         PR tree-optimization/83776
12222         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
12223         (check_array_bounds): Call it.
12225 2018-07-19  Martin Sebor  <msebor@redhat.com>
12227         * align.h (align_flags): Use member initialization.
12229 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
12231         * Makefile.in (OBJS): Add optinfo.o.
12232         * coretypes.h (class symtab_node): New forward decl.
12233         (struct cgraph_node): New forward decl.
12234         (class varpool_node): New forward decl.
12235         * dump-context.h: New file.
12236         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
12237         "tree-pass.h".
12238         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
12239         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
12240         (set_alt_dump_file): Likewise.
12241         (dump_context::~dump_context): New dtor.
12242         (dump_gimple_stmt): Move implementation to...
12243         (dump_context::dump_gimple_stmt): ...this new member function.
12244         Add the stmt to any pending optinfo, creating one if need be.
12245         (dump_gimple_stmt_loc): Move implementation to...
12246         (dump_context::dump_gimple_stmt_loc): ...this new member function.
12247         Start a new optinfo and add the stmt to it.
12248         (dump_gimple_expr): Move implementation to...
12249         (dump_context::dump_gimple_expr): ...this new member function.
12250         Add the stmt to any pending optinfo, creating one if need be.
12251         (dump_gimple_expr_loc): Move implementation to...
12252         (dump_context::dump_gimple_expr_loc): ...this new member function.
12253         Start a new optinfo and add the stmt to it.
12254         (dump_generic_expr): Move implementation to...
12255         (dump_context::dump_generic_expr): ...this new member function.
12256         Add the tree to any pending optinfo, creating one if need be.
12257         (dump_generic_expr_loc): Move implementation to...
12258         (dump_context::dump_generic_expr_loc): ...this new member
12259         function.  Add the tree to any pending optinfo, creating one if
12260         need be.
12261         (dump_printf): Move implementation to...
12262         (dump_context::dump_printf_va): ...this new member function.  Add
12263         the text to any pending optinfo, creating one if need be.
12264         (dump_printf_loc): Move implementation to...
12265         (dump_context::dump_printf_loc_va): ...this new member function.
12266         Start a new optinfo and add the stmt to it.
12267         (dump_dec): Move implementation to...
12268         (dump_context::dump_dec): ...this new member function.  Add the
12269         value to any pending optinfo, creating one if need be.
12270         (dump_context::dump_symtab_node): New member function.
12271         (dump_context::get_scope_depth): New member function.
12272         (dump_context::begin_scope): New member function.
12273         (dump_context::end_scope): New member function.
12274         (dump_context::ensure_pending_optinfo): New member function.
12275         (dump_context::begin_next_optinfo): New member function.
12276         (dump_context::end_any_optinfo): New member function.
12277         (dump_context::s_current): New global.
12278         (dump_context::s_default): New global.
12279         (dump_scope_depth): Delete global.
12280         (dumpfile_ensure_any_optinfo_are_flushed): New function.
12281         (dump_symtab_node): New function.
12282         (get_dump_scope_depth): Reimplement in terms of dump_context.
12283         (dump_begin_scope): Likewise.
12284         (dump_end_scope): Likewise.
12285         (selftest::temp_dump_context::temp_dump_context): New ctor.
12286         (selftest::temp_dump_context::~temp_dump_context): New dtor.
12287         (selftest::verify_item): New function.
12288         (ASSERT_IS_TEXT): New macro.
12289         (ASSERT_IS_TREE): New macro.
12290         (ASSERT_IS_GIMPLE): New macro.
12291         (selftest::test_capture_of_dump_calls): New test.
12292         (selftest::dumpfile_c_tests): Call it.
12293         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
12294         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
12295         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
12296         descriptive comment.
12297         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
12298         (dump_node, dump_bb): Move these unrelated decls.
12299         (class dump_manager): Add leading comment.
12300         * optinfo.cc: New file.
12301         * optinfo.h: New file.
12303 2018-07-19  Michael Collison  <michael.collison@arm.com>
12304             Richard Henderson <rth@redhat.com>
12306         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
12307         (subti): Handle op1 zero.
12308         (subvti4, usub4ti4): New.
12309         (*sub<GPI>3_compare1_imm): New.
12310         (sub<GPI>3_carryinCV): New.
12311         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
12312         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
12314 2018-07-19  Michael Collison  <michael.collison@arm.com>
12315             Richard Henderson <rth@redhat.com>
12317         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
12318         (addti3): Create simpler code if low part is already known to be 0.
12319         (addvti4, uaddvti4): New.
12320         (*add<GPI>3_compareC_cconly_imm): New.
12321         (*add<GPI>3_compareC_cconly): New.
12322         (*add<GPI>3_compareC_imm): New.
12323         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
12324         handle constants within this pattern..
12325         (*add<GPI>3_compareV_cconly_imm): New.
12326         (*add<GPI>3_compareV_cconly): New.
12327         (*add<GPI>3_compareV_imm): New.
12328         (add<GPI>3_compareV): New.
12329         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
12330         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
12331         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
12332         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
12333         with 'comparison' operator.
12334         (*add<GPI>3_compareV_cconly_imm): Ditto.
12335         (*add<GPI>3_compareV_cconly): Ditto.
12336         (*add<GPI>3_compareV_imm): Ditto.
12337         (add<GPI>3_compareV): Ditto.
12338         (add<mode>3_carryinC): Ditto.
12339         (*add<mode>3_carryinC_zero): Ditto.
12340         (*add<mode>3_carryinC): Ditto.
12341         (add<mode>3_carryinV): Ditto.
12342         (*add<mode>3_carryinV_zero): Ditto.
12343         (*add<mode>3_carryinV): Ditto.
12345 2018-07-19  Michael Collison  <michael.collison@arm.com>
12346             Richard Henderson <rth@redhat.com>
12348         * config/aarch64/aarch64-modes.def (CC_V): New.
12349         * config/aarch64/aarch64-protos.h
12350         (aarch64_addti_scratch_regs): Declare
12351         (aarch64_subvti_scratch_regs): Declare.
12352         (aarch64_expand_subvti): Declare.
12353         (aarch64_gen_unlikely_cbranch): Declare
12354         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
12355         for signed overflow using CC_Vmode.
12356         (aarch64_get_condition_code_1): Handle CC_Vmode.
12357         (aarch64_gen_unlikely_cbranch): New function.
12358         (aarch64_addti_scratch_regs): New function.
12359         (aarch64_subvti_scratch_regs): New function.
12360         (aarch64_expand_subvti): New function.
12362 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12364         * config/aarch64/aarch64-option-extensions.def: New entry for profile
12365         extension.
12366         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
12367         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
12368         extension.
12370 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12372         PR target/83009
12373         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12374         address check not strict.
12376 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12378         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
12379         Umq with Umn.
12380         (store_pair_lanes<mode>): Likewise.
12381         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
12382         enum value 'ADDR_QUERY_LDP_STP_N'.
12383         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
12384         (aarch64_print_address_internal): Add declaration.
12385         (aarch64_print_ldpstp_address): Remove.
12386         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
12387         (aarch64_print_operand): Change printing of 'y'.
12388         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
12389         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
12390         'true' rather than '1'.
12391         * config/aarch64/constraints.md (Uml): Likewise.
12392         (Uml): Rename to Umn.
12393         (Umq): Remove.
12395 2018-07-19  Richard Biener  <rguenther@suse.de>
12397         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
12398         a trailing array.
12399         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
12400         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
12401         (vn_reference_hasher): Likewise.
12402         (struct vn_tables_s): Remove obstack and alloc-pool members.
12403         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
12404         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
12405         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
12406         (vn_reference_insert_pieces): Likewise.
12407         (alloc_vn_nary_op_noinit): Adjust.
12408         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
12409         (vn_phi_eq): Adjust.
12410         (shared_lookup_phiargs): Remove.
12411         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
12412         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
12413         (visit_reference_op_call): Likewise.
12414         (copy_nary, copy_phi, copy_reference): Remove.
12415         (process_scc): Rewind the obstack when iterating.  Do not
12416         copy the elements to valid_info but just move them from one
12417         hashtable to the other.
12418         (allocate_vn_table): Adjust.
12419         (free_vn_table): Likewise.
12420         (init_scc_vn): Likewise.
12421         (free_scc_vn): Likewise.
12423 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
12425         PR target/86560
12426         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
12427         indirect_return as function type attribute.
12428         (ix86_attribute_table): Change indirect_return to function
12429         type attribute.
12430         * doc/extend.texi: Update indirect_return attribute.
12432 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
12434         * wide-int.h (widest2_int): New.
12435         * gimple-fold.c (arith_overflowed_p): Use it.
12436         * tree.h (widest2_int_cst): New.
12437         * tree-vrp.c (wide_int_binop_overflow): Rename from
12438         vrp_int_const_binop.
12439         Rewrite to work on trees.
12440         (extract_range_from_multiplicative_op_1): Abstract code to...
12441         (wide_int_range_min_max): ...here.
12442         (wide_int_range_cross_product): ...and here.
12443         (extract_range_from_binary_expr_1): Abstract overflow code to...
12444         (wide_int_range_mult_wrapping): ...here.
12445         * tree-vrp.h (wide_int_range_cross_product): New.
12446         (wide_int_range_mult_wrapping): New.
12448 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
12449             Julia Koval  <julia.koval@intel.com>
12451         * config/i386/x86-tune-costs.h (skylake_memcpy,
12452         skylake_memset): Replace rep_prefix with unrolling for size 512.
12454 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
12456         PR middle-end/86544
12457         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
12458         comparision with EQ_EXPR in last stmt.
12460 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12462         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
12463         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
12464         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
12465         previously known as "PowerPC AltiVec Built-in Functions".  Move
12466         some material to new subsubsections "PowerPC AltiVec Built-in
12467         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
12468         ISA 2.07".
12469         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
12470         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
12471         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
12472         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
12474 2018-07-18  Richard Biener  <rguenther@suse.de>
12476         PR tree-optimization/86557
12477         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
12478         EXACT_DIV_EXPR.
12480 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
12482         * config/s390/s390.c (s390_function_profiler): Generate CFI.
12484 2018-07-17  Jeff Law  <law@redhat.com>
12486         * config/arm/arm.c (get_label_padding): Update for recent
12487         changes to label_to_alignment.
12489         PR tree-optimization/86010
12490         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
12492         * config/mips/mips.c (vr4130_align_insns): Update for recent
12493         changes to label_to_alignment.
12495         * config/frv/frv.c (frv_label_align): Update for recent changes
12496         to label_to_alignment.
12498         * config/nios2/nios2.c (nios2_label_align): Update for recent
12499         changes which dropped ALIGN_LABELS_LOG.
12501 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
12503         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
12504         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
12506 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
12508         * config/arc/arc.c (arc_label_align): Use align_labels instead of
12509         deprecated align_labels_log.
12511 2018-07-17  Richard Biener  <rguenther@suse.de>
12513         PR lto/86456
12514         * dwarf2out.c (init_sections_and_labels): Always generate
12515         a debug_line_str_section for early LTO debug.
12516         (dwarf2out_finish): Reset debug_line_str_hash output early.
12517         Bump counter for extra dwarf5 .debug_loc labels to not conflict
12518         with fat LTO part.
12519         (dwarf2out_early_finish): Output debug_line_str.
12521 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12523         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
12524         index register on z196 or later.
12526 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
12528         * config/s390/s390.c (s390_default_align): Set default function
12529         alignment to 16.
12530         (s390_override_options_after_change): Call s390_default align.
12531         (s390_option_override_internal): Call s390_default_align.
12532         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
12534 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12536         PR middle-end/86542
12537         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
12538         to _looptemp_ clauses, other than the first two.
12540 2018-07-17  Martin Liska  <mliska@suse.cz>
12542         * opts.c: Do not enable OPT_falign_* for -Os.
12544 2018-07-17  Martin Liska  <mliska@suse.cz>
12546         * align.h (MAX_CODE_ALIGN): New.
12547         (MAX_CODE_ALIGN_VALUE): New.
12548         * common/config/i386/i386-common.c (ix86_handle_option):
12549         (MAX_CODE_ALIGN): Moved to align.h.
12550         * final.c (MAX_CODE_ALIGN): Likewise.
12551         * opts.c (parse_and_check_align_values):
12552         (MAX_CODE_ALIGN): Likewise.
12553         (MAX_CODE_ALIGN_VALUE): Likewise.
12555 2018-07-17  Martin Liska  <mliska@suse.cz>
12557         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
12558         in order to fulfil coding style.
12559         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
12560         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12561         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
12562         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
12563         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
12564         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
12565         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
12566         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
12567         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
12569 2018-07-17  Martin Liska  <mliska@suse.cz>
12571         * align.h: New file.
12572         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
12573         directly.
12574         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
12575         align_flags of label_to_alignment.
12576         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
12577         align_flags class.
12578         * config/m68k/m68k.c: Do not use removed align_labels_value and
12579         align_loops_value.
12580         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
12581         (LOOP_ALIGN): Likewise.
12582         (LABEL_ALIGN): Likewise.
12583         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12584         Remove not used macro.
12585         (rs6000_loop_align): Change return type to align_flags.
12586         (rs6000_loop_align_max_skip): Remove.
12587         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
12588         Change return type to align_flags.
12589         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
12590         Remove not used macro.
12591         (rs6000_loop_align):  Change return type to align_flags.
12592         (rs6000_loop_align_max_skip): Remove.
12593         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
12594         * config/rx/rx-protos.h (rx_align_for_label): Make it
12595         static function.
12596         * config/rx/rx.c (rx_align_for_label): Change return type
12597         to align_flags.
12598         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
12599         macro definitions.
12600         into align_flags class.
12601         (LABEL_ALIGN): Likewise.
12602         (LOOP_ALIGN): Likewise.
12603         * config/s390/s390.c (s390_label_align): Use align_flags
12604         class member.
12605         (s390_asm_output_function_label): Likewise.
12606         * config/sh/sh.c (sh_override_options_after_change):
12607         Use align_flags class directly without macros.
12608         (find_barrier): Likewise.
12609         (barrier_align): Likewise.
12610         (sh_loop_align): Likewise.
12611         * config/spu/spu.c (spu_option_override):
12612         Use align_flags_tuple::get_value instead of removed macros.
12613         (spu_sched_init): Likewise.
12614         * config/spu/spu.h (GTY): Likewise.
12615         * config/visium/visium.c (visium_option_override):
12616         Set "8" as default secondary alignment.
12617         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
12618         in order to guarantee secondary alignment of 8.
12619         * coretypes.h: Include align.h header file.
12620         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
12621         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
12622         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
12623         * doc/tm.texi.in: Likewise.
12624         * final.c (struct label_alignment): Remove not used structure.
12625         (LABEL_ALIGN): Change type to align_flags.
12626         (LOOP_ALIGN): Likewise.
12627         (JUMP_ALIGN): Likewise.
12628         (default_loop_align_max_skip): Remove.
12629         (default_label_align_max_skip): Likewise.
12630         (default_jump_align_max_skip): Likewise.
12631         (default_label_align_after_barrier_max_skip):
12632         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
12633         (LABEL_TO_MAX_SKIP): Remove.
12634         (label_to_alignment): Return align_flags type instead of integer.
12635         (label_to_max_skip): Remove.
12636         (align_fuzz): Use align_flags type.
12637         (compute_alignments): Use align_flags type and use align_flags::max
12638         to combine multiple alignments.
12639         (grow_label_align): Grow vec instead of C array.
12640         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
12641         (shorten_branches):  Use align_flags type and use align_flags::max
12642         to combine multiple alignments.
12643         (final_scan_insn_1): Remove usage of secondary alignment that comes
12644         from label alignment, but instead use proper secondary alignment
12645         which is computed in grow_label_align.
12646         * flags.h (struct align_flags_tuple): Move to align.h.
12647         (struct align_flags): Likewise.
12648         (state_align_loops): Rename to align_loops.
12649         (state_align_jumps): Rename to align_jumps.
12650         (state_align_labels): Rename to align_labels.
12651         (state_align_functions): Rename to align_functions.
12652         (align_loops_log): Remove.
12653         (align_jumps_log): Remove.
12654         (align_labels_log): Remove.
12655         (align_functions_log): Remove.
12656         (align_loops_max_skip): Remove.
12657         (align_jumps_max_skip): Remove.
12658         (align_labels_max_skip): Remove.
12659         (align_functions_max_skip): Remove.
12660         (align_loops_value): Remove.
12661         (align_jumps_value): Remove.
12662         (align_labels_value): Remove.
12663         (align_functions_value): Remove.
12664         * output.h (label_to_alignment): Change return type to align_flags.
12665         (label_to_max_skip): Remove.
12666         * target.def: Remove loop_align_max_skip, label_align_max_skip,
12667         jump_align_max_skip macros.
12668         * targhooks.h (default_loop_align_max_skip): Remove.
12669         (default_label_align_max_skip): Likewise.
12670         (default_jump_align_max_skip): Likewise.
12671         (default_label_align_after_barrier_max_skip): Remove.
12672         * toplev.c (read_log_maxskip): Use ::normalize function.
12673         (parse_N_M): Remove not used argument and also call ::normalize.
12674         (parse_alignment_opts): Do not pass unused arguments.
12675         * varasm.c (assemble_start_function): Use directly align_functions
12676         instead of removed macros.
12677         * system.h: Do not poison removed macros.
12679 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
12681         PR middle-end/86539
12682         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
12683         and cond temporaries don't have reference type if iterator has
12684         pointer type.  For init use &for_pre_body instead of pre_p if
12685         for_pre_body is non-empty.
12687 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12689         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
12690         double-double modes to SFmode directly directly.
12691         (trunc<mode>sf2_fprs): Delete.
12693 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12695         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
12696         for conversions between IFmode and the decimal floating point modes.
12697         (init_float128_ieee): Use the correct names for conversions between
12698         KFmode and the decimal floating point modes.
12700 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
12702         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
12703         for the conversions between TDmode and IFmode.
12704         (init_float128_ieee): Use more correct names for the conversions
12705         between TDmode and KFmode.
12707 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
12709         PR tree-optimization/86526
12710         * builtins.c (expand_builtin_memcmp): Formatting fixes.
12711         (inline_expand_builtin_string_cmp): Likewise.
12712         (inline_string_cmp): Likewise.  Use c_readstr instead of
12713         builtin_memcpy_read_str.  Add unit_mode temporary.
12715 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12717         PR middle-end/86528
12718         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
12719         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
12721 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12723         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
12724         Alphabetize prototypes of built-in functions, separating out
12725         built-in functions that are listed in this section but should be
12726         described elsewhere.
12728 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
12730         PR target/86511
12731         * expmed.c (emit_store_flag): Do not emit setcc followed by a
12732         conditional move when trapping comparison was split to a
12733         non-trapping one (and vice versa).
12735 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12737         * config/s390/s390.c (s390_function_profiler): Generate nops
12738         instead of profiler call sequences.
12739         * config/s390/s390.opt: Add the new option.
12741 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12743         * config/s390/s390.c (s390_function_profiler): Generate
12744         __mcount_loc section.
12745         * config/s390/s390.opt: Add the new option.
12747 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
12749         * common.opt: Add the new warning.
12750         * config/s390/s390.c (s390_function_profiler): Emit "brasl
12751         %r0,__fentry__" when -mfentry is specified.
12752         (s390_option_override_internal): Disallow -mfentry for 31-bit
12753         CPUs.
12754         * config/s390/s390.opt: Add the new option.
12756 2018-07-16  Richard Biener  <rguenther@suse.de>
12758         PR lto/86523
12759         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
12760         for function-local FUNCTION_DECL and RESULT_DECL immediately.
12762 2018-07-16  Martin Liska  <mliska@suse.cz>
12764         PR ipa/86529
12765         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
12766         to ::get_create.
12768 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12770         * config/arc/arcHS.md: Update ARCHS scheduling rules.
12772 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
12774         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
12775         for ARCHS4x.
12776         * config/arc/arc-cpus.def (hs4x): New cpu.
12777         (hs4xd): Likewise.
12778         * config/arc/arc-tables.opt: Regenerate.
12779         * config/arc/arc.c (arc_sched_issue_rate): New function.
12780         (TARGET_SCHED_ISSUE_RATE): Define.
12781         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
12782         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
12783         fpu_cvt.
12784         (attr tune): Add ARCHS4x tune values.
12785         (attr tune_dspmpy): Define.
12786         (*tst): Correct instruction type.
12787         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
12788         * config/arc/arcHS4x.md: New file.
12789         * config/arc/fpu.md: Update instruction type attributes.
12790         * config/arc/t-multilib: Regenerate.
12792 2018-07-16  Tom de Vries  <tdevries@suse.de>
12794         PR debug/86455
12795         * var-tracking.c (vt_initialize): Fix pre_dec handling.
12797 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
12799         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
12800         early clobber.
12802 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
12804         PR tree-optimization/86514
12805         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
12806         conversion to a boolean type from a type with greater precision.
12808 2018-07-16  Tom de Vries  <tdevries@suse.de>
12810         * var-tracking.c (vt_initialize): Print adjusted insn slim if
12811         dump_flags request TDF_SLIM.
12813 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
12815         * fold-const.c (int_const_binop_1): Abstract...
12816         (wide_int_binop): ...wide int code here.
12817         (poly_int_binop): ...poly int code here.
12818         Abstract the rest of int_const_binop_1 into int_const_binop.
12819         * fold-const.h (wide_int_binop): New.
12820         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
12821         Remove useless PLUS/MINUS_EXPR case.
12822         (zero_nonzero_bits_from_vr): Move wide int code...
12823         (zero_nonzero_bits_from_bounds): ...here.
12824         (extract_range_from_binary_expr_1): Move mask optimization code...
12825         (range_easy_mask_min_max): ...here.
12826         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
12827         (range_easy_mask_min_max): New.
12829 2018-07-15  Jeff Law  <law@redhat.com>
12831         PR target/85993
12832         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
12833         block.
12835 2018-07-14  Jim Wilson  <jimw@sifive.com>
12837         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
12839 2018-07-14  Paul Koning  <ni1d@arrl.net>
12841         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
12843 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
12845         * lto-streamer-out.c (copy_function_or_variable): Dump info about
12846         copying section.
12848 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
12849             Steve Munroe  <munroesj52@gmail.com>
12851         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
12852         (_mm_andnot_si128): Likewise.
12853         (_mm_or_si128): Likewise.
12854         (_mm_xor_si128): Likewise.
12856 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
12858         PR middle-end/78809
12859         * builtins.c (expand_builtin_memcmp): Inline the calls first
12860         when result_eq is false.
12861         (expand_builtin_strcmp): Inline the calls first.
12862         (expand_builtin_strncmp): Likewise.
12863         (inline_string_cmp): New routine. Expand a string compare
12864         call by using a sequence of char comparison.
12865         (inline_expand_builtin_string_cmp): New routine. Inline expansion
12866         a call to str(n)cmp/memcmp.
12867         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
12868         New option.
12869         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
12871 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
12873         * config/arm/driver-arm.c: Include arm-native.h.
12874         (host_detect_local_cpu): Use auto-generated data tables.
12875         (vendors, arm_cpu_table): Delete.  Move part information to ...
12876         * config/arm/arm-cpus.in: ... here.
12877         * config/arm/parsecpu.awk (gen_native): New function.
12878         (vendor, part): New CPU fields.
12879         (END): Add support for building the native CPU detection tables.
12880         * config/arm/t-arm (arm-native.h): Add build rule.
12881         (driver-arm.o): Add dependency on arm-native.h.
12883 2018-07-13  Richard Biener  <rguenther@suse.de>
12885         PR middle-end/85974
12886         * match.pd (addr1 - addr2): Allow either of the operand to
12887         have a conversion.
12889 2018-07-13  Tom de Vries  <tdevries@suse.de>
12891         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
12892         in remap_ssa_name.
12894 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
12896         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
12897         arrays instead of numbered variables.
12899 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
12901         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
12902         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
12904 2018-07-13  Richard Biener  <rguenther@suse.de>
12906         PR debug/86452
12907         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
12908         instead of get_context_die.
12910 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
12911             Richard Biener  <rguenther@suse.de>
12913         PR middle-end/86489
12914         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
12915         that the loop latch destination where phi is defined.
12917 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
12919         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
12920         (riscv_expand_epilogue): Add assertion to check interrupt mode.
12921         (riscv_set_current_function): Extract getting interrupt type to new
12922         function.
12923         (riscv_get_interrupt_type): New function.
12924         (riscv_merge_decl_attributes): New function, checking interrupt type is
12925         same.
12926         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
12928 2018-07-12  Paul Koning  <ni1d@arrl.net>
12930         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
12931         directive.
12933 2018-07-12  Paul Koning  <ni1d@arrl.net>
12935         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
12936         zero reference, add doloop_end instead.
12937         * doc/md.texi (decrement_and_branch_until_zero): Remove.
12938         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
12939         detail for doloop_end.
12941 2018-07-12  Martin Sebor  <msebor@redhat.com>
12943         PR c/86453
12944         * attribs.c (decl_attributes): Reject conflicting attributes before
12945         calling attribute handlers.
12947 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
12949         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
12950          parameter.
12951         (gcc::dump_manager::get_dump_file_name): likewise.
12952         (dump_begin): Likewise.
12953         * dumpfile.h (dump_begin): Update prototype.
12954         (gcc::dump_manager::get_dump_file_name,
12955         gcc::dump_manager::get_dump_file_name): Update prototype.
12957 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12959         * internal-fn.h (vectorizable_internal_fn_p): New function.
12960         * tree-vect-slp.c (compatible_calls_p): Likewise.
12961         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
12962         to internal functions.
12963         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
12965 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
12967         * fold-const.h (inverse_conditions_p): Declare.
12968         * fold-const.c (inverse_conditions_p): New function.
12969         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
12970         that test the inverse condition of a conditional internal function.
12971         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
12972         * internal-fn.c (internal_fn_mask_index): Handle conditional
12973         internal functions.
12974         (vectorized_internal_fn_supported_p): New function.
12975         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
12976         (any_pred_load_store): Replace with...
12977         (need_to_predicate): ...this new variable.
12978         (redundant_ssa_names): New variable.
12979         (ifcvt_can_use_mask_load_store): Move initial checks to...
12980         (ifcvt_can_predicate): ...this new function.  Handle tree codes
12981         for which a conditional internal function exists.
12982         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
12983         instead of ifcvt_can_use_mask_load_store.  Update after variable
12984         name change.
12985         (predicate_load_or_store): New function, split out from
12986         predicate_mem_writes.
12987         (check_redundant_cond_expr): New function.
12988         (value_available_p): Likewise.
12989         (predicate_rhs_code): Likewise.
12990         (predicate_mem_writes): Rename to...
12991         (predicate_statements): ...this.  Use predicate_load_or_store
12992         and predicate_rhs_code.
12993         (combine_blocks, tree_if_conversion): Update after above name changes.
12994         (ifcvt_local_dce): Handle redundant_ssa_names.
12995         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
12996         general conditional functions.
12997         * tree-vect-stmts.c (vectorizable_call): Likewise.
12999 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13000             Alan Hayward  <alan.hayward@arm.com>
13001             David Sherwood  <david.sherwood@arm.com>
13003         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
13004         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
13005         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
13006         plus and minus and convert them into IFN_COND_FMA-based sequences.
13007         (convert_mult_to_fma): Handle conditional plus and minus.
13009 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13011         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
13012         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
13013         (cond_fnms_optab): New optabs.
13014         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
13015         internal functions.
13016         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
13017         * internal-fn.h (get_conditional_internal_fn): Declare.
13018         (get_unconditional_internal_fn): Likewise.
13019         * internal-fn.c (cond_ternary_direct): New macro.
13020         (expand_cond_ternary_optab_fn): Likewise.
13021         (direct_cond_ternary_optab_supported_p): Likewise.
13022         (FOR_EACH_COND_FN_PAIR): Likewise.
13023         (get_conditional_internal_fn): New function.
13024         (get_unconditional_internal_fn): Likewise.
13025         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
13026         (gimple_match_op::gimple_match_op): Add a new overload for 5
13027         operands.
13028         (gimple_match_op::set_op): Likewise.
13029         (gimple_resimplify5): Declare.
13030         * genmatch.c (decision_tree::gen): Generate simplifications for
13031         5 operands.
13032         * gimple-match-head.c (gimple_simplify): Define an overload for
13033         5 operands.  Handle calls with 5 arguments in the top-level overload.
13034         (convert_conditional_op): Handle conversions from unconditional
13035         internal functions to conditional ones.
13036         (gimple_resimplify5): New function.
13037         (build_call_internal): Pass a fifth operand.
13038         (maybe_push_res_to_seq): Likewise.
13039         (try_conditional_simplification): Try converting conditional
13040         internal functions to unconditional internal functions.
13041         Handle 3-operand unconditional forms.
13042         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
13043         Define ternary equivalents of the current rules for binary conditional
13044         internal functions.
13045         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
13046         ternary operations.
13047         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
13048         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
13049         (optab): Handle them.
13050         (SVE_COND_FP_TERNARY): New int iterator.
13051         (sve_fmla_op, sve_fmad_op): New int attributes.
13052         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
13053         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
13054         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
13056 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
13058         * target.def (preferred_else_value): New target hook.
13059         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
13060         * doc/tm.texi: Regenerate.
13061         * targhooks.h (default_preferred_else_value): Declare.
13062         * targhooks.c (default_preferred_else_value): New function.
13063         * internal-fn.h (conditional_internal_fn_code): Declare.
13064         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
13065         (get_conditional_internal_fn): Use it.
13066         (conditional_internal_fn_code): New function.
13067         * gimple-match.h (gimple_match_cond): New struct.
13068         (gimple_match_op): Add a cond member function.
13069         (gimple_match_op::gimple_match_op): Update all forms to take a
13070         gimple_match_cond.
13071         * genmatch.c (expr::gen_transform): Use the same condition as res_op
13072         for the suboperation, but don't specify a particular else_value.
13073         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
13074         (visit_nary_op, visit_reference_op_load): Pass
13075         gimple_match_cond::UNCOND to the gimple_match_op constructor.
13076         * gimple-match-head.c: Include tree-eh.h
13077         (convert_conditional_op): New function.
13078         (maybe_resimplify_conditional_op): Likewise.
13079         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
13080         (gimple_resimplify2): Likewise.
13081         (gimple_resimplify3): Likewise.
13082         (gimple_resimplify4): Likewise.
13083         (maybe_push_res_to_seq): Return null for conditional operations.
13084         (try_conditional_simplification): New function.
13085         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
13086         constructor.
13087         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
13088         IFN_COND_* call.
13089         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
13090         function.
13091         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
13093 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
13095         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
13096         DECL_FCONTEXT
13097         (hash_tree): Do not hash DECL_FCONTEXT
13098         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
13099         Do not stream DECL_FCONTEXT.
13100         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
13101         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
13103 2018-07-12  Richard Biener  <rguenther@suse.de>
13105         PR debug/86462
13106         * dwarf2out.c (gen_block_die): Only output blocks when they have
13107         at least one !DECL_IGNORED_P variable.
13109 2018-07-12  Richard Biener  <rguenther@suse.de>
13111         PR target/84829
13112         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
13113         Remove -mieee-fp handling.
13115 2018-07-12  Richard Biener  <rguenther@suse.de>
13117         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
13118         left-over from last patch.
13120 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
13122         PR tree-optimization/86492
13123         * gimple-ssa-store-merging.c
13124         (imm_store_chain_info::coalesce_immediate_stores): Call
13125         check_no_overlap even for the merge_overlapping case.  Formatting fix.
13127 2018-07-12  Richard Biener  <rguenther@suse.de>
13129         PR middle-end/86479
13130         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
13131         move possibly trapping operations into the conditional.
13133 2018-07-12  Richard Biener  <rguenther@suse.de>
13135         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
13136         (vn_lookup_simplify_result): Remove recursion limit applied
13137         here.
13138         (vn_nary_build_or_lookup_1): Adjust.
13139         (try_to_simplify): Likewise.
13140         * gimple-match-head.c (gimple_resimplify1): Instead apply one
13141         here.
13142         (gimple_resimplify2): Likewise.
13143         (gimple_resimplify3): Likewise.
13144         (gimple_resimplify4): Likewise.
13146 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
13148         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
13149         Use __mmask64 type instead of __mmask8 for __M argument.
13150         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
13151         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
13152         __U argument.
13153         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
13154         __mmask16 for __M argument.
13155         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
13156         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
13157         to __mmask16 instead of __mmask8.
13158         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
13159         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
13160         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
13161         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
13162         instead of __mmask16 for __U argument.
13163         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
13164         __mmask16 instead of __mmask8 for __U argument.
13165         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
13166         __U argument.
13167         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
13168         __mmask16.
13169         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
13170         argument.
13171         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
13172         __U argument.
13173         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
13174         __mmask16.
13175         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
13176         of __mmask16.
13177         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
13178         __U argument.
13179         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
13180         __U argument.
13181         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
13182         __U argument.
13183         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
13184         __U argument.
13185         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
13186         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
13187         return type as well as __M argument type and all casts from __mmask8
13188         to __mmask32.
13189         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
13190         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
13191         return type as well as __M argument type and all casts from __mmask8
13192         to __mmask16.
13193         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
13194         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
13195         return type as well as __M argument type and all casts from __mmask8
13196         to __mmask32.
13197         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
13198         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
13199         return type as well as __M argument type and all casts from __mmask8
13200         to __mmask16.
13201         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
13202         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
13203         __mmask16.
13205 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
13207         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
13208         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
13209         for __U argument.
13211 2018-07-11  Paul Koning  <ni1d@arrl.net>
13213         * doc/md.texi (define_subst): Document how multiple occurrences of
13214         the same argument in the replacement pattern are handled.
13216 2018-07-11  Paul Koning  <ni1d@arrl.net>
13218         * doc/extend.texi (Common Variable Attributes): Move "mode" into
13219         alphabetical order.
13220         (Common Type Attributes): Add "mode" attribute.
13222 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
13224         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
13225         stream DECL_ORIGINAL_TYPE.
13226         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
13227         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
13228         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
13229         Do not walk original type.
13230         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
13231         external decls.
13232         (write_ts_decl_non_common_tree_pointers): Do not stream
13233         DECL_ORIGINAL_TYPE
13234         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
13235         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
13237 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
13239         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
13240         thread twice from the same starting edge.
13242 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
13244         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
13245         * gimple.c (gimple_call_nonnull_result_p): ...here...
13246         (gimple_call_nonnull_arg): ...and here.
13247         * gimple.h (gimple_call_nonnull_result_p): New.
13248         (gimple_call_nonnull_arg): New.
13250 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
13252         * config/arm/arm-cpus.in: Move information from fpu field of each
13253         cpu definition to the isa field.
13254         * config/arm/parsecpu.awk (fpu): Delete match rule.
13255         (gen_comm_data): Don't add bits from the CPU's FPU entry.
13257 2018-07-11  Richard Biener  <rguenther@suse.de>
13259         PR debug/86457
13260         * dwarf2out.c (init_sections_and_labels): Use
13261         output_asm_line_debug_info consistently.
13262         (dwarf2out_early_finish): Likewise.
13263         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
13264         type units.
13266 2018-07-11  Richard Biener  <rguenther@suse.de>
13268         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
13269         Rework father_bb setting in a way to avoid propagating constants
13270         multiple times on a loop body.
13272 2018-07-10  Mark Wielaard  <mark@klomp.org>
13274         PR debug/86459
13275         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
13277 2018-07-10  Richard Biener  <rguenther@suse.de>
13279         * hash-map.h (hash_map::iterator::operator*): Return
13280         references to key and value.
13282 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
13284         PR c++/86443
13285         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
13286         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
13287         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
13288         (gimplify_omp_for): For composite loops, move outer
13289         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
13290         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
13291         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
13292         TREE_LIST for both the original class iterator and the "last" helper
13293         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
13294         loop, remember has_decl_expr from outer composite loops for the
13295         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
13297 2018-07-09  Martin Sebor  <msebor@redhat.com>
13299         PR middle-end/77357
13300         PR middle-end/86428
13301         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
13302         accessing implicitly initialized array elements.
13303         * expr.c (string_constant): Handle string initializers of
13304         character arrays within aggregates.
13305         * gimple-fold.c (fold_array_ctor_reference): Add argument.
13306         Store element offset.  As a special case, handle zero size.
13307         (fold_nonarray_ctor_reference): Same.
13308         (fold_ctor_reference): Add argument.  Store subobject offset.
13309         * gimple-fold.h (fold_ctor_reference): Add argument.
13311 2018-07-09  Paul Koning  <ni1d@arrl.net>
13313         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
13314         (pdp11_insn_cost): New function.
13315         (pdp11_md_asm_adjust): New function.
13316         (TARGET_INVALID_WITHIN_DOLOOP): Define.
13317         (pdp11_rtx_costs): Update to match machine better.
13318         (output_addr_const_pdp11): Correct format mismatch warnings.
13319         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
13320         * config/pdp11/pdp11.md: General change to add base_cost and/or
13321         length attributes for use by new pdp11_insn_cost function.
13322         (MIN_BRANCH): Correct definition.
13323         (MIN_SOB): Ditto.
13324         (doloop_end): Use standard pattern name for looping pattern.
13325         (doloop_end_nocc): New.
13326         (movsf): Add another constraint alternative.
13327         (zero_extendqihi2): Add constraint alternatives for not in place
13328         extend.
13329         (zero_extendhisi2): Remove.
13330         (shift patterns): Add CC handling variants.
13331         (bswaphi2): New.
13332         (bswapsi2): New.
13333         (rothi3): New.
13334         (define_peephole2): New peephole to recognize mov that sets CC for
13335         subsequent test.
13337 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
13339         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
13340         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
13341         wi::add.
13343 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
13345         PR c/86420
13346         * real.c (real_nextafter): Return true if result is denormal.
13348 2018-07-09  Martin Liska  <mliska@suse.cz>
13350         * common.opt: Add back wrongly removed attribute.
13352 2018-07-09  Richard Biener  <rguenther@suse.de>
13354         PR debug/86413
13355         * dwarf2out.c (gen_block_die): For an early generated DIE
13356         always output high/low PC attributes.
13358 2018-07-09  Tom de Vries  <tdevries@suse.de>
13360         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
13361         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
13362         onto VAR_DECL with abstract origin.
13364 2018-07-07  Jim Wilson  <jimw@sifive.com>
13366         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
13368 2018-07-07  Tom de Vries  <tdevries@suse.de>
13370         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
13372 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
13374         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
13375         overflow_type.
13376         (combine_bound): Use wide-int overflow calculation instead of
13377         rolling our own.
13378         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
13379         overflow_type.
13380         * fold-const.c (int_const_binop_2): Same.
13381         (extract_muldiv_1): Same.
13382         (fold_div_compare): Same.
13383         (fold_abs_const): Same.
13384         * match.pd: Same.
13385         * poly-int.h (add): Same.
13386         (sub): Same.
13387         (neg): Same.
13388         (mul): Same.
13389         * predict.c (predict_iv_comparison): Same.
13390         * profile-count.c (slow_safe_scale_64bit): Same.
13391         * simplify-rtx.c (simplify_const_binary_operation): Same.
13392         * tree-chrec.c (tree_fold_binomial): Same.
13393         * tree-data-ref.c (split_constant_offset_1): Same.
13394         * tree-if-conv.c (idx_within_array_bound): Same.
13395         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
13396         * tree-ssa-phiopt.c (minmax_replacement): Same.
13397         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
13398         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
13399         * vr-values.c (vr_values::adjust_range_with_scev): Same.
13400         * wide-int.cc (wi::add_large): Same.
13401         (wi::mul_internal): Same.
13402         (wi::sub_large): Same.
13403         (wi::divmod_internal): Same.
13404         * wide-int.h: Change overflow type to overflow_type for neg, add,
13405         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
13406         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
13407         mul_internal, divmod_internal.
13408         (overflow_type): New enum.
13409         (accumulate_overflow): New.
13411 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
13413         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
13414         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
13416 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
13418         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
13419         argument is checked for zero before entering loop, avoid checking again.
13421 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
13423         * gimplify.h (generic_expr_could_trap_p): Set as global function.
13424         * gimplify.h (generic_expr_could_trap_p): Likwise.
13425         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
13427 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
13429         PR tree-optimization/86401
13430         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
13431         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
13432         (fold_bit_and_mask): ... here.  New helper function for match.pd.
13433         * fold-const.h (fold_bit_and_mask): Declare.
13434         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
13436 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
13438         PR target/86324
13439         * target.def (translate_mode_attribute): New hook.
13440         * targhooks.h (default_translate_mode_attribute): Declare.
13441         * targhooks.c (default_translate_mode_attribute): New function.
13442         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
13443         * doc/tm.texi: Regenerate.
13444         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
13445         (rs6000_translate_mode_attribute): New function.
13447 2018-07-06  Paul Koning  <ni1d@arrl.net>
13449         * doc/md.texi (define_split): Document DONE and FAIL.
13450         (define_peephole2): Ditto.
13452 2018-07-05  Jeff Law  <law@redhat.com>
13454         PR tree-optimization/86010
13455         * tree-ssa-dse.c (compute_trims): More aggressively trim at
13456         both the head and tail of mem* and str* calls.
13458 2018-07-05  Jim Wilson  <jimw@sifive.com>
13460         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
13462 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
13464         * config/aarch64/aarch64-simd.md: correct flags text for
13465         MIN_EXPR replacement.
13467 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
13469         * configure: Regenerated.
13471 2018-07-05  Carl Love  <cel@us.ibm.com>
13473         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
13474         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
13475         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
13476         VSX_BUILTIN_DOUBLEL_V4SF.
13478 2018-07-05  Martin Sebor  <msebor@redhat.com>
13480         PR c++/86400
13481         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
13482         than its domain to compute its the upper bound of a char array.
13484 2018-07-05  Nathan Sidwell  <nathan@acm.org>
13486         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
13487         * doc/cpp.texi: Update comment.
13488         * doc/tm.texi: Rebuilt.
13489         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
13490         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
13491         * doc/extend.texi (Backwards Compatibility): Clarify it is system
13492         headers affected by extern "C".
13493         * system.h: Poison NO_IMPLICIT_EXTERN_C.
13494         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
13495         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
13496         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
13497         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
13498         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
13499         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
13500         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
13501         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
13502         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
13503         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
13504         config/sparc/sp64-elf.h, config/spu/spu.h,
13505         config/stormy16/stormy16.h, config/v850/v850.h,
13506         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
13507         define NO_IMPLICIT_EXTERN_C.
13508         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
13510 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13512         PR target/84711
13513         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
13514         instead of GET_MODE_SIZE when comparing Units.
13516 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13518         PR target/84711
13519         * rtlanal.c (set_noop_p): Constrain on mode change,
13520         include hard-reg-set.h
13522 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
13524         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
13526 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
13528         Revert
13529         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13531         PR sanitizer/84250
13532         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13533         libasan.
13534         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13536 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
13538         PR sanitizer/84250
13539         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
13540         libasan.
13541         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
13543 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
13545         PR middle-end/86380
13546         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
13548 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
13550         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
13551         neg_*_op* variables.
13553 2018-07-04  Martin Liska  <mliska@suse.cz>
13555         * tree-switch-conversion.c: Define
13556         max_ratio_for_speed and max_ratio_for_size constants.
13558 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13559             Martin Liska  <mliska@suse.cz>
13561         PR middle-end/66240
13562         PR target/45996
13563         PR c/84100
13564         * common.opt: Rename align options with 'str_' prefix.
13565         * common/config/i386/i386-common.c (set_malign_value): New
13566         function.
13567         (ix86_handle_option): Use it to set -falign-* options/
13568         * config/aarch64/aarch64-protos.h (struct tune_params): Change
13569         type from int to string.
13570         * config/aarch64/aarch64.c: Update default values from int
13571         to string.
13572         * config/alpha/alpha.c (alpha_override_options_after_change):
13573         Likewise.
13574         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
13575         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13576         max skip conditionally.
13577         * config/i386/freebsd.h (SUBALIGN_LOG): New.
13578         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13579         max skip conditionally.
13580         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13581         max skip conditionally.
13582         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
13583         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13584         max skip conditionally.
13585         * config/i386/i386.c (struct ptt): Change type from int to
13586         string.
13587         (ix86_default_align): Set default values.
13588         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
13589         max skip conditionally.
13590         * config/i386/iamcu.h (SUBALIGN_LOG): New.
13591         (ASM_OUTPUT_MAX_SKIP_ALIGN):
13592         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
13593         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13594         max skip conditionally.
13595         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
13596         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
13597         * config/i386/x86-64.h (SUBALIGN_LOG): New.
13598         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
13599         max skip conditionally.
13600         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13601         * config/ia64/ia64.c (ia64_option_override): Set default values
13602         for alignment options.
13603         * config/m68k/m68k.c: Handle new str_align_* options.
13604         * config/mips/mips.c (mips_set_compression_mode): Change
13605         type of constants.
13606         (mips_option_override): Set default values for options.
13607         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
13608         Likewise.
13609         * config/rs6000/rs6000.c (rs6000_option_override_internal):
13610         Likewise.
13611         * config/rx/rx.c (rx_option_override): Likewise.
13612         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
13613         (LABEL_ALIGN): Use align_labels_log.
13614         (LOOP_ALIGN): Use align_loops_align.
13615         * config/s390/s390.c (s390_asm_output_function_label): Use new
13616         macros.
13617         * config/sh/sh.c (sh_override_options_after_change):
13618         Change type of constants.
13619         * config/spu/spu.c (spu_sched_init): Likewise.
13620         * config/sparc/sparc.c (sparc_option_override): Set default
13621         values for options.
13622         * config/visium/visium.c (visium_option_override): Likewise.
13623         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
13624         emit p2align format with last argument if it's not needed.
13625         * doc/invoke.texi: Document extended format of -falign-*.
13626         * final.c: Use align_labels alignment.
13627         * flags.h (struct target_flag_state): Change type to use
13628         align_flags.
13629         (struct align_flags_tuple): New.
13630         (struct align_flags): Likewise.
13631         (align_loops_log): Redefine macro to use new types.
13632         (align_loops_max_skip): Redefine macro to use new types.
13633         (align_jumps_log): Redefine macro to use new types.
13634         (align_jumps_max_skip): Redefine macro to use new types.
13635         (align_labels_log): Redefine macro to use new types.
13636         (align_labels_max_skip): Redefine macro to use new types.
13637         (align_functions_log): Redefine macro to use new types.
13638         (align_loops): Redefine macro to use new types.
13639         (align_jumps): Redefine macro to use new types.
13640         (align_labels): Redefine macro to use new types.
13641         (align_functions): Redefine macro to use new types.
13642         (align_functions_max_skip): Redefine macro to use new types.
13643         (align_loops_value): New macro.
13644         (align_jumps_value): New macro.
13645         (align_labels_value): New macro.
13646         (align_functions_value): New macro.
13647         * function.c (invoke_set_current_function_hook): Propagate
13648         alignment values from flags to global variables default in
13649         topleev.h.
13650         * ipa-icf.c (sem_function::equals_wpa): Use
13651         cl_optimization_option_eq instead of memcmp.
13652         * lto-streamer.h (cl_optimization_stream_out): Support streaming
13653         of string types.
13654         (cl_optimization_stream_in): Likewise.
13655         * optc-save-gen.awk: Support strings in cl_optimization.
13656         * opth-gen.awk: Likewise.
13657         * opts.c (finish_options): Remove error checking of invalid
13658         value ranges.
13659         (MAX_CODE_ALIGN): Remove.
13660         (MAX_CODE_ALIGN_VALUE): Likewise.
13661         (parse_and_check_align_values): New function.
13662         (check_alignment_argument): Likewise.
13663         (common_handle_option): Use check_alignment_argument.
13664         * opts.h (parse_and_check_align_values): Declare.
13665         * toplev.c (init_alignments): Remove.
13666         (read_log_maxskip): New.
13667         (parse_N_M): Likewise.
13668         (parse_alignment_opts): Likewise.
13669         (backend_init_target): Remove usage of init_alignments.
13670         * toplev.h (parse_alignment_opts): Declare.
13671         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
13672         argument.
13673         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
13674         * tree.c (cl_option_hasher::equal): New.
13675         * varasm.c: Use new global macros.
13677 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
13679         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
13680         Use a simpler align directive also if MAXSKIP = ALIGN-1.
13681         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13682         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13683         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13684         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
13685         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
13686         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
13687         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
13688         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13689         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13690         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13691         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
13693 2018-07-04  Martin Liska  <mliska@suse.cz>
13694             Jonathan Wakely  <jwakely@redhat.com>
13696         * coverage.c: Use correct type.
13697         * doc/invoke.texi: Language correction.
13699 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
13701         PR target/85620
13702         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
13703         ENDBRANCH for non-tail call which may return via indirect branch.
13704         * doc/extend.texi: Document indirect_return attribute.
13706 2018-07-03  Martin Sebor  <msebor@redhat.com>
13708         PR tree-optimization/86274
13709         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
13710         precondition.
13711         (format_floating): Correct handling of infinities and NaNs.
13713 2018-07-03  Martin Sebor  <msebor@redhat.com>
13715         * print-tree.c (print_real_cst): New function.
13716         (print_node_brief): Call it.
13717         (print_node): Ditto.
13719 2018-07-03  Jeff Law  <law@redhat.com>
13721         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
13722         into a single pattern.
13724         * config/h8300/h8300.md (ors code_iterator): New.
13725         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
13726         a single pattern and single splitter.
13727         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
13728         (iorqi3_1, xorqi3_1): Likewise.
13729         (iorqi3, xorqi3 expanders): Similarly.
13731         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
13732         (movmd_internal) into a single pattern using the P mode iterator.
13733         (movmd splitters): Similarly.
13734         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
13735         (movsd splitters): Similarly.
13737         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
13738         ADDB, ADDW and ADDL into a single ADD attribute which selects the
13739         right table based on the size of the operand.
13740         * config/h8300/h8300.md (length_table): Corresponding changes. All
13741         references to "addb", "addw" and "addl" changed to "add".
13742         (btst patterns): Merge two variants into a single pattern.
13743         (tstqi, tsthi): Likewise.
13744         (addhi3_incdec, addsi3_incdec): Likewise.
13745         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
13746         (mulhi3, mulsi3): Likewise.
13747         (udivhi3, udivsi3): Likewise.
13748         (divhi3, divsi3): Likewise.
13749         (andorqi3, andorhi3, andorsi3): Likewise.
13751 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
13753         PR target/85694
13754         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
13755         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
13757 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13759         PR tree-optimization/85694
13760         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
13761         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
13762         UNSPEC_URHADD.
13763         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
13764         (<u>avg<mode>3_ceil): New patterns.
13766 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
13768         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
13769         scan-tree-dump directive.
13770         * gcc.dg/vect/slp-perm-2.c: Likewise.
13771         * gcc.dg/vect/slp-perm-3.c: Likewise.
13772         * gcc.dg/vect/slp-perm-5.c: Likewise.
13773         * gcc.dg/vect/slp-perm-6.c: Likewise.
13774         * gcc.dg/vect/slp-perm-7.c: Likewise.
13775         * gcc.dg/vect/slp-perm-8.c: Likewise.
13777 2018-07-03  Marek Polacek  <polacek@redhat.com>
13779         PR middle-end/86202
13780         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
13782 2018-07-03  Richard Biener  <rguenther@suse.de>
13784         PR ipa/86389
13785         * tree-ssa-structalias.c (find_func_clobbers): Properly
13786         handle indirect calls.
13788 2018-07-03  Jeff Law  <law@redhat.com>
13790         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
13791         (shifts): New code iterator.
13792         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
13793         expander.  Fix HImode handling on H8/SX.
13794         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
13795         (subqi3, subhi3, subsi3 expanders): Likewise.
13796         (andqi3, andhi3, andsi3 expanders): Likewise.
13797         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
13798         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
13799         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
13800         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
13801         (zero_extendqihi2, zero_extendqisi2): Likewise.
13802         (extendqihi2, extendqisi2): Likewise.
13803         (rotlqi3, rotlhi3, rotlsi3): Likewise.
13804         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
13805         (rotlqi3_1, rotlhi3_1): Likewise.
13806         (logicalhi3_sn, logicalsi3_sn): Likewise.
13807         (logicalhi3, logicalsi3): Likewise.
13809 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13811         * tree-vect-patterns.c (vect_recog_rotate_pattern)
13812         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
13813         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
13814         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
13815         type to append_pattern_def_seq instead of creating a stmt_vec_info
13816         directly.
13817         (build_mask_conversion): Likewise.  Remove vinfo argument.
13818         (vect_add_conversion_to_patterm): Likewise, renaming to...
13819         (vect_add_conversion_to_pattern): ...this.
13820         (vect_recog_mask_conversion_pattern): Update call to
13821         build_mask_conversion.  Pass the vector type to
13822         append_pattern_def_seq here too.
13823         (vect_recog_gather_scatter_pattern): Update call to
13824         vect_add_conversion_to_pattern.
13826 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13828         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
13829         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
13830         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
13831         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
13832         STMT_VINFO_PATTERN_DEF_SEQ to null here.
13833         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
13834         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
13835         append_pattern_def_seq instead of new_pattern_def_seq.
13836         (vect_recog_divmod_pattern): Do both of the above.
13837         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
13838         is null.
13840 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13842         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
13843         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
13844         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
13845         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
13846         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
13847         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
13848         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
13849         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
13850         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
13851         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
13852         parameter with a single stmt_vec_info.
13853         (vect_recog_func_ptr): Likewise.
13854         (vect_recog_gather_scatter_pattern): Likewise, folding in...
13855         (vect_try_gather_scatter_pattern): ...this.
13856         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
13857         the stmt_vec_info of the statement to be matched.  Don't clear
13858         STMT_VINFO_RELATED_STMT.
13859         (vect_pattern_recog): Update call accordingly.
13861 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13863         PR tree-optimization/85694
13864         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
13865         (uavgM3_ceil): Document new optabs.
13866         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
13867         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
13868         functions.
13869         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
13870         (savg_ceil_optab): New optabs.
13871         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
13872         (vect_vect_recog_func_ptrs): Add it.
13873         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
13874         constant directly from the associated lhs.
13876 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13878         * tree-vect-patterns.c (vect_split_statement): New function.
13879         (vect_convert_input): Use it to try to split an existing cast.
13881 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13883         * poly-int.h (print_hex): New function.
13884         * dumpfile.h (dump_dec, dump_hex): Declare.
13885         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
13886         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
13887         min_input_precision, operation_precision and operation_sign.
13888         * tree-vect-patterns.c (vect_get_range_info): New function.
13889         (vect_same_loop_or_bb_p, vect_single_imm_use)
13890         (vect_operation_fits_smaller_type): Delete.
13891         (vect_look_through_possible_promotion): Add an optional
13892         single_use_p parameter.
13893         (vect_recog_over_widening_pattern): Rewrite to use new
13894         stmt_vec_info infomration.  Handle one operation at a time.
13895         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
13896         (vect_truncatable_operation_p, vect_set_operation_type)
13897         (vect_set_min_input_precision): New functions.
13898         (vect_determine_min_output_precision_1): Likewise.
13899         (vect_determine_min_output_precision): Likewise.
13900         (vect_determine_precisions_from_range): Likewise.
13901         (vect_determine_precisions_from_users): Likewise.
13902         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
13903         (vect_vect_recog_func_ptrs): Put over_widening first.
13904         Add cast_forwprop.
13905         (vect_pattern_recog): Call vect_determine_precisions.
13907 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
13909         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
13910         statements that have been replaced by further pattern statements.
13911         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
13913 2018-07-03  Richard Biener  <rguenther@suse.de>
13915         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
13916         always set *dt.  Dump vectype in vectype overload.
13917         * dumpfile.h (dump_gimple_expr): New function.
13918         (dump_gimple_expr_loc): Likewise.
13919         * dumpfile.c (dump_gimple_expr): New function.
13920         (dump_gimple_expr_loc): Likewise.
13922 2018-07-02  Jeff Law  <law@redhat.com>
13924         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
13925         the H8/300, H8/300H and H8/S variants into a single pattern.
13926         (movhi_h8300, movqi_h8300hs): Similarly.
13927         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
13928         (QHI mode iterator): New.
13930         * config/h8300/h8300.md: Remove trailing whitespace.
13932 2018-07-02  Jim Wilson  <jimw@sifive.com>
13934         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
13935         instead of emit_insn for interrupt returns.
13936         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
13937         (riscv_sret, riscv_uret): Likewise.
13939 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
13941         * pretty-print.c (selftest::test_pp_format): Move save and restore
13942         of quotes to class auto_fix_quotes, and add an instance.
13943         * selftest.c: Include "intl.h".
13944         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
13945         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
13946         * selftest.h (selftest::auto_fix_quotes): New class.
13948 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
13950         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
13951         (aarch64_sve_prepare_conditional_op): Remove.
13952         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
13953         Allow aarch64_simd_reg_or_zero as select operand; remove
13954         the aarch64_sve_prepare_conditional_op call.
13955         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
13956         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13957         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
13958         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
13959         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
13960         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
13961         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
13962         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
13963         and a splitters to match all of the *_any patterns.
13964         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
13966         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
13967         (SVE_COND_FP_BINARY_REV): Remove.
13968         (sve_int_op_rev, sve_fp_op_rev): New.
13969         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
13970         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
13971         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
13972         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
13973         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
13974         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
13975         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
13976         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
13977         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
13979         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
13980         Remove match_dup 1 from the inner unspec.
13981         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
13983         * config/aarch64/aarch64.md (movprfx): New attr.
13984         (length): Default movprfx to 8.
13985         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
13986         (*madd<SVE_I>, *msub<SVE_I): Likewise.
13987         (*<su>mul<SVE_I>3_highpart): Likewise.
13988         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
13989         (*v<ASHIFT><SVE_I>3): Likewise.
13990         (*<su><MAXMIN><SVE_I>3): Likewise.
13991         (*<su><MAXMIN><SVE_F>3): Likewise.
13992         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
13993         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
13994         (*div<SVE_F>4): Likewise.
13996 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
13998         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
13999         in dump string.
14001 2018-07-02  Richard Biener  <rguenther@suse.de>
14003         PR tree-optimization/86363
14004         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
14005         memset argument refers to a non-variable address.
14007 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
14009         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
14010         {PLUS,MINUS}_EXPR code to...
14011         (adjust_symbolic_bound): ...here,
14012         (combine_bound): ...here,
14013         (set_value_range_with_overflow): ...and here.
14015 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
14017         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
14018         code...
14019         (extract_range_from_abs_expr): ...here.
14021 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
14023         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
14024         -fno-omit-frame-pointer when not optimizing.
14026 2018-07-02  Martin Liska  <mliska@suse.cz>
14028         PR ipa/86279
14029         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
14030         (propagate_nothrow): Likewise.
14032 2018-07-02  Martin Liska  <mliska@suse.cz>
14034         PR ipa/86323
14035         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
14037 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
14039         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
14040         function in r262149, changing "loc" param from source_location to
14041         const dump_location_t &.
14042         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
14043         declaration, as above.
14045 2018-07-01  Paul Koning  <ni1d@arrl.net>
14047         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14048         -munit-asm, -mgnu-asm, -mdec-asm.
14049         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
14050         (pdp11_output_labelref): New.
14051         (pdp11_output_def): New.
14052         (pdp11_output_addr_vec_elt): New.
14053         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
14054         %# and %@ format codes.
14055         (pdp11_option_override): New.
14056         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
14057         (pdp11_output_ident): New.
14058         (pdp11_asm_named_section): New.
14059         (pdp11_asm_init_sections): New.
14060         (pdp11_file_start): New.
14061         (pdp11_file_end): New.
14062         (output_ascii): Use .ascii/.asciz for -mdec-asm.
14063         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
14064         %o, like %c but octal.
14065         (pdp11_option_override): New.
14066         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
14067         -mdec-asm.
14068         (DATA_SECTION_ASM_OP): Ditto.
14069         (READONLY_DATA_SECTION_ASM_OP): New.
14070         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
14071         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
14072         (ASM_OUTPUT_LABELREF): Ditto.
14073         (ASM_OUTPUT_DEF): Ditto.
14074         (ASM_OUTPUT_EXTERNAL): New.
14075         (ASM_OUTPUT_SOURCE_FILENAME): New.
14076         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
14077         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
14078         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
14079         %# and %@ format codes.
14080         * config/pdp11/pdp11.opt (mgnu-asm): New.
14081         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
14082         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
14083         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
14085 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
14087         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
14088         dereferencing path[] beyond its length.
14089         (debug_path): New.
14090         (debug_all_paths): New.
14091         (rewire_first_differing_edge): New.
14092         (adjust_paths_after_duplication): New.
14093         (duplicate_thread_path): Call adjust_paths_after_duplication.
14094         Add new argument.
14095         (thread_through_all_blocks): Add new argument to
14096         duplicate_thread_path.
14098 2018-06-30  Jim Wilson  <jimw@sifive.com>
14100         * config/riscv/predicates.md (p2m1_shift_operand): New.
14101         (high_mask_shift_operand): New.
14102         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
14103         pattern using p2m1_shift_operand.
14104         (lshsi3_zero_extend_3+2): New combiner pattern using
14105         high_mask_shift_operand.
14107 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14109         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
14110         split out from...
14111         (vect_recog_rotate_pattern): ...here.
14112         (vect_convert_input): Try to insert casts of invariants in the
14113         preheader.
14114         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
14115         preheader to be empty.
14117 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14119         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
14120         vector type.  If given, install it in the new statement's
14121         STMT_VINFO_VECTYPE.
14122         (vect_element_precision): New function.
14123         (vect_unpromoted_value): New struct.
14124         (vect_unpromoted_value::vect_unpromoted_value): New function.
14125         (vect_unpromoted_value::set_op): Likewise.
14126         (vect_look_through_possible_promotion): Likewise.
14127         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
14128         (vect_widened_op_tree, vect_convert_input): Likewise.
14129         (vect_convert_inputs, vect_convert_output): Likewise.
14130         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
14131         to handle the optional cast of the multiplication result and
14132         vect_widened_op_tree to detect the widened multiplication itself.
14133         Do not require the input and output of promotion casts to have
14134         the same sign, but base the signedness of the operation on the
14135         input rather than the result.  If the pattern includes two
14136         promotions, check that those promotions have the same sign.
14137         Do not restrict the MULT_EXPR handling to a double-width result;
14138         handle quadruple-width results and wider.  Use vect_convert_inputs
14139         to convert the inputs to the common type.
14140         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
14141         to handle the optional cast of the ABS result.  Also allow a sign
14142         change or a sign extension between the ABS and MINUS.
14143         Use vect_widened_op_tree to detect the widened subtraction and use
14144         vect_convert_inputs to convert the inputs to the common type.
14145         (vect_handle_widen_op_by_const): Delete.
14146         (vect_recog_widen_op_pattern): New function.
14147         (vect_recog_widen_mult_pattern): Use it.
14148         (vect_recog_widen_shift_pattern): Likewise.
14149         (vect_recog_widen_sum_pattern): Use
14150         vect_look_through_possible_promotion to handle the promoted
14151         PLUS_EXPR operand.
14153 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14155         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
14156         the containing gimple_seq *.
14157         * gimple-iterator.h (gsi_for_stmt): Declare it.
14158         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
14159         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
14160         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
14161         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
14162         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
14163         checks.
14164         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
14165         split out from...
14166         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
14167         statement being replaced is part of an existing pattern
14168         definition sequence, inserting the new pattern statements before
14169         the original one.
14170         (vect_pattern_recog_1): Don't return a bool.  If the statement
14171         is already part of a pattern, instead apply pattern matching
14172         to the pattern definition statements.  Don't clear the
14173         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
14174         (vect_pattern_recog): Don't break after the first match;
14175         continue processing the pattern definition statements instead.
14176         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
14178 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14180         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
14181         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
14182         (vect_recog_widen_sum_pattern): Use it.
14184 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14186         * tree-vect-loop.c (vectorizable_reduction): Assert that the
14187         phi is not a pattern statement and has not been replaced by
14188         a pattern statement.
14189         * tree-vect-patterns.c (type_conversion_p): Don't check
14190         STMT_VINFO_IN_PATTERN_P.
14191         (vect_recog_vector_vector_shift_pattern): Likewise.
14192         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
14193         the pattern statement rather than the original statement; check
14194         directly for a WIDEN_MULT_EXPR here.
14195         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
14196         vect_is_simple_use to return the pattern statement rather
14197         than the original statement; use is_pattern_stmt_p to check
14198         for such a pattern statement.
14199         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
14200         to return the pattern statement rather than the original statement;
14201         don't do the same transformation here.
14202         (vect_is_simple_use): If the defining statement has been replaced
14203         by a pattern statement, return the pattern statement instead.
14204         Remove the corresponding (local) transformation from the vectype
14205         overload.
14207 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
14209         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
14210         end and default to null.
14211         * tree-vect-loop.c (vect_create_epilog_for_reduction)
14212         (vectorizable_reduction): Update calls accordingly, dropping the
14213         gimple ** argument if the passed-back statement isn't needed.
14214         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
14215         (vect_recog_rotate_pattern): Likewise.
14216         (vect_recog_mask_conversion_pattern): Likewise.
14217         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
14218         (vect_mask_constant_operand_p): Likewise.
14219         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
14220         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
14221         (get_group_load_store_type, get_load_store_type): Likewise.
14222         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
14223         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
14224         (vectorizable_conversion, vectorizable_assignment): Likewise.
14225         (vectorizable_shift, vectorizable_operation): Likewise.
14226         (vectorizable_store, vect_is_simple_cond): Likewise.
14227         (vectorizable_condition, vectorizable_comparison): Likewise.
14228         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
14229         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
14230         and move it to the end.  Cope with null def_stmt_outs.
14232 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
14234         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
14236 2018-06-29  Jeff Law  <law@redhat.com>
14238         * config/v850/v850.c (v850_legitimate_address_p): Handle large
14239         displacements for TARGET_V850E2V3 and newer.
14240         (TARGET_LRA_P): Remove.  Defaults to LRA now.
14241         * config/v850/v850.md (sign23byte_load): Remove.
14242         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
14243         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
14245 2018-06-29  Martin Liska  <mliska@suse.cz>
14247         PR lto/85759
14248         * coverage.c (coverage_init): Mangle full path name.
14249         * doc/invoke.texi: Document the change.
14250         * gcov-io.c (mangle_path): New.
14251         * gcov-io.h (mangle_path): Likewise.
14252         * gcov.c (mangle_name): Use mangle_path for path mangling.
14254 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14256         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
14257         if starting source register is not even.
14259 2018-06-29  Martin Liska  <mliska@suse.cz>
14261         PR tree-optimization/86263
14262         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
14263         Make edge redirection.
14265 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
14267         * dumpfile.c (dump_loc): Add indentation based on scope depth.
14268         (dump_scope_depth): New variable.
14269         (get_dump_scope_depth): New function.
14270         (dump_begin_scope): New function.
14271         (dump_end_scope): New function.
14272         * dumpfile.h (get_dump_scope_depth): New declaration.
14273         (dump_begin_scope): New declaration.
14274         (dump_end_scope): New declaration.
14275         (class auto_dump_scope): New class.
14276         (AUTO_DUMP_SCOPE): New macro.
14277         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
14278         AUTO_DUMP_SCOPE.
14280 2018-06-29  Richard Biener  <rguenther@suse.de>
14282         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
14283         compute_all_dependences succeeds.
14284         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
14285         exceed --param loop-max-datarefs-for-datadeps.
14287 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
14289         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
14291 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
14293         PR target/86348
14294         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
14295         alternative 0 in preferred_for_speed attribute.
14297 2018-06-28  Paul Koning  <ni1d@arrl.net>
14299         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
14300         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
14301         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
14302         * config/pdp11/pdp11.md: Correct "length" attribute calculation
14303         for shift insn patterns.
14305 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
14307         * cgraph.c (cgraph_node::get_body): Replace assignments to
14308         "dump_file" with calls to set_dump_file.
14309         * dumpfile.c (alt_dump_file): Make static, and group with...
14310         (alt_flags): ...this definition.
14311         (dumps_are_enabled): New variable.
14312         (refresh_dumps_are_enabled): New function.
14313         (set_dump_file): New function.
14314         (set_alt_dump_file): New function.
14315         (gcc::dump_manager::dump_start): Replace assignments to
14316         "dump_file" and "alt_dump_file" with calls to set_dump_file and
14317         set_alt_dump_file.
14318         (gcc::dump_manager::dump_finish): Likewise.
14319         * dumpfile.h (alt_dump_file): Delete decl.
14320         (dumps_are_enabled): New variable decl.
14321         (set_dump_file): New function decl.
14322         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
14323         global.
14324         * tree-nested.c (lower_nested_functions): Replace assignments to
14325         "dump_file" with calls to set_dump_file.
14327 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14329         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
14330         goto_locus of each outgoing edge of each basic block.
14332 2018-06-28  Richard Biener  <rguenther@suse.de>
14334         * dwarf2out.c (decl_scope_table): Remove.
14335         (push_decl_scope): Likewise.
14336         (pop_decl_scope): Likewise.
14337         (gen_type_die_for_member): Do not call push/pop_decl_scope.
14338         (gen_struct_or_union_type_die): Likewise.
14339         (gen_tagged_type_die): Likewise.
14340         (dwarf2out_init): Do not initialize decl_scope_table.
14341         (dwarf2out_c_finalize): Do not free it.
14343 2018-06-28  Richard Biener  <rguenther@suse.de>
14345         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
14346         deciding whether to not re-use a DIE.
14348 2018-06-28  Richard Biener  <rguenther@suse.de>
14350         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
14351         DW_AT_abstract_origin attribute.
14353 2018-06-28  Martin Liska  <mliska@suse.cz>
14355         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
14356         Use newly introduced constants.
14357         * tree-switch-conversion.h (struct jump_table_cluster):
14358         Define max_ratio_for_size and max_ratio_for_speed.
14360 2018-06-28  Martin Liska  <mliska@suse.cz>
14362         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
14363         Add new checking assert to catch invalid state.
14364         (jump_table_cluster::can_be_handled): Handle single case
14365         clusters.
14366         (jump_table_cluster::is_beneficial): Bail out for such case.
14367         (bit_test_cluster::find_bit_tests):
14368         Add new checking assert to catch invalid state.
14369         (bit_test_cluster::can_be_handled): Handle single case
14370         clusters.
14371         (bit_test_cluster::is_beneficial): Bail out for such case.
14372         (switch_decision_tree::analyze_switch_statement):
14373         Fix comment.
14375 2018-06-28  Martin Liska  <mliska@suse.cz>
14377         * common.opt: Introduce -completion option.
14378         * gcc.c (driver_handle_option): Handle it.
14379         (driver::main): Print completions if completion
14380         is set.
14381         * opt-suggestions.c (option_proposer::get_completions):
14382         New function.
14383         (option_proposer::suggest_completion): Likewise.
14384         (option_proposer::find_param_completions): Likewise.
14385         (verify_autocompletions): Likewise.
14386         (test_completion_valid_options): Likewise.
14387         (test_completion_valid_params): Likewise.
14388         (in_completion_p): Likewise.
14389         (empty_completion_p): Likewise.
14390         (test_completion_partial_match): Likewise.
14391         (test_completion_garbage): Likewise.
14392         (opt_proposer_c_tests): Likewise.
14393         * opt-suggestions.h: Declare new functions.
14394         * opts.c (common_handle_option): Handle OPT__completion_.
14395         * selftest-run-tests.c (selftest::run_tests): Add
14396         opt_proposer_c_tests.
14397         * selftest.c (assert_str_startswith): New.
14398         * selftest.h (assert_str_startswith): Likewise.
14399         (opt_proposer_c_tests): New.
14400         (ASSERT_STR_STARTSWITH): Likewise.
14402 2018-06-28  Martin Liska  <mliska@suse.cz>
14404         * Makefile.in: Add opt-suggestions.o.
14405         * gcc-main.c: Include opt-suggestions.h.
14406         * gcc.c (driver::driver): Likewise.
14407         (driver::~driver): Remove m_option_suggestions.
14408         (driver::build_option_suggestions): Moved to option_proposer.
14409         (driver::suggest_option): Likewise.
14410         (driver::handle_unrecognized_options): Use option_proposer.
14411         * gcc.h (class driver): Add new memver m_option_proposer.
14412         * opt-suggestions.c: New file.
14413         * opt-suggestions.h: New file.
14415 2018-06-28  Martin Liska  <mliska@suse.cz>
14417         * vec.h (class auto_string_vec): New (moved from auto_argvec).
14418         (auto_string_vec::~auto_string_vec): Likewise.
14420 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
14422         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
14423         prevent_decl_creation_for_types fields up and add reset_location field.
14424         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
14425         statement if id->reset_location is true.
14426         (copy_edges_for_bb): Do not set goto_locus on the new edges if
14427         id->reset_location is true.
14428         (copy_phis_for_bb): Force input_location on the arguments if
14429         id->reset_location is true.
14430         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
14431         is set on the function to be inlined.
14433 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
14435         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
14437 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
14439         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
14440         registers for Pmode.
14441         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
14442         hard registers for the clobbered pseudo.
14444 2018-06-27  Paul Koning  <ni1d@arrl.net>
14446         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
14447         mutually exclusive options.
14448         * config/pdp11/constraints.md (h): New constraint.
14449         (O): Update definition to match shift code generation.
14450         (D): New constraint.
14451         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
14452         (CCFP): Remove.
14453         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
14454         function.
14455         (output_jump): Change arguments.
14456         (pdp11_fixed_cc_regs): New function.
14457         (pdp11_cc_mode): Ditto.
14458         (pdp11_expand_shift): Ditto.
14459         (pdp11_assemble_shift): Ditto.
14460         (pdp11_small_shift): Ditto.
14461         (pdp11_branch_cost): Remove.
14462         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
14463         from output.
14464         (pdp11_register_move_cost): Update for CC registers.
14465         (pdp11_rtx_costs): Add case for LSHIFTRT.
14466         (pdp11_output_jump): Add CCNZ mode conditional branches.
14467         (notice_update_cc_on_set): Remove.
14468         (pdp11_cc_mode): New function.
14469         (simple_memory_operand): Correct pre/post decrement case.
14470         (no_side_effect_operand): New function.
14471         (pdp11_regno_reg_class): Add CC_REGS class.
14472         (pdp11_fixed_cc_regs): New function.
14473         (pdp11_small_shift): New function.
14474         (pdp11_expand_shift): New function to expand shift insns.
14475         (pdp11_assemble_shift): New function to output shifts.
14476         (pdp11_branch_cost): Remove.
14477         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
14478         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
14479         (WCHAR_TYPE): Ditto.
14480         (PTRDIFF_TYPE): Ditto.
14481         (ADJUST_INSN_LENGTH): New macro.
14482         (FIXED_REGISTERS): Add CC registers.
14483         (CALL_USED_REGISTERS): Ditto.
14484         (reg_class): Ditto.
14485         (REG_CLASS_NAMES): Ditto.
14486         (REG_CLASS_CONTENTS): Ditto.
14487         (SELECT_CC_MODE): Use new function.
14488         (TARGET_FLAGS_REGNUM): New macro.
14489         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
14490         (cc0_reg_rtx): Remove.
14491         (CC_STATUS_MDEP): Remove.
14492         (CC_STATUS_MDEFP_INIT): Remove.
14493         (CC_IN_FPU): Remove.
14494         (NOTICE_UPDATE_CC): Remove.
14495         (REGISTER_NAMES): Add CC registers.
14496         (BRANCH_COST): Change to constant 1.
14497         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
14498         handling.
14499         * config/pdp11/pdp11.opt (mbcopy): Remove.
14500         (mbcopy-builtin): Remove.
14501         (mbranch-cheap): Remove.
14502         (mbranch-expensive): Remove.
14503         * config/pdp11/predicates.md (expand_shift_operand): Update to
14504         match shift code generation.
14505         (ccnz_operator): New predicate.
14506         * doc/invoke.texi (PDP-11 Options): Remove deleted options
14507         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
14508         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
14509         exclusive options.
14510         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
14511         description of O constraint.
14513 2018-06-27  Jeff Law  <law@redhat.com>
14514             Austin Law  <austinklaw@gmail.com>
14516         * config/v850/v850.md (addsi3_set_flags): New pattern.
14517         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
14518         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
14519         (zero_extendhisi2_v850_set_flags): Likewise.
14520         (zero_extendqisi2_v850_set_flags): Likewise.
14521         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
14522         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
14523         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
14525         * config/v850/v850-protos.h (notice_update_cc): Remove.
14526         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
14527         (v850_print_operand): Handle 'D' and "d".
14528         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
14529         Add handling of arithmetic/logical operations compared against zero.
14530         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
14531         Do not look at v850_compare_op, instead get mode from last argument.
14532         (v850_gen_compare): Remove
14533         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
14534         after reload for prologue insns.
14535         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
14536         patterns.
14537         (construct_save_jarl): Likewise.
14538         (TARGET_FLAGS_REGNUM): Define.
14539         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
14540         (NOTICE_UPDATE_CC): Remove.
14541         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
14542         than cc0.  Conditionalize on reload_completed.
14543         (cmpsi_insn, setfcc_insn): Likewise.
14544         (tst1 splitter): Turn into define_and_split which sets the flags
14545         after reload.
14546         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
14547         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
14548         (cstoresf4, cstoredf4): Clobber the flags.
14549         (cmpsi, cmpsf, cmpdf): Remove expanders.
14550         (setf_insn): Remove pattern.
14551         (addsi3): Turn into define_and_split which clobbers the flags after
14552         reload and a suitable pattern (addsi3_clobber_flags) for use after
14553         reload.
14554         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
14555         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
14556         (ashrsi3, ashrsi3_v850e2): Likewise.
14557         (bins): Clobber the flags.
14558         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
14559         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
14560         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
14561         (call_value_internal_short, call_value_internal_long): Likewise.
14562         (callt_save_interrupt, callt_return_interrupt): Likewise.
14563         (save_interrupt, return_interrupt): Likewise.
14564         (callt_save_all_interrupt, save_all_interrupt): Likewise.
14565         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
14566         (restore_all_interrupt, _restore_all_interrupt): Likewise.
14567         (All FP comparisons): Only allow after reload has completed.
14568         (trfsr): Likewise.
14569         (divh, divhu): Tweak output template.
14570         (branch_z_normal, branch_z_invert): Remove
14571         (branch_nz_normal, branch_nz_invert): Likewise.
14572         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
14574         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
14575         * config/v850/v850.c (notice_update_cc): Remove.
14576         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
14577         (CC_NO_CARRY): Likewise.
14578         (NOTICE_UPDATE_CC): Define to nothing.
14579         * config/v850/v850.md: Remove block comment on cc0 handling
14580         Remove "cc" attribute from all patterns.  Remove cc_status handling
14581         from all patterns.  Minor formatting fixes.
14583 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14585         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
14586         (cortex-a76.cortex-a55): Likewise.
14587         * config/aarch64/aarch64-tune.md: Regenerate.
14588         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
14589         cortex-a76.cortex-a55.
14591 2018-06-27  Jeff Law  <law@redhat.com>
14593         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
14594         (MULTILIB_DIRNAMES): Similarly.
14596 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
14598         * gimple.h (gimple_return_retbnd): Delete.
14599         (gimple_return_set_retbnd): Likewise.
14600         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
14601         gimple_return_set_retbnd.
14602         * gimple-pretty-print.c (dump_gimple_return): Remove call to
14603         gimple_return_retbnd and adjust.
14604         * tree-inline.h (struct copy_body_data): Remove retbnd field.
14605         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
14606         Explicitly return NULL in a couple more cases.  Move assertion
14607         on debug statements and remove unreachable code.
14608         (reset_debug_binding): Do not test id->retbnd.
14609         (expand_call_inline): Do not set it.
14611 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
14613         * configure.ac: Add --disable-gcov option.
14614         * configure: Regenerate.
14615         * Makefile.in: Honour @enable_gcov@.
14616         * doc/install.texi: Document --disable-gcov.
14618 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14620         * config/arm/arm-cpus.in (cortex-a76): New entry.
14621         (cortex-a76.cortex-a55): Likewise.
14622         * config/arm/arm-tables.opt: Regenerate.
14623         * config/arm/arm-tune.md: Likewise.
14624         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
14625         * doc/invoke.texi (ARM Options): Document cortex-a76 and
14626         cortex-a76.cortex-a55.
14628 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
14630         PR target/85769
14631         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
14633 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
14635         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
14636         comment.
14637         (EPILOGUE_USES): Likewise.
14639 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
14641         * tree-inline.c (remap_location): New function extracted from...
14642         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
14643         (copy_phis_for_bb): ...here.  Call remap_location.
14644         (copy_cfg_body): Adjust call to copy_edges_for_bb.
14646 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
14648         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
14649         unaligned vsx for 16B memset.
14651 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
14653         PR target/86285
14654         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
14655         ieee128_float_type_node to long_double_type_node unless
14656         TARGET_LONG_DOUBLE_128 is set.
14658 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
14660         * cfgloop.c (get_loop_location): Convert return type from
14661         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
14662         by implicit construction from rtx_insn *, and using
14663         dump_user_location_t::from_function_decl for the fallback case.
14664         * cfgloop.h (get_loop_location): Convert return type from
14665         location_t to dump_user_location_t.
14666         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
14667         dump_printf_loc to pass in a dump_location_t rather than a
14668         location_t, via the gimple stmt.
14669         * coverage.c (get_coverage_counts): Update calls to
14670         dump_printf_loc to pass in dump_location_t rather than a
14671         location_t.
14672         * doc/optinfo.texi (Dump types): Convert example of
14673         dump_printf_loc from taking "locus" to taking "insn".  Update
14674         description of the "_loc" calls to cover dump_location_t.
14675         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
14676         "selftest.h".
14677         (dump_user_location_t::dump_user_location_t): New constructors,
14678         from gimple *stmt and rtx_insn *.
14679         (dump_user_location_t::from_function_decl): New function.
14680         (dump_loc): Make static.
14681         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
14682         const dump_location_t &.
14683         (dump_generic_expr_loc): Delete.
14684         (dump_printf_loc): Convert param "loc" from location_t to
14685         const dump_location_t &.
14686         (selftest::test_impl_location): New function.
14687         (selftest::dumpfile_c_tests): New function.
14688         * dumpfile.h: Include "profile-count.h".
14689         (class dump_user_location_t): New class.
14690         (struct dump_impl_location_t): New struct.
14691         (class dump_location_t): New class.
14692         (dump_printf_loc): Convert 2nd param from source_location to
14693         const dump_location_t &.
14694         (dump_generic_expr_loc): Delete.
14695         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
14696         const dump_location_t &.
14697         * gimple-fold.c (fold_gimple_assign): Update call to
14698         dump_printf_loc to pass in a dump_location_t rather than a
14699         location_t, via the gimple stmt.
14700         (gimple_fold_call): Likewise.
14701         * gimple-loop-interchange.cc
14702         (loop_cand::analyze_iloop_reduction_var): Update for change to
14703         check_reduction_path.
14704         (tree_loop_interchange::interchange): Update for change to
14705         find_loop_location.
14706         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
14707         change in return-type of find_loop_location.
14708         (graphite_regenerate_ast_isl): Likewise.
14709         * graphite-optimize-isl.c (optimize_isl): Likewise.
14710         * graphite.c (graphite_transform_loops): Likewise.
14711         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
14712         pass in a dump_location_t rather than a location_t, via the
14713         gimple stmt.
14714         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
14715         * ipa.c (walk_polymorphic_call_targets): Likewise.
14716         * loop-unroll.c (report_unroll): Convert "locus" param from
14717         location_t to dump_location_t.
14718         (decide_unrolling): Update for change to get_loop_location's
14719         return type.
14720         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
14721         location_t to dump_user_location_t.
14722         (grid_find_single_omp_among_assignments_1): Updates calls to
14723         dump_printf_loc to pass in a dump_location_t rather than a
14724         location_t, via the gimple stmt.
14725         (grid_parallel_clauses_gridifiable): Convert "tloc" from
14726         location_t to dump_location_t.  Updates calls to dump_printf_loc
14727         to pass in a dump_location_t rather than a location_t, via the
14728         gimple stmt.
14729         (grid_inner_loop_gridifiable_p): Likewise.
14730         (grid_dist_follows_simple_pattern): Likewise.
14731         (grid_gfor_follows_tiling_pattern): Likewise.
14732         (grid_target_follows_gridifiable_pattern): Likewise.
14733         (grid_attempt_target_gridification): Convert initialization
14734         of local "grid" from memset to zero-initialization; FIXME: does
14735         this require C++11?  Update call to dump_printf_loc to pass in a
14736         optinfo_location rather than a location_t, via the gimple stmt.
14737         * profile.c (read_profile_edge_counts): Updates call to
14738         dump_printf_loc to pass in a dump_location_t rather than a
14739         location_t
14740         (compute_branch_probabilities): Likewise.
14741         * selftest-run-tests.c (selftest::run_tests): Call
14742         dumpfile_c_tests.
14743         * selftest.h (dumpfile_c_tests): New decl.
14744         * tree-loop-distribution.c (pass_loop_distribution::execute):
14745         Update for change in return type of find_loop_location.
14746         * tree-parloops.c (parallelize_loops): Likewise.
14747         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
14748         "locus" from location_t to dump_user_location_t.
14749         (canonicalize_loop_induction_variables): Likewise.
14750         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
14751         for change in return type of find_loop_location.
14752         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
14753         to dump_printf_loc to pass in a dump_location_t rather than a
14754         location_t, via the stmt.
14755         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
14756         Likewise.
14757         * tree-vect-loop-manip.c (find_loop_location): Convert return
14758         type from source_location to dump_user_location_t.
14759         (vect_do_peeling): Update for above change.
14760         (vect_loop_versioning): Update for change in type of
14761         vect_location.
14762         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
14763         from location_t to dump_user_location_t.
14764         (vect_estimate_min_profitable_iters): Update for change in type
14765         of vect_location.
14766         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
14767         location_t to dump_location_t.
14768         (vect_slp_bb): Update for change in type of vect_location.
14769         * tree-vectorizer.c (vect_location): Convert from source_location
14770         to dump_user_location_t.
14771         (try_vectorize_loop_1): Update for change in vect_location's type.
14772         (vectorize_loops): Likewise.
14773         (increase_alignment): Likewise.
14774         * tree-vectorizer.h (vect_location): Convert from source_location
14775         to dump_user_location_t.
14776         (find_loop_location): Convert return type from source_location to
14777         dump_user_location_t.
14778         (check_reduction_path): Convert 1st param from location_t to
14779         dump_user_location_t.
14780         * value-prof.c (check_counter): Update call to dump_printf_loc to
14781         pass in a dump_user_location_t rather than a location_t; update
14782         call to error_at for change in type of "locus".
14783         (check_ic_target): Update call to dump_printf_loc to
14784         pass in a dump_user_location_t rather than a location_t, via the
14785         call_stmt.
14787 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
14789         * config/s390/s390.h (enum processor_flags): Do not use
14790         default tune parameter when -march was specified.
14792 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
14794         PR target/86314
14795         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
14796         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
14797         operands.
14799 2018-06-26  Richard Biener  <rguenther@suse.de>
14801         PR tree-optimization/86287
14802         PR bootstrap/86316
14803         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
14804         (vect_analyze_loop): Initialize n_stmts.
14806 2018-06-26  Richard Biener  <rguenther@suse.de>
14808         PR middle-end/86271
14809         * fold-const.c (fold_convertible_p): Pointer extension
14810         isn't valid.
14812 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
14814         PR debug/86064
14815         * dwarf2out.c (loc_list_has_views): Adjust comments.
14816         (dw_loc_list): Split single cross-partition range with
14817         nonzero locview.
14819 2018-06-25  Jeff Law  <law@redhat.com>
14821         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
14822         on -mbig-switch by default.
14824         * config/v850/predicates.md (const_float_1_operand): Fix match_code
14825         test.
14826         (const_float_0_operand): Remove unused predicate.
14827         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
14828         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
14829         (recipsf2): New expander.  Original pattern now called
14830         (recipsf2_insn).
14831         (recipdf2, recipdf2_insn): Similarly.
14832         (rsqrtsf2, rsqrtsf2_insn): Similarly
14833         (rsqrtdf2, rsqrtdf2_insn): Similarly
14835 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
14837         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
14838         Simplify logic for FreeBSD (twice).
14840 2018-06-25  Martin Sebor  <msebor@redhat.com>
14842         PR tree-optimization/86204
14843         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
14844         a strnlen result if it's less than the length of the string.
14846 2018-06-25  Martin Sebor  <msebor@redhat.com>
14848         PR tree-optimization/85700
14849         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
14850         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
14851         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
14853 2018-06-25  Martin Sebor  <msebor@redhat.com>
14855         * doc/extend.texi (Zero-length arrays): Update and clarify.
14857 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
14859         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
14860         added IEEE/IBM long double multilib support on PowerPC little
14861         endian Linux systems.
14862         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
14863         (MULTILIB_DEFAULTS): Likewise.
14864         * config/rs6000/rs6000.c (rs6000_option_override_internal):
14865         Likewise.
14866         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
14867         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
14868         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
14870 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
14872         PR middle-end/86311
14873         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
14874         (REORDER_45): Likewise.
14876 2018-06-25  Jeff Law  <law@redhat.com>
14878         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
14879         dividend to 32 bits.  Adjust length.
14880         (udivmodhi4): Cleanup output template.  Fix length.
14882 2018-06-25  Carl Love  <cel@us.ibm.com>
14884         * config/rs6000/vsx.md: Change word selector to prefered location.
14886 2018-06-25  Richard Biener  <rguenther@suse.de>
14888         PR tree-optimization/86304
14889         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
14890         epilogue-if-converted loops as well.
14892 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
14894         * lto-section-out.c (lto_begin_section): Do not print section
14895         name for noaddr and unnumbered dumps.
14897 2018-06-25  Richard Biener  <rguenther@suse.de>
14899         * tree-vectorizer.h (struct vec_info_shared): New structure
14900         with parts split out from struct vec_info and loop_nest from
14901         struct _loop_vec_info.
14902         (struct vec_info): Adjust accordingly.
14903         (struct _loop_vec_info): Likewise.
14904         (LOOP_VINFO_LOOP_NEST): Adjust.
14905         (LOOP_VINFO_DATAREFS): Likewise.
14906         (LOOP_VINFO_DDRS): Likewise.
14907         (struct _bb_vec_info): Likewise.
14908         (BB_VINFO_DATAREFS): Likewise.
14909         (BB_VINFO_DDRS): Likewise.
14910         (struct _stmt_vec_info): Add dr_aux member.
14911         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
14912         (DR_MISALIGNMENT_UNINITIALIZED): New.
14913         (set_dr_misalignment): Adjust.
14914         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
14915         (vect_analyze_loop): Adjust prototype.
14916         (vect_analyze_loop_form): Likewise.
14917         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
14918         Compute dependences lazily.
14919         (vect_record_base_alignments): Use shared datarefs/ddrs.
14920         (vect_verify_datarefs_alignment): Likewise.
14921         (vect_analyze_data_refs_alignment): Likewise.
14922         (vect_analyze_data_ref_accesses): Likewise.
14923         (vect_analyze_data_refs): Likewise.
14924         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
14925         constructor parameter for shared part.
14926         (vect_analyze_loop_form): Pass in shared part and adjust.
14927         (vect_analyze_loop_2): Pass in storage for the number of
14928         stmts.  Move loop nest finding to the caller.  Compute
14929         datarefs lazily.
14930         (vect_analyze_loop): Pass in shared part.
14931         (vect_transform_loop): Verify shared datarefs are unchanged.
14932         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
14933         constructor parameter for shared part.
14934         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
14935         (vect_slp_bb): Verify shared datarefs are unchanged before
14936         transform.
14937         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
14938         change.
14939         (new_stmt_vec_info): Initialize DR_AUX misalignment to
14940         DR_MISALIGNMENT_UNINITIALIZED.
14941         * tree-vectorizer.c (vec_info::vec_info): Add constructor
14942         parameter for shared part.
14943         (vec_info::~vec_info): Adjust.
14944         (vec_info_shared::vec_info_shared): New.
14945         (vec_info_shared::~vec_info_shared): Likewise.
14946         (vec_info_shared::save_datarefs): Likewise.
14947         (vec_info_shared::check_datarefs): Likewise.
14948         (try_vectorize_loop_1): Construct shared part live for analyses
14949         of a single loop for multiple vector sizes.
14950         * tree-parloops.c (gather_scalar_reductions): Adjust.
14952 2018-06-25  Richard Biener  <rguenther@suse.de>
14954         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
14955         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
14956         (vect_analyze_data_refs): Remove similar code from here and
14957         simplify accordingly.
14959 2018-06-25  Richard Biener  <rguenther@suse.de>
14961         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
14962         for reverse storage order accesses rather than asserting
14963         they cannot happen here.
14965 2018-06-25  Tom de Vries  <tdevries@suse.de>
14967         PR debug/86257
14968         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
14969         Use data16 instead of .byte for insn prefix.
14971 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
14973         PR C++/86082
14974         * parser.c (make_char_string_pack): Pass this literal chars
14975         through cpp_interpret_string.
14976         (cp_parser_userdef_numeric_literal): Check the result of
14977         make_char_string_pack.
14979 2018-06-24  Maya Rashish  <coypu@sdf.org>
14981         * ginclude/stddef.h: Simplify conditions around avoiding
14982         re-definition of __size_t.
14984 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
14986         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
14987         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
14989 2018-06-22  Maya Rashish  <coypu@sdf.org>
14991         * doc/invoke.texi (mno-fancy-math-387): Update for changes
14992         made to OpenBSD and NetBSD through the years.
14994 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14996         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
14997         behavior of vec_pack (vector double, vector double) to match
14998         behavior of vec_float2 (vector double, vector double).
15000 2018-06-22  Olivier Hainque  <hainque@adacore.com>
15002         * gimplify.c (gimplify_function_tree): Prevent creation
15003         of a trampoline for the address of the current function
15004         passed to entry/exit instrumentation hooks.
15006 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
15008         PR target/86222
15009         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
15010         correctly.
15012 2018-06-22  Martin Liska  <mliska@suse.cz>
15014         PR tree-optimization/86263
15015         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15016         Bail out if is_enabled is false.
15017         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
15018         New declaration.
15019         (jump_table_cluster::is_enabled): New function.
15021 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
15023         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
15024         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
15025         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
15026         (lto_input_ts_binfo_tree_pointers): Likewise.
15027         * tree-streamer-out.c (streamer_write_tree_bitfields,
15028         write_ts_binfo_tree_pointers): Likewise.
15029         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
15031 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
15033         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
15035 2018-06-22  Martin Liska  <mliska@suse.cz>
15037         * symbol-summary.h (get): Make it pure and inline move
15038         functionality from ::get function.
15039         (get): Remove and inline into ::get and ::get_create.
15040         (get_create): Move code from ::get function.
15042 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15044         PR target/85994
15045         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
15046         -x assembler-with-cpp.
15048 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
15050         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
15051         _FILE_OFFSET_BITS=64 for C++.
15053 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
15055         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
15056         conversion insn that shows up when pr85657-3.c is compiled using
15057         IEEE 128-bit long double.
15058         (neg<mode>2_internal): Use the correct mode to check whether the
15059         mode is IBM extended.
15060         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
15061         multiply and divide external functions from being created more
15062         than once.
15064 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15066         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
15067         functions.
15068         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
15069         the edge can be forwarded.
15070         (cfg_layout_merge_blocks): Likewise.
15072 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15074         * except.c (finish_eh_generation): Commit edge insertions only after
15075         the EH edges have been redirected from post-landing to landing pads.
15077 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15079         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
15080         create_tmp_var_for to create the FRAME decl.
15081         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
15083 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
15085         * tree-inline.c (copy_edges_for_bb): Minor tweak.
15086         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
15087         debug statement when resetting its value.
15088         (expand_call_inline): Copy the locus of the call onto the assignment
15089         of the return value, if any.  Use local variable in more cases.
15091 2018-06-21  Martin Liska  <mliska@suse.cz>
15093         * ipa-pure-const.c (propagate_nothrow): Use
15094         funct_state_summaries->get.
15095         (dump_malloc_lattice): Likewise.
15096         (propagate_malloc): Likewise.
15098 2018-06-21  Richard Biener  <rguenther@suse.de>
15100         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
15101         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
15102         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
15103         comment.
15104         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
15105         BLOCK_ABSTRACT_ORIGIN unconditionally.
15107 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
15109         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
15110         deleting it.
15111         * ipa-reference.c (ipa_reference_c_finalize): Delete
15112         ipa_ref_opt_sum_summaries and set it to NULL.
15114 2018-06-21  Tom de Vries  <tdevries@suse.de>
15116         PR tree-optimization/85859
15117         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
15118         test with comment from bb_no_side_effects_p.
15120 2018-06-21  Richard Biener  <rguenther@suse.de>
15122         PR tree-optimization/86232
15123         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
15124         max for constant niter.
15126 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15128         * config/aarch64/aarch64-simd.md
15129         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
15131 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
15133         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
15134         Make opernads of the unspec commutative.
15136 2018-06-21  Richard Biener  <rguenther@suse.de>
15138         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
15139         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
15140         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
15141         (vect_analyze_data_ref_dependence): Re-order checks to deal with
15142         NULL DR_STEP.
15143         (vect_record_base_alignments): Do not record base alignment
15144         for gathers or scatters.
15145         (vect_compute_data_ref_alignment): Drop return value that is always
15146         true.  Bail out early for gathers or scatters.
15147         (vect_enhance_data_refs_alignment): Bail out early for gathers
15148         or scatters.
15149         (vect_find_same_alignment_drs): Likewise.
15150         (vect_analyze_data_refs_alignment): Remove dead code.
15151         (vect_slp_analyze_and_verify_node_alignment): Likewise.
15152         (vect_analyze_data_refs): For possible gathers or scatters do
15153         not create an alternate DR, just check their possible validity
15154         and mark them.  Adjust DECL_NONALIASED handling to not rely
15155         on DR_BASE_ADDRESS.
15156         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
15157         update inits of gathers or scatters.
15158         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
15159         Also copy gather/scatter flag to pattern vinfo.
15161 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
15163         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
15164         behavior of vec_packsu (vector unsigned long long, vector unsigned
15165         long long) to match behavior of vec_packs with same signature.
15167 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
15168             Thomas Schwinge <thomas@codesourcery.com>
15169             Cesar Philippidis  <cesar@codesourcery.com>
15171         * gimplify.c (gimplify_scan_omp_clauses): Add support for
15172         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15173         (gimplify_adjust_omp_clauses): Likewise.
15174         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
15175         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
15176         (gimplify_omp_target_update): Update handling of acc update and
15177         enter/exit data.
15178         * omp-low.c (install_var_field): Remove unused parameter
15179         base_pointers_restrict.
15180         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
15181         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
15182         FINALIZE}
15183         (omp_target_base_pointers_restrict_p): Delete.
15184         (scan_omp_target): Update call to scan_sharing_clauses.
15185         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
15186         FINALIZE}.
15187         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
15188         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
15189         (convert_local_omp_clauses): Likewise.
15190         * tree-pretty-print.c (dump_omp_clause): Likewise.
15191         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
15192         FINALIZE}.
15193         (omp_clause_code_name): Likewise.
15195 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
15197         PR debug/86194
15198         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
15199         be narrowed.
15201         PR tree-optimization/86231
15202         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
15203         anti-range don't overwrite *vr0min before using it to compute *vr0max.
15205 2018-06-20  Tom de Vries  <tdevries@suse.de>
15207         PR tree-optimization/86097
15208         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
15209         iv type if signedness of iv type is not the same as that of *nit.
15211 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
15213         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
15214         EDGE_EH edges, verify they are all EDGE_EH.
15216 2018-06-20  Maya Rashish  <coypu@sdf.org>
15218         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
15220 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15222         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
15223         * config/aarch64/aarch64.c (xgene1_tunings): Add
15224         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
15225         (aarch64_mode_valid_for_sched_fusion_p):
15226         Allow 16-byte modes.
15227         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
15228         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
15229         128-bit modes.
15230         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
15231         New pattern.
15232         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
15233         * config/aarch64/iterators.md (VQ2): New mode iterator.
15235 2018-06-20  Martin Liska  <mliska@suse.cz>
15237         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
15238         Change default ratio from 10 to 8.
15240 2018-06-20  Martin Liska  <mliska@suse.cz>
15242         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
15243         New.
15244         (bit_test_cluster::find_bit_tests): Likewise.
15245         (switch_decision_tree::analyze_switch_statement): Find clusters.
15246         * tree-switch-conversion.h (struct jump_table_cluster): Document
15247         hierarchy.
15249 2018-06-20  Martin Liska  <mliska@suse.cz>
15251         * tree-switch-conversion.c (switch_conversion::collect):
15252         Record m_uniq property.
15253         (switch_conversion::expand): Bail out for special conditions.
15254         (group_cluster::~group_cluster): New.
15255         (group_cluster::group_cluster): Likewise.
15256         (group_cluster::dump): Likewise.
15257         (jump_table_cluster::emit): New.
15258         (switch_decision_tree::fix_phi_operands_for_edges): New.
15259         (struct case_node): Remove struct.
15260         (jump_table_cluster::can_be_handled): New.
15261         (case_values_threshold): Moved to header.
15262         (reset_out_edges_aux): Likewise.
15263         (jump_table_cluster::is_beneficial): New.
15264         (bit_test_cluster::can_be_handled): Likewise.
15265         (add_case_node): Remove.
15266         (bit_test_cluster::is_beneficial): New.
15267         (case_bit_test::cmp): New.
15268         (bit_test_cluster::emit): New.
15269         (expand_switch_as_decision_tree_p): Remove.
15270         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
15271         (fix_phi_operands_for_edge): Likewise.
15272         (switch_decision_tree::analyze_switch_statement): New.
15273         (compute_cases_per_edge): Move ...
15274         (switch_decision_tree::compute_cases_per_edge): ... here.
15275         (try_switch_expansion): Likewise.
15276         (switch_decision_tree::try_switch_expansion): Likewise.
15277         (record_phi_operand_mapping): Likewise.
15278         (switch_decision_tree::record_phi_operand_mapping): Likewise.
15279         (emit_case_decision_tree): Likewise.
15280         (switch_decision_tree::emit): Likewise.
15281         (balance_case_nodes): Likewise.
15282         (switch_decision_tree::balance_case_nodes): Likewise.
15283         (dump_case_nodes): Likewise.
15284         (switch_decision_tree::dump_case_nodes): Likewise.
15285         (emit_jump): Likewise.
15286         (switch_decision_tree::emit_jump): Likewise.
15287         (emit_cmp_and_jump_insns): Likewise.
15288         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
15289         (emit_case_nodes): Likewise.
15290         (switch_decision_tree::emit_case_nodes): Likewise.
15291         (conditional_probability): Remove.
15292         * tree-switch-conversion.h (enum cluster_type): New.
15293         (PRINT_CASE): New.
15294         (struct cluster): Likewise.
15295         (cluster::cluster): Likewise.
15296         (struct simple_cluster): Likewise.
15297         (simple_cluster::simple_cluster): Likewise.
15298         (struct group_cluster): Likewise.
15299         (struct jump_table_cluster): Likewise.
15300         (struct bit_test_cluster): Likewise.
15301         (struct min_cluster_item): Likewise.
15302         (struct case_tree_node): Likewise.
15303         (case_tree_node::case_tree_node): Likewise.
15304         (jump_table_cluster::case_values_threshold): Likewise.
15305         (struct case_bit_test): Likewise.
15306         (struct switch_decision_tree): Likewise.
15307         (struct switch_conversion): Likewise.
15308         (switch_decision_tree::reset_out_edges_aux): Likewise.
15310 2018-06-20  Martin Liska  <mliska@suse.cz>
15312         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
15313         (hoist_edge_and_branch_if_true): Likewise.
15314         (expand_switch_using_bit_tests_p): Likewise.
15315         (struct case_bit_test): Likewise.
15316         (case_bit_test_cmp): Likewise.
15317         (emit_case_bit_tests): Likewise.
15318         (switch_conversion::switch_conversion): New class.
15319         (struct switch_conv_info): Remove old struct.
15320         (collect_switch_conv_info): More to ...
15321         (switch_conversion::collect): ... this.
15322         (check_range): Likewise.
15323         (switch_conversion::check_range): Likewise.
15324         (check_all_empty_except_final): Likewise.
15325         (switch_conversion::check_all_empty_except_final): Likewise.
15326         (check_final_bb): Likewise.
15327         (switch_conversion::check_final_bb): Likewise.
15328         (create_temp_arrays): Likewise.
15329         (switch_conversion::create_temp_arrays): Likewise.
15330         (free_temp_arrays): Likewise.
15331         (gather_default_values): Likewise.
15332         (switch_conversion::gather_default_values): Likewise.
15333         (build_constructors): Likewise.
15334         (switch_conversion::build_constructors): Likewise.
15335         (constructor_contains_same_values_p): Likewise.
15336         (switch_conversion::contains_same_values_p): Likewise.
15337         (array_value_type): Likewise.
15338         (switch_conversion::array_value_type): Likewise.
15339         (build_one_array): Likewise.
15340         (switch_conversion::build_one_array): Likewise.
15341         (build_arrays): Likewise.
15342         (switch_conversion::build_arrays): Likewise.
15343         (gen_def_assigns): Likewise.
15344         (switch_conversion::gen_def_assigns): Likewise.
15345         (prune_bbs): Likewise.
15346         (switch_conversion::prune_bbs): Likewise.
15347         (fix_phi_nodes): Likewise.
15348         (switch_conversion::fix_phi_nodes): Likewise.
15349         (gen_inbound_check): Likewise.
15350         (switch_conversion::gen_inbound_check): Likewise.
15351         (process_switch): Use the newly created class.
15352         (switch_conversion::expand): New.
15353         (switch_conversion::~switch_conversion): New.
15354         * tree-switch-conversion.h: New file.
15356 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15358         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
15359         tree-vect-patterns.c.
15360         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
15361         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
15362         (vect_recog_sad_pattern): Likewise.
15363         (vect_recog_widen_sum_pattern): Likewise.
15364         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
15365         (vect_recog_widen_shift_pattern): Remove the type_in argument.
15366         (vect_recog_rotate_pattern): Likewise.
15367         (vect_recog_mult_pattern): Likewise.
15368         (vect_recog_vector_vector_shift_pattern): Likewise.
15369         (vect_recog_divmod_pattern): Likewise.
15370         (vect_recog_mixed_size_cond_pattern): Likewise.
15371         (vect_recog_bool_pattern): Likewise.
15372         (vect_recog_mask_conversion_pattern): Likewise.
15373         (vect_try_gather_scatter_pattern): Likewise.
15374         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
15375         (vect_recog_over_widening_pattern): Likewise.
15376         (vect_recog_gather_scatter_pattern): Likewise.
15377         (vect_recog_func_ptr): Move from tree-vectorizer.h
15378         (vect_vect_recog_func_ptrs): Move further down the file.
15379         (vect_recog_func): Likewise.  Remove the third argument.
15380         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
15381         (vect_pattern_recog_1): Expect the pattern function to do any
15382         necessary target tests.  Also expect it to provide a vector type.
15383         Remove the type_in handling.
15385 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15387         * tree-vect-patterns.c (vect_pattern_detected): New function.
15388         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
15389         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
15390         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
15391         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
15392         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
15393         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
15394         (vect_recog_mask_conversion_pattern)
15395         (vect_try_gather_scatter_pattern): Likewise.
15397 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15399         * tree-vect-patterns.c (vect_get_internal_def): New function.
15400         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
15401         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
15402         (search_type_for_mask_1): Use it.
15404 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15406         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15407         redundant WIDEN_SUM_EXPR handling.
15408         (vect_recog_sad_pattern): Likewise.
15410 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15412         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
15413         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
15414         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
15415         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
15416         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
15418 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15420         * tree-vect-stmts.c (vectorizable_call): Make sure that we
15421         use the stmt_vec_info of the original bb statement for the
15422         new zero assignment, even if the call is part of a pattern.
15424 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15426         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
15427         that the sequence is attached to the original statement rather
15428         than the pattern statement.
15429         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
15430         PATTERN_DEF_SEQ from the original statement rather than
15431         the main pattern statement.
15432         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
15433         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
15434         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
15436 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
15438         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
15439         definition statements before the early exit for statements that aren't
15440         live or relevant.
15441         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
15442         split out from...
15443         (vect_transform_loop): ...here.  Process pattern definition
15444         statements without first checking whether the main pattern
15445         statement is live or relevant.
15447 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15449         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
15450         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
15452 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
15454         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
15455         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
15456         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
15457         (expand_block_compare): Change select_block_compare_mode call.
15458         (expand_strncmp_align_check): Use new functions, fix comment.
15459         (emit_final_str_compare_gpr): New function.
15460         (expand_strn_compare): Refactor and clean up code.
15461         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
15463 2018-06-19  Tony Reix  <tony.reix@atos.com>
15464             Damien Bergamini  <damien.bergamini@atos.com>
15465             David Edelsohn  <dje.gcc@gmail.com>
15467         * collect2.c (static_obj): New variable.
15468         (static_libs): New variable.
15469         (is_in_list): Uncomment declaration.
15470         (main): Track AIX libraries linked statically.
15471         (is_in_list): Uncomment definition.
15472         (scan_prog_file): Don't add AIX shared libraries initializer
15473         to constructor list if linking statically.
15475 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
15477         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
15478         constant.
15479         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
15481 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
15483         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
15484         blocks.
15486 2018-06-19  Martin Liska  <mliska@suse.cz>
15488         * config/i386/i386.c (ix86_can_inline_p): Do not use
15489         ipa_fn_summaries::get_create.
15490         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
15491         get.
15492         (devirtualization_time_bonus): Likewise.
15493         (ipcp_propagate_stage): Likewise.
15494         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
15495         (edge_set_predicate): Likewise.
15496         (evaluate_conditions_for_known_args): Likewise.
15497         (evaluate_properties_for_edge): Likewise.
15498         (ipa_call_summary::reset): Tranform to ...
15499         (ipa_call_summary::~ipa_call_summary): ... this.
15500         (ipa_fn_summary::reset): Transform to ...
15501         (ipa_fn_summary::~ipa_fn_summary): ... this.
15502         (ipa_fn_summary_t::remove): Rename to ...
15503         (ipa_fn_summary_t::remove_callees): ... this.
15504         (ipa_fn_summary_t::duplicate): Use placement new
15505         instead of memory copy.
15506         (ipa_call_summary_t::duplicate): Likewise.
15507         (ipa_call_summary_t::remove): Remove.
15508         (dump_ipa_call_summary): Change get_create to get.
15509         (ipa_dump_fn_summary): Dump only when summary exists.
15510         (analyze_function_body): Use symbol_summary::get instead
15511         of get_create.
15512         (compute_fn_summary): Likewise.
15513         (estimate_edge_devirt_benefit): Likewise.
15514         (estimate_edge_size_and_time): Likewise.
15515         (inline_update_callee_summaries): Likewise.
15516         (remap_edge_change_prob): Likewise.
15517         (remap_edge_summaries): Likewise.
15518         (ipa_merge_fn_summary_after_inlining): Likewise.
15519         (write_ipa_call_summary): Likewise.
15520         (ipa_fn_summary_write): Likewise.
15521         (ipa_free_fn_summary): Likewise.
15522         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
15523         (struct ipa_call_summary): Likewise.
15524         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
15525         of get_create.
15526         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
15527         (estimate_size_after_inlining): Likewise.
15528         (estimate_growth): Likewise.
15529         (growth_likely_positive): Likewise.
15530         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
15531         (inline_call): Likewise.
15532         * ipa-inline.c (caller_growth_limits): Likewise.
15533         (can_inline_edge_p): Likewise.
15534         (can_inline_edge_by_limits_p): Likewise.
15535         (compute_uninlined_call_time): Likewise.
15536         (compute_inlined_call_time): Likewise.
15537         (want_inline_small_function_p): Likewise.
15538         (edge_badness): Likewise.
15539         (update_caller_keys): Likewise.
15540         (update_callee_keys): Likewise.
15541         (inline_small_functions): Likewise.
15542         (inline_to_all_callers_1): Likewise.
15543         (dump_overall_stats): Likewise.
15544         (early_inline_small_functions): Likewise.
15545         (early_inliner): Likewise.
15546         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
15547         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
15548         * ipa-pure-const.c (malloc_candidate_p): Likewise.
15549         * ipa-split.c (execute_split_functions): Likewise.
15550         * symbol-summary.h: Likewise.
15551         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
15553 2018-06-19  Richard Biener  <rguenther@suse.de>
15555         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
15556         (vectorize_loops): ... here.  Fix dbgcnt handling.
15557         (try_vectorize_loop): Wrap try_vectorize_loop_1.
15559 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
15561         PR target/86197
15562         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
15563         ieee128 argument takes up only one (vector) register, not two (floating
15564         point) registers.
15566 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
15568         * gimplify.c (gimplify_init_constructor): Really never clear for an
15569         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
15571 2018-06-19  Richard Biener  <rguenther@suse.de>
15573         PR tree-optimization/86179
15574         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
15575         after failed recognition.
15577 2018-06-18  Martin Sebor  <msebor@redhat.com>
15579         PR middle-end/85602
15580         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
15581         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
15582         Handle integer subtraction.
15583         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
15584         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
15586 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15588         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
15589         param from rtx to rtx_insn *.
15590         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
15591         param.
15592         (frv_ifcvt_modify_insn): Likwise.
15593         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
15594         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
15595         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
15596         as_a <rtx_insn *> cast to local "unprotected_region" once
15597         it's been established that it's not NULL or pc_rtx.
15598         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
15599         param "sethi" from rtx to rtx_insn *.
15600         (nds32_group_float_insns): Likewise for param "insn".
15601         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
15602         param.
15603         (vax_output_int_subtract): Likewise.
15604         * config/vax/vax.c (vax_output_int_add): Likewise for param
15605         "insn".
15606         (vax_output_int_subtract): Likewise.
15607         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
15608         (emit_pattern_after): Likewise for param "after".
15609         (emit_insn_after): Likewise.
15610         (emit_jump_insn_after): Likewise.
15611         (emit_call_insn_after): Likewise.
15612         (emit_debug_insn_after): Likewise.
15613         (emit_pattern_before): Likewise for param "before".
15614         (emit_insn_before): Likewise.
15615         (emit_jump_insn_before): Likewise.
15616         * final.c (get_insn_template): Likewise for param "insn", removing
15617         a cast.
15618         * output.h (get_insn_template): Likewise for 2nd param.
15619         * rtl.h (emit_insn_before): Likewise.
15620         (emit_jump_insn_before): Likewise.
15621         (emit_debug_insn_before_noloc): Likewise.
15622         (emit_insn_after): Likewise.
15623         (emit_jump_insn_after): Likewise.
15624         (emit_call_insn_after): Likewise.
15625         (emit_debug_insn_after): Likewise.
15626         (set_insn_deleted): Likewise for param.
15628 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
15630         PR target/85358
15631         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
15632         floating point modes, so that IFmode is numerically greater than
15633         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
15634         to declare the ordering.  This prevents IFmode from being
15635         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
15636         machine.  Include rs6000-modes.h to share the fractional values
15637         between genmodes* and the rest of the compiler.
15638         (IFmode): Likewise.
15639         (KFmode): Likewise.
15640         (TFmode): Likewise.
15641         * config/rs6000/rs6000-modes.h: New file.
15642         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
15643         meaning of rs6000_long_double_size so that 126..128 selects an
15644         appropriate 128-bit floating point type.
15645         (rs6000_option_override_internal): Likewise.
15646         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
15647         (TARGET_LONG_DOUBLE_128): Change the meaning of
15648         rs6000_long_double_size so that 126..128 selects an appropriate
15649         128-bit floating point type.
15650         (LONG_DOUBLE_TYPE_SIZE): Update comment.
15651         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
15652         source and destination to match the standard usage.
15653         (truncifkf2): Likewise.
15654         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
15655         ISA 2.07 to use an explicit clobber, instead of passing in a
15656         temporary.
15657         (copysign<mode>3_soft): Likewise.
15659 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
15661         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
15662         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
15663         (vect_slp_analyze_instance_dependence): Likewise.
15664         (vect_enhance_data_refs_alignment): Likewise.
15665         (vect_analyze_data_refs_alignment): Likewise.
15666         (vect_slp_analyze_and_verify_instance_alignment
15667         (vect_analyze_data_ref_accesses): Likewise.
15668         (vect_prune_runtime_alias_test_list): Likewise.
15669         (vect_analyze_data_refs): Likewise.
15670         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
15671         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15672         (vect_analyze_scalar_cycles_1): Likewise.
15673         (vect_get_loop_niters): Likewise.
15674         (vect_analyze_loop_form_1): Likewise.
15675         (vect_update_vf_for_slp): Likewise.
15676         (vect_analyze_loop_operations): Likewise.
15677         (vect_analyze_loop): Likewise.
15678         (vectorizable_induction): Likewise.
15679         (vect_transform_loop): Likewise.
15680         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
15681         * tree-vect-slp.c (vect_analyze_slp): Likewise.
15682         (vect_make_slp_decision): Likewise.
15683         (vect_detect_hybrid_slp): Likewise.
15684         (vect_slp_analyze_operations): Likewise.
15685         (vect_slp_bb): Likewise.
15686         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
15687         (vectorizable_bswap): Likewise.
15688         (vectorizable_call): Likewise.
15689         (vectorizable_simd_clone_call): Likewise.
15690         (vectorizable_conversion): Likewise.
15691         (vectorizable_assignment): Likewise.
15692         (vectorizable_shift): Likewise.
15693         (vectorizable_operation): Likewise.
15694         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
15696 2018-06-18  Martin Sebor  <msebor@redhat.com>
15698         PR tree-optimization/81384
15699         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
15700         * builtins.c (expand_builtin_strnlen): New function.
15701         (expand_builtin): Call it.
15702         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
15703         * builtins.def (BUILT_IN_STRNLEN): New.
15704         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
15705         Warn for bounds in excess of maximum object size.
15706         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
15707         single-value ranges.  Handle strnlen.
15708         (handle_builtin_strlen): Handle strnlen.
15709         (strlen_check_and_optimize_stmt): Same.
15710         * doc/extend.texi (Other Builtins): Document strnlen.
15712 2018-06-18  Maya Rashish  <coypu@sdf.org>
15714         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
15715         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
15716         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
15718         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
15719         here to ...
15720         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
15722 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15724         * tree.c (escaped_string::escape): Replace cast to char * by
15725         const_cast<char *> (unescaped).
15727 2018-06-18  Nick Clifton  <nickc@redhat.com>
15729         PR 84195
15730         * tree.c (escaped_string): New class.  Converts an unescaped
15731         string into its escaped equivalent.
15732         (warn_deprecated_use): Use the new class to convert the
15733         deprecation message, if present.
15734         (test_escaped_strings): New self test.
15735         (test_c_tests): Add test_escaped_strings.
15736         * doc/extend.texi (deprecated): Add a note that the
15737         deprecation message is affected by the -fmessage-length
15738         option, and that control characters will be escaped.
15739         (#pragma GCC error): Document this pragma.
15740         (#pragma GCC warning): Likewise.
15741         * doc/invoke.texi (-fmessage-length): Document this option's
15742         effect on the #warning and #error preprocessor directives and
15743         the deprecated attribute.
15745 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
15747         * tree.c (decl_value_expr_lookup): Revert latest change.
15748         (decl_value_expr_insert): Likewise.
15750 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
15752         * gimplify.c (nonlocal_vlas): Delete.
15753         (nonlocal_vla_vars): Likewise.
15754         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
15755         referenced VLAs.
15756         (gimplify_body): Do not create and destroy nonlocal_vlas.
15757         * tree-nested.c: Include diagnostic.h.
15758         (use_pointer_in_frame): Tweak.
15759         (lookup_field_for_decl): Add assertion and declare the transformation.
15760         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
15761         internal error when the reference is in a wrong context.  Do not
15762         create a debug decl by default.
15763         (note_nonlocal_block_vlas): Delete.
15764         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
15765         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
15766         create a debug decl by default.
15767         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
15768         call is in a wrong context.
15769         (fixup_vla_decls): New function.
15770         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
15771         debug variables were created.
15772         * tree.c (decl_value_expr_lookup): Add checking assertion.
15773         (decl_value_expr_insert): Likewise.
15775 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15777         PR middle-end/82479
15778         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
15779         * tree-scalar-evolution.c (interpret_expr): Likewise.
15780         (expression_expensive_p): Likewise.
15781         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
15782         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
15783         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
15784         (ssa_defined_by_minus_one_stmt_p): New.
15786 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
15788         PR middle-end/64946
15789         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
15790         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
15791         * dojump.c (do_jump): Likewise.
15792         * expr.c (expand_expr_real_2): Check operand type's sign.
15793         * fold-const.c (const_unop): Handle ABSU_EXPR.
15794         (fold_abs_const): Likewise.
15795         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
15796         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
15797         (strip_sign_op_1): Likesise.
15798         * match.pd: Add new pattern to generate ABSU_EXPR.
15799         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
15800         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
15801         * tree-eh.c (operation_could_trap_helper_p): Likewise.
15802         * tree-inline.c (estimate_operator_cost): Likewise.
15803         * tree-pretty-print.c (dump_generic_node): Likewise.
15804         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
15805         * tree.def (ABSU_EXPR): New.
15807 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
15809         PR middle-end/86095
15810         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
15811         documented as preserved for backward compatibility only.
15812         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
15814         PR rtl-optimization/86108
15815         * bb-reorder.c (create_forwarder_block): Renamed to ...
15816         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
15817         jump from new landing pad to the second part.
15818         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
15819         Adjust callers.
15821 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
15823         PR middle-end/85878
15824         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
15825         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
15826         Only call store_expr for halves if the mode is the same.
15828         PR middle-end/86123
15829         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
15830         Fix up comment formatting.
15832 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15834         * typed-splay-tree.h (typed_splay_tree::remove): New function.
15835         (typed_splay_tree::closure,
15836         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
15837         (typed_splay_tree::typed_splay_tree,
15838         typed_splay_tree::operator =): Declared private.
15839         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
15840         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
15841         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
15842         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
15843         typed_splay_tree::splay_tree_splay,
15844         typed_splay_tree::splay_tree_foreach_helper,
15845         typed_splay_tree::splay_tree_insert,
15846         typed_splay_tree::splay_tree_remove,
15847         typed_splay_tree::splay_tree_lookup,
15848         typed_splay_tree::splay_tree_predecessor,
15849         typed_splay_tree::splay_tree_successor,
15850         typed_splay_tree::splay_tree_min,
15851         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
15852         (typed_splay_tree::root, typed_splay_tree::comp,
15853         typed_splay_tree::delete_key,
15854         typed_splay_tree::delete_value): New data members.
15855         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
15856         typed_splay_tree::remove.
15858 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
15860         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
15861         -mginv and -mno-ginv to the assembler.
15862         * config/mips/mips.opt (-mcrc): New option.
15863         (-mginv): Likewise.
15864         * doc/invoke.text (-mcrc): Document.
15865         (-mginv): Likewise.
15867 2018-06-15  Nick Clifton  <nickc@redhat.com>
15869         PR 84195
15870         * tree.c (escaped_string): New class.  Converts an unescaped
15871         string into its escaped equivalent.
15872         (warn_deprecated_use): Use the new class to convert the
15873         deprecation message, if present.
15874         (test_escaped_strings): New self test.
15875         (test_c_tests): Add test_escaped_strings.
15876         * doc/extend.texi (deprecated): Add a note that the
15877         deprecation message is affected by the -fmessage-length
15878         option, and that control characters will be escaped.
15879         (#pragma GCC error): Document this pragma.
15880         (#pragma GCC warning): Likewise.
15881         * doc/invoke.texi (-fmessage-length): Document this option's
15882         effect on the #warning and #error preprocessor directives and
15883         the deprecated attribute.
15885 2018-06-15  Richard Biener  <rguenther@suse.de>
15887         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
15888         here, also noting vector size used.
15889         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
15890         size used in MSG_OPTIMIZED_LOCATIONS dump.
15891         (pass_slp_vectorize::execute): Adjust.
15893 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
15895         PR target/85968
15896         * config/arc/arc.c (arc_return_address_register): Fix
15897         if-condition.
15899 2018-06-15  Richard Biener  <rguenther@suse.de>
15901         PR middle-end/86159
15902         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
15903         leave useless conversion stripping to force_gimple_operand_gsi.
15904         (gimplify_build2): Likewise.
15905         (gimplify_build1): Likewise.
15907 2018-06-15  Richard Biener  <rguenther@suse.de>
15909         PR middle-end/86076
15910         * tree-cfg.c (move_stmt_op): unshare invariant addresses
15911         before adjusting their block.
15913 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15915         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
15916         multilibs for *-*-rtems*.
15917         * config/riscv/t-rtems: New file.
15919 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
15921         PR middle-end/86122
15922         * match.pd ((A +- CST1) +- CST2): Punt if last resort
15923         unsigned_type_for returns NULL.
15925         PR target/85945
15926         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
15927         subregs of multi-word pseudos unless the float mode has word size.
15929 2018-06-14  Richard Biener  <rguenther@suse.de>
15931         PR middle-end/86139
15932         * tree-vect-generic.c (build_word_mode_vector_type): Remove
15933         duplicate and harmful type_hash_canon.
15934         * tree.c (type_hash_canon): Assert we didn't find ourselves.
15936 2018-06-14  Richard Biener  <rguenther@suse.de>
15938         PR ipa/86124
15939         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
15940         NULL cgraph_node.
15942 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
15944         * config/rtems.h (STDINT_LONG32): Define.
15946 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
15947             Prachi Godbole  <prachi.godbole@imgtec.com>
15949         * config/mips/mips-cpus.def: Define P6600.
15950         * config/mips/mips-tables.opt: Regenerate.
15951         * config/mips/mips.c (mips_ucbranch_type): New enum.
15952         (mips_rtx_cost_data): Add support for P6600.
15953         (mips_issue_rate): Likewise.
15954         (mips_multipass_dfa_lookahead): Likewise.
15955         (mips_avoid_hazard): Likewise.
15956         (mips_reorg_process_insns): Likewise.
15957         (mips_classify_branch_p6600): New function.
15958         * config/mips/mips.h (TUNE_P6600): New define.
15959         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
15960         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
15961         * config/mips/mips.md: Include p6600.md.
15962         (processor): Add p6600.
15963         * config/mips/p6600.md: New file.
15964         * doc/invoke.texi: Add p6600 to supported architectures.
15966 2018-06-13  Martin Sebor  <msebor@redhat.com>
15968         PR tree-optimization/86114
15969         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
15970         of integer types.
15971         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
15973 2018-06-13  Richard Biener  <rguenther@suse.de>
15975         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
15976         Properly set vector type of the intermediate stmt.
15977         * tree-vect-stmts.c (vectorizable_operation): The destination
15978         var always has vectype_out type.
15980 2018-06-13  Jeff Law  <law@redhat.com>
15982         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
15983         integer 0 for argument to print_rtl_with_bb.
15984         (rl78_reorg): Likewise.
15986 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
15988         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
15989         from rtx to rtx_insn *.
15990         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
15991         "label".
15992         (add_sched_insns_for_speculation): Likewise for local "target",
15993         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
15994         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
15995         from rtx_insn ** to rtx_code_label **.
15996         (reorg_emit_nops): Likewise.
15997         (c6x_reorg): Likewise for local "call_labels".
15998         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
15999         rtx to rtx_insn *.
16000         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
16001         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
16002         the loops over LABEL_REFS.
16003         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
16004         braf_label.
16005         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
16006         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
16007         (split_branches): Strengthen local "olabel" from rtx to
16008         rtx_insn *, adding a safe_as_a cast.
16009         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
16010         to "rtx_insn *".
16011         (add_insn_after): Likewise for first two params.
16012         (add_insn_before): Likewise.
16013         (remove_insn): Likewise for param.
16014         (emit_pattern_before_noloc): Likewise for second and third params.
16015         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
16016         (emit_call_insn_before_noloc): Likewise.
16017         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
16018         to "rtx_insn *".
16019         (emit_barrier_before): Likewise.
16020         (emit_label_before): Strengthen "label" param from "rtx" to
16021         "rtx_code_label *".  Strengthen "before" param from "rtx" to
16022         "rtx_insn *".
16023         (emit_insn_after_1): Strengthen "after" param from "rtx" to
16024         "rtx_insn *".
16025         (emit_pattern_after_noloc): Likewise.
16026         (emit_insn_after_noloc): Likewise.
16027         (emit_jump_insn_after_noloc): Likewise.
16028         (emit_call_insn_after_noloc): Likewise.
16029         (emit_debug_insn_after_noloc): Likewise.
16030         (emit_barrier_after): Likewise.
16031         (emit_label_after): Likewise for both params.
16032         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
16033         "loc" param from "int" to "location_t".
16034         (emit_insn_after_setloc): Likewise.
16035         (emit_jump_insn_after_setloc): Likewise.
16036         (emit_call_insn_after_setloc): Likewise.
16037         (emit_debug_insn_after_setloc): Likewise.
16038         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
16039         "loc" param from "int" to "location_t".
16040         (emit_pattern_before): Convert NULL_RTX to NULL.
16041         (emit_insn_before_setloc): Convert "loc" param from "int" to
16042         "location_t".
16043         (emit_jump_insn_before_setloc): Likewise.
16044         (emit_call_insn_before_setloc): Likewise.
16045         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
16046         rtx_insn *.  Convert "loc" param from "int" to "location_t".
16047         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
16048         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
16049         Convert 3rd param from "int" to "location_t".
16050         (emit_barrier_before, emit_barrier_after, next_real_insn):
16051         Strengthen param from rtx to rtx_insn *.
16052         (emit_label_before): Strengthen 1st param from "rtx" to
16053         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
16054         "rtx_insn *".
16055         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
16056         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
16057         Strengthen 2nd param from "rtx" to "rtx_insn *".
16058         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
16059         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
16060         Likewise. Convert 3rd param from "int" to "location_t".
16061         (emit_label_after): Strengthen 1st param from "rtx" to
16062         "rtx_code_label *".
16063         (next_real_insn, remove_insn): Strengthen param from "rtx" to
16064         "rtx_insn *".
16065         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
16066         from "rtx" to "rtx_insn *".
16068 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
16070         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
16071         bodies streamed in with -Q.
16072         * dumpfile.c (dump_files): Add lto-stream-out dump file.
16073         * dumpfile.h (tree_dump_index): Add lto_stream_out.
16074         * gimple-streamer-out.c: Include gimple-pretty-print.h
16075         (output_bb): Dump stmts streamed.
16076         * lto-section-out.c: Include print-tree.h
16077         (lto_begin_section): Dump sections created.
16078         (lto_output_decl_index): Dump decl encoded.
16079         * lto-streamer-out.c: Include print-tree.h
16080         (create_output_block): Dump output block created.
16081         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
16082         (output_function): Dump function output.
16083         (output_constructor): Dump constructor streamed.
16084         (write_global_stream): Output indexes encoded.
16085         (produce_asm_for_decls): Dump streams encoded.
16086         * lto-streamer.c (streamer_dump_file): New global var.
16087         * lto-streamer.h (streamer_dump_file): Declare.
16088         * passes.c (ipa_write_summaries): Initialize streamer dump.
16089         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
16090         in.
16092 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
16094         PR target/86048
16095         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
16096         offsets for register save directives.  Emit a second batch of save
16097         directives, if need be, when the function accesses prior frames.
16099 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16101         * config/arc/fpu.md (fmasf4): Force operand to register.
16102         (fnmasf4): Likewise.
16104 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16106         * config/arc/arc-protos.h (arc_pad_return): Remove.
16107         * config/arc/arc.c (machine_function): Remove force_short_suffix
16108         and size_reason.
16109         (arc_print_operand): Adjust printing of '&'.
16110         (arc_verify_short): Remove conditional printing of short suffix.
16111         (arc_final_prescan_insn): Remove reference to size_reason.
16112         (pad_return): New function.
16113         (arc_reorg): Call pad_return.
16114         (arc_pad_return): Remove.
16115         (arc_init_machine_status): Remove reference to force_short_suffix.
16116         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
16117         (attr length): When attribute iscompact is true force to 2
16118         regardless; in the case of maybe check if we want to force the
16119         instruction to have 4 bytes length.
16120         (nopv): Change it to generate 4 byte long nop as well.
16121         (blockage): New pattern.
16122         (simple_return): Remove call to arc_pad_return.
16123         (p_return_i): Likewise.
16125 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16127         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
16129 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16131         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
16132         ARC cores.
16134 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16136         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
16137         for ARC700 and ARCv2.
16139 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
16141         PR target/86076
16142         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
16143         operands[2] instead of operands[1].
16146 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
16148         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
16149         case, check whether the outer register overlaps an unallocatable
16150         register, not just whether it fits the required class.
16152 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
16154         * poly-int.h (can_div_trunc_p): Add new overload in which all values
16155         are poly_ints.
16156         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
16157         (memrefs_conflict_p): Likewise.
16158         (init_alias_analysis): Likewise.
16159         * cfgexpand.c (expand_debug_expr): Likewise.
16160         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
16161         * cse.c (fold_rtx): Likewise.
16162         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
16163         * expr.c (emit_block_move_hints): Likewise.
16164         (clear_storage_hints, push_block, emit_push_insn): Likewise.
16165         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
16166         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
16167         (emit_group_store): Likewise.
16168         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
16169         to read the PRE/POST_MODIFY increment.
16170         * calls.c (store_one_arg): Use strip_offset.
16171         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
16172         poly_int_rtx_p.
16173         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
16174         by a VEC_SELECT.
16175         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
16176         (simplify_binary_operation_1): Extend CONST_INT handling to
16177         poly_int_rtx_p.
16178         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
16179         than a HOST_WIDE_INT.
16180         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
16181         poly_int64.
16182         (adjust_mems, add_stores): Update accodingly.
16183         (vt_canonicalize_addr): Track polynomial offsets.
16184         (emit_note_insn_var_location): Likewise.
16185         (vt_add_function_parameter): Likewise.
16186         (vt_initialize): Likewise.
16188 2018-06-12  Jeff Law  <law@redhat.com>
16190         * config.gcc (alpha*-*-freebsd*): Remove.
16191         * config/alpha/freebsd.h: Remove.
16193 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
16195         PR other/69968
16196         * spellcheck-tree.c (levenshtein_distance): Rename to...
16197         (get_edit_distance): ...this, and update for underlying renaming.
16198         * spellcheck-tree.h (levenshtein_distance): Rename to...
16199         (get_edit_distance): ...this.
16200         * spellcheck.c (levenshtein_distance): Rename to...
16201         (get_edit_distance): ...this.  Convert from Levenshtein distance
16202         to Damerau-Levenshtein distance by supporting transpositions of
16203         adjacent characters.  Rename "v1" to "v_next" and "v0" to
16204         "v_one_ago".
16205         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
16206         (selftest::test_edit_distance_unit_test_oneway): ...this, and
16207         update for underlying renaming.
16208         (selftest::levenshtein_distance_unit_test): Rename to...
16209         (selftest::test_get_edit_distance_unit): ...this, and update for
16210         underlying renaming.
16211         (selftest::test_find_closest_string): Add example from PR 69968
16212         where transposition helps
16213         (selftest::test_metric_conditions): Update for renaming.
16214         (selftest::test_metric_conditions): Likewise.
16215         (selftest::spellcheck_c_tests): Likewise.
16216         * spellcheck.h (levenshtein_distance): Rename both overloads to...
16217         (get_edit_distance): ...this.
16218         (best_match::consider): Update for renaming.
16220 2018-06-12  Martin Sebor  <msebor@redhat.com>
16222         PR tree-optimization/85259
16223         * builtins.c (compute_objsize): Handle constant offsets.
16224         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
16225         true iff a warning has been issued.
16226         * gimple.h (gimple_nonartificial_location): New function.
16227         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
16228         gimple_nonartificial_location and handle -Wno-system-headers.
16229         (handle_builtin_stxncpy): Same.
16231 2018-06-12  Martin Sebor  <msebor@redhat.com>
16233         PR c/85931
16234         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
16236 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
16238         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16239         BUILTIN_VEC_XST entries for pointer to double and long long.
16241 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
16243         PR target/85990
16244         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16245         Update comments.
16246         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
16247         Likewise.
16249 2018-06-12  Martin Liska  <mliska@suse.cz>
16251         * doc/options.texi: Document IntegerRange.
16253 2018-06-12  Martin Liska  <mliska@suse.cz>
16255         * config/i386/i386.opt: Make MPX-related options as Deprecated.
16256         * opt-functions.awk: Handle Deprecated flag.
16257         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
16258         and report error.
16259         (read_cmdline_option): Report warning for a deprecated option.
16260         * opts.h (struct cl_option): Add new field cl_deprecated.
16261         (CL_ERR_DEPRECATED): New.
16263 2018-06-12  Martin Liska  <mliska@suse.cz>
16265         * doc/options.texi: Document Deprecated option flag.
16267 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16269         * config/arc/arc-arch.h (arc_extras): New enum.
16270         (arc_cpu_t):Add field extra.
16271         (arc_cpu_types): Consider the extras.
16272         * config/arc/arc-cpus.def: Add extras info.
16273         * config/arc/arc-opts.h (processor_type): Consider extra field.
16274         * config/arc/arc.c (arc_override_options): Handle extra field.
16276 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
16278         * config/arc/arc-arch.h: Update ARC_OPTX macro.
16279         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
16280         field.
16281         * config/arc/arc.c (arc_init): Update pic warning.
16282         (irq_range): Update irq range parsing warnings.
16283         (arc_override_options): Update various warning messages.
16284         (arc_handle_aux_attribute): Likewise.
16286 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
16288         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
16290 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
16292         * doc/sourcebuild.texi: Document usage of line number 0 in verify
16293         compiler messages directives.
16295 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
16297         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
16298         * config/mips/mips-tables.opt: Regenerate.
16299         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
16300         mips64r6.
16301         * doc/invoke.texi: Document -march=i6500.
16303 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
16305         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
16306         (i6400_gpmul): Add cpu_unit.
16307         (i6400_gpdiv): Likewise.
16308         (i6400_msa_add_d): Update reservations.
16309         (i6400_msa_int_add) Likewise.
16310         (i6400_msa_short_logic3) Likewise.
16311         (i6400_msa_short_logic2) Likewise.
16312         (i6400_msa_short_logic) Likewise.
16313         (i6400_msa_move) Likewise.
16314         (i6400_msa_cmp) Likewise.
16315         (i6400_msa_short_float2) Likewise.
16316         (i6400_msa_div_d) Likewise.
16317         (i6400_msa_long_logic1) Likewise.
16318         (i6400_msa_long_logic2) Likewise.
16319         (i6400_msa_mult) Likewise.
16320         (i6400_msa_long_float2) Likewise.
16321         (i6400_msa_long_float4) Likewise.
16322         (i6400_msa_long_float5) Likewise.
16323         (i6400_msa_long_float8) Likewise.
16324         (i6400_fpu_fadd): Include frint type.
16325         (i6400_fpu_store): New define_insn_reservation.
16326         (i6400_fpu_load): Likewise.
16327         (i6400_fpu_move): Likewise.
16328         (i6400_fpu_fcmp): Likewise.
16329         (i6400_fpu_fmadd): Likewise.
16330         (i6400_int_mult): Include imul3nc type and update reservation.
16331         (i6400_int_div): Include idiv3 type and update reservation.
16332         (i6400_int_load): Update to check type not move_type.
16333         (i6400_int_store): Likewise.
16334         (i6400_int_prefetch): Set zero latency.
16336 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
16338         * gcc.c: Document new %@{...} sequence.
16339         (LINK_COMMAND_SPEC): Use it for the -L switches.
16340         (cpp_unique_options): Use it for the -I switches.
16341         (at_file_argbuf): New global variable.
16342         (in_at_file): Likewise.
16343         (alloc_args): Create at_file_argbuf.
16344         (clear_args): Truncate at_file_argbuf.
16345         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
16346         (open_at_file): New function.
16347         (close_at_file): Likewise.
16348         (create_at_file): Delete.
16349         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
16350         <'o'>: Likewise.
16351         <'@'>: New case.
16352         (validate_switches_from_spec): Deal with %@{...} sequence.
16353         (validate_switches): Likewise.
16354         (driver::finalize): Call clear_args.
16356 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
16358         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
16360 2018-06-11  Martin Sebor  <msebor@redhat.com>
16362         * doc/invoke.texi (-Wall): List -Wc++17-compat.
16363         (Wno-class-memaccess): Add @opindex.
16364         (Wno-templates, Wno-multiple-inheritance): Same.
16365         (Wno-virtual-inheritance, Wno-namespaces): Same.
16366         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
16367         (Wno-format-overflow, Wno-format-truncation): Same.
16368         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
16369         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
16370         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
16371         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
16372         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
16373         (Wno-misspelled-isr): Same.
16375 2018-06-11  Martin Sebor  <msebor@redhat.com>
16377         * PR tree-optimization/86083
16378         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
16380 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
16382         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
16384 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
16386         PR target/85755
16387         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
16388         on the correct operand.
16389         (*movdi_internal64): Ditto.
16391 2018-06-11  Martin Liska  <mliska@suse.cz>
16393         PR tree-optimization/86089
16394         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
16396 2018-06-11  Julia Koval  <julia.koval@intel.com>
16398         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
16399         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
16400         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
16402 2018-06-11  Olivier Hainque  <hainque@adacore.com>
16404         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
16405         for Ada with strict dwarf2.
16407 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
16409         PR target/85755
16410         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
16411         addresses.
16413 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
16415         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
16417 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
16419         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
16420         TARGET_ELF.
16422 2018-06-08  Martin Liska  <mliska@suse.cz>
16424         * tree-cfg.h (debug_function): Fix argument type to match
16425         implementation.
16427 2018-06-08  Martin Liska  <mliska@suse.cz>
16429         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
16430         Remove usage of MPX-related (and removed) fields.
16431         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
16433 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
16435         * cfg.c (debug): Use TDF_NONE rather than 0.
16436         * cfghooks.c (debug): Likewise.
16437         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
16438         (struct dump_option_value_info): Convert to...
16439         (struct kv_pair): ...this template type.
16440         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
16441         rather than 0.
16442         (optinfo_verbosity_options): Likewise.
16443         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
16444         OPTGROUP_NONE.
16445         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
16446         than int for "optgroup_flags" param.
16447         (dump_generic_expr_loc): Use dump_flags_t rather than int for
16448         "dump_kind" param.
16449         (dump_dec): Likewise.
16450         (dump_finish): Use TDF_NONE rather than 0.
16451         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
16452         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
16453         than 0.  Update for change to option_ptr.
16454         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
16455         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
16456         0.  Update for changes to optinfo_verbosity_options and
16457         optgroup_options.
16458         (opt_info_switch_p): Convert optgroup_flags from int to
16459         optgroup_flags_t.
16460         (dump_basic_block): Use dump_flags_t rather than int
16461         for "dump_kind" param.
16462         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
16463         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
16464         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
16465         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
16466         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
16467         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
16468         TDF_NONE): Convert from macros to...
16469         (enum dump_flag): ...this new enum.
16470         (dump_flags_t): Update to use enum.
16471         (operator|, operator&, operator~, operator|=, operator&=):
16472         Implement for dump_flags_t.
16473         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
16474         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
16475         Convert from macros to...
16476         (enum optgroup_flag): ...this new enum.
16477         (optgroup_flags_t): New typedef.
16478         (operator|, operator|=): Implement for optgroup_flags_t.
16479         (struct dump_file_info): Convert field "alt_flags" to
16480         dump_flags_t.  Convert field "optgroup_flags" to
16481         optgroup_flags_t.
16482         (dump_basic_block): Use dump_flags_t rather than int for param.
16483         (dump_generic_expr_loc): Likewise.
16484         (dump_dec): Likewise.
16485         (dump_register): Convert param "optgroup_flags" to
16486         optgroup_flags_t.
16487         (opt_info_enable_passes): Likewise.
16488         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
16489         than 0.
16490         * gimple-pretty-print.c (debug): Likewise.
16491         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
16492         (merged_store_group::apply_stores): Likewise.
16493         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
16494         * gimple.c (verify_gimple_pp): Likewise.
16495         * graphite-poly.c (print_pbb_body): Likewise.
16496         * passes.c (pass_manager::register_one_dump_file): Convert
16497         local "optgroup_flags" to optgroup_flags_t.
16498         * print-tree.c (print_node): Use TDF_NONE rather than 0.
16499         (debug): Likewise.
16500         (debug_body): Likewise.
16501         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
16502         to optgroup_flags_t.
16503         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
16504         than 0.
16505         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
16506         (convert_mult_to_fma): Likewise.
16507         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
16508         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
16509         * tree-vect-data-refs.c (dump_lower_bound): Convert param
16510         "dump_kind" to dump_flags_t.
16512 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16514         * config/rs6000/rs6000.c (min, max): Delete.
16516 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
16518         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
16519         -mabi=spe and -mabi=no-spe.
16521 2018-06-08  Martin Liska  <mliska@suse.cz>
16523         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
16524         where we expect an existing summary.
16526 2018-06-08  Martin Liska  <mliska@suse.cz>
16528         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
16529         * ipa-inline.h (estimate_edge_growth): Likewise.
16531 2018-06-08  Martin Liska  <mliska@suse.cz>
16533         * cgraph.c (function_version_hasher::hash): Use
16534         cgraph_node::get_uid ().
16535         (function_version_hasher::equal):
16536         * cgraph.h (cgraph_node::get_uid): New method.
16537         * ipa-inline.c (update_caller_keys): Use
16538         cgraph_node::get_uid ().
16539         (update_callee_keys): Likewise.
16540         * ipa-utils.c (searchc): Likewise.
16541         (ipa_reduced_postorder): Likewise.
16542         * lto-cgraph.c (input_node): Likewise.
16543         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
16544         * symbol-summary.h (symtab_insertion): Likewise.
16545         (symtab_removal): Likewise.
16546         (symtab_duplication): Likewise.
16547         * tree-pretty-print.c (dump_function_header): Likewise.
16548         * tree-sra.c (convert_callers_for_node): Likewise.
16550 2018-06-08  Martin Liska  <mliska@suse.cz>
16552         * cgraph.c (symbol_table::create_edge): Always assign a new
16553         unique number.
16554         (symbol_table::free_edge): Do not recycle numbers.
16555         * cgraph.h (cgraph_edge::get): New method.
16556         * symbol-summary.h (symtab_removal): Use it.
16557         (symtab_duplication): Likewise.
16558         (call_summary::hashable_uid): Remove.
16560 2018-06-08  Martin Liska  <mliska@suse.cz>
16562         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
16563         (initialize_growth_caches): Remove.
16564         (free_growth_caches): Likewise.
16565         (do_estimate_edge_time): Use edge_growth_cache.
16566         (do_estimate_edge_size): Likewise.
16567         (do_estimate_edge_hints): Likewise.
16568         * ipa-inline.c (reset_edge_caches): Likewise.
16569         (recursive_inlining): Likewise.
16570         (inline_small_functions): Likewise.
16571         * ipa-inline.h (initialize_growth_caches): Remove.
16572         (estimate_edge_size): Likewise.
16573         (estimate_edge_time): Likewise.
16574         (estimate_edge_hints): Likewise.
16575         (reset_edge_growth_cache): Likewise.
16576         * symbol-summary.h (call_summary::remove): New method.
16578 2018-06-08  Martin Liska  <mliska@suse.cz>
16580         * ipa-cp.c (class edge_clone_summary): New summary.
16581         (grow_edge_clone_vectors): Remove.
16582         (ipcp_edge_duplication_hook): Remove.
16583         (class edge_clone_summary_t): New call_summary class.
16584         (ipcp_edge_removal_hook): Remove.
16585         (edge_clone_summary_t::duplicate): New function.
16586         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
16587         (create_specialized_node): Likewise.
16588         (ipcp_driver): Initialize edge_clone_summaries and do not
16589         register hooks.
16591 2018-06-08  Martin Liska  <mliska@suse.cz>
16593         * symbol-summary.h (get): New function.
16594         (call_summary::m_initialize_when_cloning): New class member.
16596 2018-06-08  Martin Liska  <mliska@suse.cz>
16598         * cgraph.c (cgraph_node::remove): Do not recycle uid.
16599         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
16600         (symbol_table::allocate_cgraph_symbol): Do not set uid.
16601         * passes.c (uid_hash_t): Record removed_nodes by their uids.
16602         (remove_cgraph_node_from_order): Use the removed_nodes set.
16603         (do_per_function_toporder): Likwise.
16604         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
16605         instead of summary_uid.
16606         (symtab_removal): Likewise.
16607         (symtab_duplication): Likewise.
16609 2018-06-08  Martin Liska  <mliska@suse.cz>
16611         * ipa-cp.c (ipcp_store_bits_results): Use
16612         ipcp_transformation_sum.
16613         (ipcp_store_vr_results): Likewise.
16614         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
16615         to ...
16616         (ipcp_transformation_initialize): ... this.
16617         (ipa_set_node_agg_value_chain):
16618         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
16619         (write_ipcp_transformation_info): Likewise.
16620         (read_ipcp_transformation_info): Likewise.
16621         (ipcp_update_bits): Likewise.
16622         (ipcp_update_vr): Likewise.
16623         (ipcp_transform_function): Likewise.
16624         * ipa-prop.h: Rename ipcp_transformation_summary to
16625         ipcp_transformation.
16626         (class ipcp_transformation_t): New function summary.
16627         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
16628         (ipa_get_agg_replacements_for_node): Likewise.
16630 2018-06-08  Martin Liska  <mliska@suse.cz>
16632         * ipa-pure-const.c (struct funct_state_d): Do it class instead
16633         of struct.
16634         (class funct_state_summary_t): New function_summary class.
16635         (has_function_state): Remove.
16636         (get_function_state): Likewise.
16637         (set_function_state): Likewise.
16638         (add_new_function): Likewise.
16639         (funct_state_summary_t::insert): New function.
16640         (duplicate_node_data): Remove.
16641         (remove_node_data): Remove.
16642         (funct_state_summary_t::duplicate): New function.
16643         (register_hooks): Create new funct_state_summaries.
16644         (pure_const_generate_summary): Use it.
16645         (pure_const_write_summary): Likewise.
16646         (pure_const_read_summary): Likewise.
16647         (propagate_pure_const): Likewise.
16648         (propagate_nothrow): Likewise.
16649         (dump_malloc_lattice): Likewise.
16650         (propagate_malloc): Likewise.
16651         (execute): Do not register hooks, just remove summary
16652         instead.
16653         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
16654         constructor.
16656 2018-06-08  Martin Liska  <mliska@suse.cz>
16658         * ipa-reference.c (remove_node_data): Remove.
16659         (duplicate_node_data): Likewise.
16660         (class ipa_ref_var_info_summary_t): New class.
16661         (class ipa_ref_opt_summary_t): Likewise.
16662         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
16663         (get_reference_optimization_summary): Use
16664         ipa_ref_opt_sum_summaries.
16665         (set_reference_vars_info): Remove.
16666         (set_reference_optimization_summary): Likewise.
16667         (ipa_init): Create summaries.
16668         (init_function_info): Use function summary.
16669         (ipa_ref_opt_summary_t::duplicate): New function.
16670         (ipa_ref_opt_summary_t::remove): New function.
16671         (get_read_write_all_from_node): Fix GNU coding style.
16672         (propagate): Use function summary.
16673         (write_node_summary_p): Fix GNU coding style.
16674         (stream_out_bitmap): Likewise.
16675         (ipa_reference_read_optimization_summary): Use function summary.
16676         (ipa_reference_c_finalize): Do not release hooks.
16678 2018-06-08  Martin Liska  <mliska@suse.cz>
16680         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
16681         (analyze_function_body): Extract multiple calls of get_create.
16682         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16683         * ipa-inline.c (recursive_inlining): Use ::get method.
16684         * ipa-inline.h (estimate_edge_growth): Likewise.
16686 2018-06-08  Martin Liska  <mliska@suse.cz>
16688         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
16689         HSA_INVALID.
16690         (hsa_function_summary::hsa_function_summary): Use the new enum
16691         value.
16692         (hsa_gpu_implementation_p): Use hsa_summaries::get.
16693         * hsa-gen.c (hsa_get_host_function): Likewise.
16694         (get_brig_function_name): Likewise.
16695         * ipa-hsa.c (process_hsa_functions): Likewise.
16696         (ipa_hsa_write_summary): Likewise.
16697         * symbol-summary.h (symtab_duplication): Use ::get function/
16698         (get): New function.
16700 2018-06-08  Martin Liska  <mliska@suse.cz>
16702         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
16703         of get.
16704         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
16705         (hsa_register_kernel): Likewise.
16706         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
16707         * hsa-gen.c (hsa_get_host_function): Likewise.
16708         (get_brig_function_name): Likewise.
16709         (generate_hsa): Likewise.
16710         (pass_gen_hsail::execute): Likewise.
16711         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
16712         (devirtualization_time_bonus): Likewise.
16713         (ipcp_propagate_stage): Likewise.
16714         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
16715         (edge_set_predicate): Likewise.
16716         (evaluate_conditions_for_known_args): Likewise.
16717         (evaluate_properties_for_edge): Likewise.
16718         (ipa_fn_summary::reset): Likewise.
16719         (ipa_fn_summary_t::duplicate): Likewise.
16720         (dump_ipa_call_summary): Likewise.
16721         (ipa_dump_fn_summary): Likewise.
16722         (analyze_function_body): Likewise.
16723         (compute_fn_summary): Likewise.
16724         (estimate_edge_devirt_benefit): Likewise.
16725         (estimate_edge_size_and_time): Likewise.
16726         (estimate_calls_size_and_time): Likewise.
16727         (estimate_node_size_and_time): Likewise.
16728         (inline_update_callee_summaries): Likewise.
16729         (remap_edge_change_prob): Likewise.
16730         (remap_edge_summaries): Likewise.
16731         (ipa_merge_fn_summary_after_inlining): Likewise.
16732         (ipa_update_overall_fn_summary): Likewise.
16733         (read_ipa_call_summary): Likewise.
16734         (inline_read_section): Likewise.
16735         (write_ipa_call_summary): Likewise.
16736         (ipa_fn_summary_write): Likewise.
16737         (ipa_free_fn_summary): Likewise.
16738         * ipa-hsa.c (process_hsa_functions): Likewise.
16739         (ipa_hsa_write_summary): Likewise.
16740         (ipa_hsa_read_section): Likewise.
16741         * ipa-icf.c (sem_function::merge): Likewise.
16742         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
16743         (do_estimate_edge_time): Likewise.
16744         (estimate_size_after_inlining): Likewise.
16745         (estimate_growth): Likewise.
16746         (growth_likely_positive): Likewise.
16747         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
16748         (inline_call): Likewise.
16749         * ipa-inline.c (caller_growth_limits): Likewise.
16750         (can_inline_edge_p): Likewise.
16751         (can_inline_edge_by_limits_p): Likewise.
16752         (compute_uninlined_call_time): Likewise.
16753         (compute_inlined_call_time): Likewise.
16754         (want_inline_small_function_p): Likewise.
16755         (edge_badness): Likewise.
16756         (update_caller_keys): Likewise.
16757         (update_callee_keys): Likewise.
16758         (recursive_inlining): Likewise.
16759         (inline_small_functions): Likewise.
16760         (inline_to_all_callers_1): Likewise.
16761         (dump_overall_stats): Likewise.
16762         (early_inline_small_functions): Likewise.
16763         (early_inliner): Likewise.
16764         * ipa-inline.h (estimate_edge_growth): Likewise.
16765         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
16766         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
16767         * ipa-prop.h (IPA_NODE_REF): Likewise.
16768         (IPA_EDGE_REF): Likewise.
16769         * ipa-pure-const.c (malloc_candidate_p): Likewise.
16770         (propagate_malloc): Likewise.
16771         * ipa-split.c (execute_split_functions): Likewise.
16772         * symbol-summary.h: Rename get to get_create.
16773         (get): Likewise.
16774         (get_create): Likewise.
16775         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
16777 2018-06-08  Martin Liska  <mliska@suse.cz>
16779         * symbol-summary.h (release): Move definition out of class
16780         declaration.
16781         (symtab_removal): Likewise.
16782         (symtab_duplication): Likewise.
16784 2018-06-08  Martin Liska  <mliska@suse.cz>
16786         * symbol-summary.h (function_summary): Move constructor
16787         implementation out of class declaration.
16788         (release): Likewise.
16789         (symtab_insertion): Likewise.
16790         (symtab_removal): Likewise.
16791         (symtab_duplication): Likewise.
16792         (get): Likewise.
16794 2018-06-08  Martin Liska  <mliska@suse.cz>
16796         * Makefile.in: Remove support for MPX (macros, related functions,
16797         fields in cgraph_node, ...).
16798         * builtin-types.def (BT_BND): Likewise.
16799         (BT_FN_BND_CONST_PTR): Likewise.
16800         (BT_FN_CONST_PTR_BND): Likewise.
16801         (BT_FN_VOID_PTR_BND): Likewise.
16802         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
16803         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
16804         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
16805         (expand_builtin_mempcpy_with_bounds): Likewise.
16806         (expand_builtin_memset_with_bounds): Likewise.
16807         (expand_builtin_memset_args): Likewise.
16808         (std_expand_builtin_va_start): Likewise.
16809         (expand_builtin): Likewise.
16810         (expand_builtin_with_bounds): Likewise.
16811         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
16812         (DEF_LIB_BUILTIN_CHKP): Likewise.
16813         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
16814         (DEF_CHKP_BUILTIN): Likewise.
16815         (BUILT_IN_MEMCPY): Likewise.
16816         (BUILT_IN_MEMMOVE): Likewise.
16817         (BUILT_IN_MEMPCPY): Likewise.
16818         (BUILT_IN_MEMSET): Likewise.
16819         (BUILT_IN_STPCPY): Likewise.
16820         (BUILT_IN_STRCAT): Likewise.
16821         (BUILT_IN_STRCHR): Likewise.
16822         (BUILT_IN_STRCPY): Likewise.
16823         (BUILT_IN_STRLEN): Likewise.
16824         (BUILT_IN_MEMCPY_CHK): Likewise.
16825         (BUILT_IN_MEMMOVE_CHK): Likewise.
16826         (BUILT_IN_MEMPCPY_CHK): Likewise.
16827         (BUILT_IN_MEMSET_CHK): Likewise.
16828         (BUILT_IN_STPCPY_CHK): Likewise.
16829         (BUILT_IN_STRCAT_CHK): Likewise.
16830         (BUILT_IN_STRCPY_CHK): Likewise.
16831         * calls.c (store_bounds): Likewise.
16832         (emit_call_1): Likewise.
16833         (special_function_p): Likewise.
16834         (maybe_warn_nonstring_arg): Likewise.
16835         (initialize_argument_information): Likewise.
16836         (finalize_must_preallocate): Likewise.
16837         (compute_argument_addresses): Likewise.
16838         (expand_call): Likewise.
16839         * cfgexpand.c (expand_call_stmt): Likewise.
16840         (expand_return): Likewise.
16841         (expand_gimple_stmt_1): Likewise.
16842         (pass_expand::execute): Likewise.
16843         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
16844         (cgraph_node::remove): Likewise.
16845         (cgraph_node::dump): Likewise.
16846         (cgraph_node::verify_node): Likewise.
16847         * cgraph.h (chkp_function_instrumented_p): Likewise.
16848         (symtab_node::get_alias_target): Likewise.
16849         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
16850         (cgraph_local_p): Likewise.
16851         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
16852         (cgraph_edge::rebuild_references): Likewise.
16853         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
16854         (walk_polymorphic_call_targets): Likewise.
16855         (cgraph_node::expand_thunk): Likewise.
16856         (symbol_table::output_weakrefs): Likewise.
16857         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
16858         (ix86_handle_option): Likewise.
16859         * config/i386/constraints.md: Likewise.
16860         * config/i386/i386-builtin-types.def (BND): Likewise.
16861         (VOID): Likewise.
16862         (PVOID): Likewise.
16863         (ULONG): Likewise.
16864         * config/i386/i386-builtin.def (BDESC_END): Likewise.
16865         (BDESC_FIRST): Likewise.
16866         (BDESC): Likewise.
16867         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
16868         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
16869         * config/i386/i386.c (enum reg_class): Likewise.
16870         (ix86_target_string): Likewise.
16871         (ix86_option_override_internal): Likewise.
16872         (ix86_conditional_register_usage): Likewise.
16873         (ix86_valid_target_attribute_inner_p): Likewise.
16874         (ix86_set_indirect_branch_type): Likewise.
16875         (ix86_set_current_function): Likewise.
16876         (ix86_function_arg_regno_p): Likewise.
16877         (init_cumulative_args): Likewise.
16878         (ix86_function_arg_advance): Likewise.
16879         (ix86_function_arg): Likewise.
16880         (ix86_pass_by_reference): Likewise.
16881         (ix86_function_value_regno_p): Likewise.
16882         (ix86_function_value_1): Likewise.
16883         (ix86_function_value_bounds): Likewise.
16884         (ix86_return_in_memory): Likewise.
16885         (ix86_setup_incoming_vararg_bounds): Likewise.
16886         (ix86_va_start): Likewise.
16887         (indirect_thunk_need_prefix): Likewise.
16888         (print_reg): Likewise.
16889         (ix86_print_operand): Likewise.
16890         (ix86_expand_call): Likewise.
16891         (ix86_output_function_return): Likewise.
16892         (reg_encoded_number): Likewise.
16893         (BDESC_VERIFYS): Likewise.
16894         (ix86_init_mpx_builtins): Likewise.
16895         (ix86_init_builtins): Likewise.
16896         (ix86_emit_cmove): Likewise.
16897         (ix86_emit_move_max): Likewise.
16898         (ix86_expand_builtin): Likewise.
16899         (ix86_builtin_mpx_function): Likewise.
16900         (ix86_get_arg_address_for_bt): Likewise.
16901         (ix86_load_bounds): Likewise.
16902         (ix86_store_bounds): Likewise.
16903         (ix86_load_returned_bounds): Likewise.
16904         (ix86_store_returned_bounds): Likewise.
16905         (ix86_class_likely_spilled_p): Likewise.
16906         (ix86_hard_regno_mode_ok): Likewise.
16907         (x86_order_regs_for_local_alloc): Likewise.
16908         (ix86_mitigate_rop): Likewise.
16909         (ix86_bnd_prefixed_insn_p): Likewise.
16910         (ix86_mpx_bound_mode): Likewise.
16911         (ix86_make_bounds_constant): Likewise.
16912         (ix86_initialize_bounds): Likewise.
16913         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
16914         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
16915         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
16916         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
16917         (TARGET_CHKP_BOUND_MODE): Likewise.
16918         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
16919         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
16920         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
16921         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
16922         * config/i386/i386.h (TARGET_MPX): Likewise.
16923         (TARGET_MPX_P): Likewise.
16924         (VALID_BND_REG_MODE): Likewise.
16925         (FIRST_BND_REG): Likewise.
16926         (LAST_BND_REG): Likewise.
16927         (enum reg_class): Likewise.
16928         (BND_REG_P): Likewise.
16929         (BND_REGNO_P): Likewise.
16930         (BNDmode): Likewise.
16931         (ADJUST_INSN_LENGTH): Likewise.
16932         * config/i386/i386.md: Likewise.
16933         * config/i386/i386.opt: Likewise.
16934         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
16935         (defined): Likewise.
16936         (LINK_MPX): Likewise.
16937         (MPX_SPEC): Likewise.
16938         (LIBMPX_SPEC): Likewise.
16939         (LIBMPXWRAPPERS_SPEC): Likewise.
16940         (CHKP_SPEC): Likewise.
16941         * config/i386/predicates.md: Likewise.
16942         * dbxout.c (dbxout_type): Likewise.
16943         * doc/extend.texi: Likewise.
16944         * doc/invoke.texi: Likewise.
16945         * doc/md.texi: Likewise.
16946         * doc/tm.texi: Likewise.
16947         * doc/tm.texi.in: Likewise.
16948         * dwarf2out.c (is_base_type): Likewise.
16949         (gen_formal_types_die): Likewise.
16950         (gen_subprogram_die): Likewise.
16951         (gen_type_die_with_usage): Likewise.
16952         (gen_decl_die): Likewise.
16953         (dwarf2out_late_global_decl): Likewise.
16954         * expr.c (expand_assignment): Likewise.
16955         (emit_storent_insn): Likewise.
16956         (store_expr_with_bounds): Likewise.
16957         (store_expr): Likewise.
16958         (expand_expr_real_1): Likewise.
16959         * expr.h (store_expr_with_bounds): Likewise.
16960         * function.c (use_register_for_decl): Likewise.
16961         (struct bounds_parm_data): Likewise.
16962         (assign_parms_augmented_arg_list): Likewise.
16963         (assign_parm_find_entry_rtl): Likewise.
16964         (assign_parm_is_stack_parm): Likewise.
16965         (assign_parm_load_bounds): Likewise.
16966         (assign_bounds): Likewise.
16967         (assign_parms): Likewise.
16968         (expand_function_start): Likewise.
16969         * gcc.c (CHKP_SPEC): Likewise.
16970         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16971         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
16972         (wrestrict_dom_walker::check_call): Likewise.
16973         * gimple.c (gimple_build_call_from_tree): Likewise.
16974         * gimple.h (enum gf_mask): Likewise.
16975         (gimple_call_with_bounds_p): Likewise.
16976         (gimple_call_set_with_bounds): Likewise.
16977         * gimplify.c (gimplify_init_constructor): Likewise.
16978         * ipa-cp.c (initialize_node_lattices): Likewise.
16979         (propagate_constants_across_call): Likewise.
16980         (find_more_scalar_values_for_callers_subset): Likewise.
16981         * ipa-hsa.c (process_hsa_functions): Likewise.
16982         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
16983         * ipa-icf.c (sem_function::merge): Likewise.
16984         * ipa-inline.c (early_inliner): Likewise.
16985         * ipa-pure-const.c (warn_function_noreturn): Likewise.
16986         (warn_function_cold): Likewise.
16987         (propagate_pure_const): Likewise.
16988         * ipa-ref.h (enum GTY): Likewise.
16989         * ipa-split.c (find_retbnd): Likewise.
16990         (consider_split): Likewise.
16991         (split_function): Likewise.
16992         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
16993         * ipa.c (walk_polymorphic_call_targets): Likewise.
16994         (symbol_table::remove_unreachable_nodes): Likewise.
16995         (process_references): Likewise.
16996         (cgraph_build_static_cdtor_1): Likewise.
16997         * lto-cgraph.c (lto_output_node): Likewise.
16998         (output_refs): Likewise.
16999         (compute_ltrans_boundary): Likewise.
17000         (input_overwrite_node): Likewise.
17001         (input_node): Likewise.
17002         (input_cgraph_1): Likewise.
17003         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
17004         * passes.c (pass_manager::execute_early_local_passes): Likewise.
17005         (class pass_chkp_instrumentation_passes): Likewise.
17006         (make_pass_chkp_instrumentation_passes): Likewise.
17007         * passes.def: Likewise.
17008         * rtl.h (struct GTY): Likewise.
17009         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
17010         * stor-layout.c (layout_type): Likewise.
17011         * symtab.c: Likewise.
17012         * target.def: Likewise.
17013         * targhooks.c (default_chkp_bound_type): Likewise.
17014         (default_chkp_bound_mode): Likewise.
17015         (default_builtin_chkp_function): Likewise.
17016         (default_chkp_function_value_bounds): Likewise.
17017         (default_chkp_make_bounds_constant): Likewise.
17018         (default_chkp_initialize_bounds): Likewise.
17019         * targhooks.h (default_chkp_bound_type): Likewise.
17020         (default_chkp_bound_mode): Likewise.
17021         (default_builtin_chkp_function): Likewise.
17022         (default_chkp_function_value_bounds): Likewise.
17023         (default_chkp_make_bounds_constant): Likewise.
17024         (default_chkp_initialize_bounds): Likewise.
17025         * toplev.c (compile_file): Likewise.
17026         (process_options): Likewise.
17027         * tree-core.h (DEF_BUILTIN): Likewise.
17028         (DEF_BUILTIN_CHKP): Likewise.
17029         * tree-inline.c (declare_return_variable): Likewise.
17030         (remap_gimple_stmt): Likewise.
17031         (copy_bb): Likewise.
17032         (initialize_inlined_parameters): Likewise.
17033         (expand_call_inline): Likewise.
17034         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
17035         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
17036         (make_pass_ipa_chkp_produce_thunks): Likewise.
17037         (make_pass_chkp): Likewise.
17038         (make_pass_chkp_opt): Likewise.
17039         (make_pass_chkp_instrumentation_passes): Likewise.
17040         * tree-pretty-print.c (dump_generic_node): Likewise.
17041         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
17042         * tree-ssa-dce.c (propagate_necessity): Likewise.
17043         (eliminate_unnecessary_stmts): Likewise.
17044         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
17045         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
17046         * tree-ssa-sccvn.h: Likewise.
17047         * tree-ssa-strlen.c (get_string_length): Likewise.
17048         (valid_builtin_call): Likewise.
17049         (adjust_last_stmt): Likewise.
17050         (handle_builtin_strchr): Likewise.
17051         (handle_builtin_strcpy): Likewise.
17052         (handle_builtin_stxncpy): Likewise.
17053         (handle_builtin_memcpy): Likewise.
17054         (handle_builtin_strcat): Likewise.
17055         (strlen_check_and_optimize_stmt): Likewise.
17056         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
17057         * tree-streamer-in.c: Likewise.
17058         * tree-streamer.c (record_common_node): Likewise.
17059         * tree.c (tree_code_size): Likewise.
17060         (wide_int_to_tree_1): Likewise.
17061         (type_contains_placeholder_1): Likewise.
17062         (build_common_tree_nodes): Likewise.
17063         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
17064         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
17065         (POINTER_BOUNDS_P): Likewise.
17066         (BOUNDED_TYPE_P): Likewise.
17067         (BOUNDED_P): Likewise.
17068         (CALL_WITH_BOUNDS_P): Likewise.
17069         (pointer_bounds_type_node): Likewise.
17070         * value-prof.c (gimple_ic): Likewise.
17071         * var-tracking.c (vt_add_function_parameters): Likewise.
17072         * varasm.c (make_decl_rtl): Likewise.
17073         (assemble_start_function): Likewise.
17074         (output_constant): Likewise.
17075         (maybe_assemble_visibility): Likewise.
17076         * varpool.c (ctor_for_folding): Likewise.
17077         * chkp-builtins.def: Remove.
17078         * ipa-chkp.c: Remove.
17079         * ipa-chkp.h: Remove.
17080         * rtl-chkp.c: Remove.
17081         * rtl-chkp.h: Remove.
17082         * tree-chkp-opt.c: Remove.
17083         * tree-chkp.c: Remove.
17084         * tree-chkp.h: Remove.
17086 2018-06-07  Carl Love  <cel@us.ibm.com>
17088         * config/rs6000/vsx.md (vextract_fp_from_shorth,
17089         vextract_fp_from_shortl): Add BE support.
17091 2018-06-07  Paul Koning  <ni1d@arrl.net>
17093         * compare-elim.c (try_merge_compare): Don't merge compare if
17094         address contains a side effect.
17095         (try_eliminate_compare): Likewise.
17097 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
17099         * config.gcc: Support "tremont".
17100         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
17101         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
17102         PROCESSOR_TREMONT.
17103         * config/i386/i386.c (m_TREMONT): Define.
17104         (processor_target_table): Add "tremont".
17105         (PTA_TREMONT): Define.
17106         (ix86_lea_outperforms): Add TARGET_TREMONT.
17107         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
17108         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
17109         and M_INTEL_GOLDMONT_PLUS.
17110         (fold_builtin_cpu): Add "tremont".
17111         (ix86_add_stmt_cost): Add TARGET_TREMONT.
17112         (ix86_option_override_internal): Add "tremont".
17113         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
17114         (processor_type): Add PROCESSOR_TREMONT.
17115         * config/i386/x86-tune.def: Add m_TREMONT.
17116         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
17118 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17120         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
17121         symbol defined for msp430i* devices to be lower case.
17123 2018-06-07  Richard Biener  <rguenther@suse.de>
17125         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
17126         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
17127         Properly wrap signed arithmetic if overflow wraps.
17129 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
17131         PR tree-optimization/69615
17132         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
17133         of a cast from a same precision integral SSA_NAME in a bb dominated
17134         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
17135         cast to utype if rhs2 has already a compatible type.
17137 2018-06-07  Richard Biener  <rguenther@suse.de>
17139         PR tree-optimization/85935
17140         * graphite-scop-detection.c (find_params_in_bb): Analyze
17141         condition operands with respect to the correct loop.  Assert
17142         the analysis doesn't fail.
17144 2018-06-04  Carl Love  <cel@us.ibm.com>
17146         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
17147         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
17148         as it is slightly cheaper.
17149         (first_match_or_eos_index_<mode>):
17150         Calculate index using natural element order.
17151         (first_match_index_<mode>):
17152         Calculate index using natural element order.
17153         (first_match_or_eos_index_<mode>):
17154         Calculate index using natural order.
17155         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
17156         for BE and LE modes.
17157         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
17158         P9V_BUILTIN_VCLZLSBB_V16QI.
17159         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
17160         specific.
17162 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17164         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
17165         indentation and line wrap for many prototypes.  Add missing
17166         @smallexample directives around block of prototypes for vec_xl and
17167         vec_xst.
17169 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
17171         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
17172         track if we pass or return IEEE 128-bit floating point.
17173         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
17174         C++ mangling that is compatible with GCC 8.1.
17175         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
17176         (init_cumulative_args): Note if we pass or return IEEE 128-bit
17177         floating point types.
17178         (rs6000_function_arg_advance_1): Likewise.
17179         (rs6000_mangle_type): Optionally generate mangled names that match
17180         what GCC 8.1 generated for IEEE 128-bit floating point types.
17181         (rs6000_globalize_decl_name): If we have an external function that
17182         passes or returns IEEE 128-bit types, generate a weak reference
17183         from the mangled name used in GCC 8.1 to the current mangled
17184         name.
17185         (rs6000_init_builtins): Make __ibm128 use the long double type if
17186         long double is IBM extended double.  Make __float128 use the long
17187         double type if long double is IEEE 128-bit.
17189         PR target/85657
17190         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
17191         macro for __ibm128 built-in functions.
17192         (PACK_IF): Add __ibm128 pack/unpack functions.
17193         (UNPACK_IF): Likewise.
17194         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
17195         enable long double built-in functions if long double is IEEE
17196         128-bit floating point.
17197         (rs6000_invalid_builtin): Update long double built-in function
17198         error message.
17199         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
17200         functions, adjust the built-in function to use the long double
17201         built-in function if __ibm128 and long double are the same type.
17202         * doc/extend.texi (PowerPC builtins): Update documention for
17203         __builtin_{,un}pack_longdouble.  Add documentation for
17204         __builtin_{,un}pack_ibm128.
17206 2018-06-06  Jim Wilson  <jimw@sifive.com>
17208         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
17209         (struct machine_function): New field interrupt_mode.
17210         (riscv_handle_type_attribute): New function.  Add forward declaration.
17211         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
17212         (riscv_expand_epilogue): Check interrupt_mode field.
17213         (riscv_set_current_function): Check interrupt attribute args and
17214         set interrupt_mode field.
17215         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
17216         (riscv_sret, riscv_uret): New.
17217         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
17218         new arguments to interrupt attribute.
17220 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
17222         PR target/63177
17223         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
17224         Don't handle -mcpu=power8 if -mpower9-vector is also used.
17226 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17228         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
17229         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
17230         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
17231         several redundant entries.
17233 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
17235         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
17236         type from "rtx" to "rtx_insn *".
17237         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
17238         for local "call_insn", removing cast.
17239         (ix86_expand_call): Likewise, introducing a "call_insn" local.
17241 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
17243         PR tree-optimization/86066
17244         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
17245         for BIT_INSERT_EXPR stores.
17247 2018-06-06  Richard Biener  <rguenther@suse.de>
17249         PR tree-optimization/86062
17250         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
17251         component refs ontop
17252         of to be offsetted base.
17254 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17256         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
17257         to be static and remove check on interrupt attribute name.
17259 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17261         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17262         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
17264 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
17266         PR target/79924
17267         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
17268         second argument.
17269         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
17270         Remove second argument, change how error is called.
17271         (aarch64_layout_arg): Remove second argument from
17272         aarch64_err_no_fpadvsimd call.
17273         (aarch64_init_cumulative_args): Ditto.
17274         (aarch64_gimplify_va_arg_expr): Ditto.
17275         * config/aarch64/aarch64.md (mov<mode>): Ditto.
17277 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
17279         * config/i386/i386.md (simple_return_indirect_internal): New expander.
17280         (*simple_return_indirect_internal<mode>): Rename from
17281         simple_return_indirect_internal.  Use W mode iterator.
17282         (rstorssp): New expander.
17283         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
17284         (clrssbsy): New expander.
17285         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
17287 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17289         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
17290         __typeof__.
17291         (cmse_check_pointed_object): Likewise.
17293 2018-06-05  Martin Liska  <mliska@suse.cz>
17295         PR gcov-profile/47618
17296         * doc/invoke.texi: Document how -fprofile-dir format
17297         is extended.
17299 2018-06-05  Richard Biener  <rguenther@suse.de>
17301         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
17302         removal pretend DOM info isn't available so we do not update
17303         it and only remove edges, not dominated blocks.  Actually free
17304         DOM info in case we removed something.  Remove unreachable blocks.
17305         (mfb_keep_latches): Work with either DOM info or marked backedges.
17306         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
17307         first.  Mark backedges if DOM info isn't available.
17308         (Re-)compute DOM info after cleanup_control_flow_pre.
17310 2018-06-05  Richard Biener  <rguenther@suse.de>
17312         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
17313         (locus_discrim_hasher::hash): Adjust.
17314         (locus_discrim_hasher::equal): Likewise.
17315         (next_discriminator_for_locus): Work on line directly.
17316         (same_line_p): Pass in expanded locus1 as well.
17317         (assign_discriminators): Avoid redundant location expansions.
17319 2018-06-05  Richard Biener  <rguenther@suse.de>
17321         PR tree-optimization/86046
17322         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
17323         if required after clearing TREE_ADDRESSABLE.
17325 2018-06-05  Richard Biener  <rguenther@suse.de>
17327         PR tree-optimization/86047
17328         * tree-ssa-loop.c (for_each_index): Glob handling of all
17329         decls and constants and really handle all of them.
17331 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17333         PR target/81497
17334         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
17335         qualifier_void_pointer and qualifier_const_void_pointer.
17336         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
17337         (arm_init_builtins): Handle the above.
17338         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
17339         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
17340         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
17341         void intrinsics.
17343 2018-06-05  Martin Liska  <mliska@suse.cz>
17345         * auto-profile.c (read_autofdo_file): Do not use
17346         gcov_ctr_summary struct.
17347         (afdo_callsite_hot_enough_for_early_inline): Likewise.
17348         * coverage.c (struct counts_entry): Likewise.
17349         (read_counts_file): Read just single summary entry.
17350         (get_coverage_counts): Use gcov_summary struct.
17351         * coverage.h (get_coverage_counts): Likewise.
17352         * gcov-dump.c (dump_working_sets): Likewise.
17353         (tag_summary): Dump just single summary.
17354         * gcov-io.c (gcov_write_summary): Write just histogram
17355         summary.
17356         (gcov_read_summary): Read just single summary.
17357         (compute_working_sets): Use gcov_summary struct.
17358         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
17359         of GCOV_COUNTERS_SUMMABLE.
17360         (GCOV_COUNTERS_SUMMABLE): Remove.
17361         (GCOV_FIRST_VALUE_COUNTER): Replace with
17362         GCOV_COUNTER_V_INTERVAL.
17363         (struct gcov_ctr_summary): Remove.
17364         (struct gcov_summary): Directly use fields of former
17365         gcov_ctr_summary.
17366         (compute_working_sets): Use gcov_summary struct.
17367         * gcov.c (read_count_file): Do not use ctrs fields.
17368         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
17369         struct.
17370         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
17371         struct.
17372         * profile.c: Likewise.
17373         * profile.h: Likewise.
17375 2018-06-05  Martin Liska  <mliska@suse.cz>
17377         PR gcov-profile/84846
17378         * gcov.c (output_lines): Print working directory only
17379         in intermediate format.
17381 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
17383         * config/s390/s390-builtin-types.def: Add void function type.
17384         * config/s390/s390-builtins.def: Use the function type for the
17385         tbeginc builtin.
17387 2018-06-04  Jim Wilson  <jimw@sifive.com>
17389         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
17390         to int.
17391         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
17392         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
17393         handle EH_RETURN_DATA_REGNO registers properly.
17394         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
17395         (riscv_expand_epilogue): Update comment.  Change argument name and
17396         type.  Update code to use new name and type.  Pass new args to
17397         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
17398         EXCEPTION_RETURN.
17399         * config/riscv/riscv.md (NORMAL_RETURN): New.
17400         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
17401         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
17402         (eh_return): Call gen_eh_return_internal and emit barrier.
17403         (eh_return_internal): Call riscv_expand_epilogue.
17405 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
17407         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
17408         bit_insertion field and declare can_be_merged_into method.
17409         (merged_store_group::can_be_merged_into): New method.
17410         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
17411         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
17412         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
17414 2018-06-04  Richard Biener  <rguenther@suse.de>
17416         PR tree-optimization/85955
17417         * builtins.c (fold_builtin_sincos): Convert pointers to
17418         destination to appropriate type before dereferencing.
17420 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
17422         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
17424 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
17426         * expr.c (expand_expr_real_1): Force the operand into memory if
17427         its TYPE_MODE is BLKmode and if there is no integer mode for
17428         the number of bits being extracted.
17430 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17432         PR target/85832
17433         PR target/86036
17434         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
17435         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
17437 2018-06-04  Richard Biener  <rguenther@suse.de>
17439         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
17440         (cleanup_tree_cfg_noloop): ... single caller.  Do
17441         start_recording_case_labels later.
17443 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
17445         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
17446         to _IMMINTRIN_H_INCLUDED.
17447         * config/i386/pconfigintrin.h: Ditto.
17448         * config/i386/waitpkgintrin.h: Ditto.
17449         * config/i386/immintrin.h: Add includes for sgxintrin.h,
17450         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
17451         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
17452         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
17453         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
17454         waitpkgintrin.h and cldemoteintrin.h.
17456 2018-06-04  Richard Biener  <rguenther@suse.de>
17458         PR tree-optimization/86038
17459         * tracer.c (find_best_successor): Check probability for
17460         being initialized, bail out if not.
17462 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
17464         PR target/86003
17465         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
17466         of bits to ignore when comparing architectures.
17468 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
17470         PR tree-optimization/69615
17471         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
17472         maximum or minimum of the type, try to merge it also as if
17473         range1 is + [-, x - 1] or + [x + 1, -].
17475         PR c++/86025
17476         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
17478 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
17480         PR tree-optimization/86034
17481         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
17482         the unsigned bitfield type in a bit insertion sequence if it does not
17483         have a larger precision than the bitfield size.
17484         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
17486 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
17488         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
17490 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17492         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
17493         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
17494         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
17495         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
17497 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
17499         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
17500         Disable -fdelete-null-pointer-checks for ELF toolchain.
17502 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17503             Kito Cheng  <kito.cheng@gmail.com>
17505         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
17506         (nds32le-*-*, nds32be-*-*): Integrate checking process.
17507         (nds32*-*-*): Add glibc and uclibc conditions.
17508         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
17509         (TARGET_EXCEPT_UNWIND_INFO): Define.
17510         * config/nds32/elf.h: New file.
17511         * config/nds32/linux.h: New file.
17512         * config/nds32/nds32-elf.opt: New file.
17513         * config/nds32/nds32-linux.opt: New file.
17514         * config/nds32/nds32-fp-as-gp.c
17515         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
17516         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
17517         TARGET_LINUX_ABI.
17518         (nds32_asm_file_end): Ditto.
17519         (nds32_print_operand): Ditto.
17520         (nds32_insert_attributes): Ditto.
17521         (nds32_init_libfuncs): New function.
17522         (TARGET_HAVE_TLS): Define.
17523         (TARGET_INIT_LIBFUNCS): Define.
17524         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
17525         spec content.
17526         (TARGET_ELF): Apply different mcmodel setting.
17527         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
17528         been migrated into elf.h and linux.h files.
17529         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
17530         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
17531         (mcmodel): The content has been migrated into nds32-elf.opt and
17532         nds32-linux.opt files.
17533         * config/nds32/t-elf: New file.
17534         * config/nds32/t-linux: New file.
17536 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17537             Shiva Chen  <shiva0217@gmail.com>
17539         * config/nds32/constants.md (unspec_volatile_element): Add
17540         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
17541         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
17542         optimization.
17543         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
17544         (make_pass_nds32_fp_as_gp): Declare.
17545         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
17546         optmization pass.
17547         (nds32_asm_function_end_prologue): Remove unused asm output.
17548         (nds32_asm_function_begin_epilogue): Remove unused asm output.
17549         (nds32_asm_file_start): Output necessary fp_as_gp information.
17550         (nds32_option_override): Adjust register usage.
17551         (nds32_expand_prologue): Consider fp_as_gp situation.
17552         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
17553         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
17554         (epilogue): Ditto.
17555         (return): Ditto.
17556         (simple_return): Ditto.
17557         (omit_fp_begin): Output special directive for fp_as_gp.
17558         (omit_fp_end): Output special directive for fp_as_gp.
17559         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
17560         mforbid-fp-as-gp): New options.
17562 2018-06-01  Mark Wielaard  <mark@klomp.org>
17564         * dwarf2out.c (dwarf2out_finish): Remove generation of
17565         DW_AT_loclists_base.
17567 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
17569         * gimple-ssa-store-merging.c: Include gimple-fold.h.
17570         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
17571         (struct merged_store_group): Add bit_insertion field.
17572         (dump_char_array): Use standard hexadecimal format.
17573         (merged_store_group::merged_store_group): Set bit_insertion to false.
17574         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
17575         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
17576         also print the mask in the dump file.
17577         (pass_store_merging::gate): Minor tweak.
17578         (imm_store_chain_info::coalesce_immediate): Fix wrong association
17579         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
17580         stores with INTEGER_CST stores.
17581         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
17582         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
17583         and use it throughout.  Generate bit insertion sequences if need be.
17584         (pass_store_merging::process_store): Remove redundant condition.
17585         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
17587 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
17589         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
17590         the 128-bit floating point types.  Fix function comment.
17592 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17594         * config/aarch64/aarch64-simd.md
17595         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
17596         mnemonics.
17597         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
17598         mnemonics.
17600 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17602         PR tree-optimization/85989
17603         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
17604         variable.
17605         (backprop::intersect_uses): Check it when deciding whether this
17606         is a backedge reference.
17607         (backprop::process_block): Add each phi to m_visited_phis
17608         after visiting it, then clear it at the end.
17610 2018-06-01  Richard Biener  <rguenther@suse.de>
17612         * tree-vectorizer.h (vect_dr_stmt): New function.
17613         (vect_get_load_cost): Adjust.
17614         (vect_get_store_cost): Likewise.
17615         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
17616         Use vect_dr_stmt instead of DR_SMTT.
17617         (vect_record_base_alignments): Likewise.
17618         (vect_calculate_target_alignment): Likewise.
17619         (vect_compute_data_ref_alignment): Likewise and make static.
17620         (vect_update_misalignment_for_peel): Likewise.
17621         (vect_verify_datarefs_alignment): Likewise.
17622         (vector_alignment_reachable_p): Likewise.
17623         (vect_get_data_access_cost): Likewise.  Pass down
17624         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
17625         (vect_get_peeling_costs_all_drs): Likewise.
17626         (vect_peeling_hash_get_lowest_cost): Likewise.
17627         (vect_enhance_data_refs_alignment): Likewise.
17628         (vect_find_same_alignment_drs): Likewise.
17629         (vect_analyze_data_refs_alignment): Likewise.
17630         (vect_analyze_group_access_1): Likewise.
17631         (vect_analyze_group_access): Likewise.
17632         (vect_analyze_data_ref_access): Likewise.
17633         (vect_analyze_data_ref_accesses): Likewise.
17634         (vect_vfa_segment_size): Likewise.
17635         (vect_small_gap_p): Likewise.
17636         (vectorizable_with_step_bound_p): Likewise.
17637         (vect_prune_runtime_alias_test_list): Likewise.
17638         (vect_analyze_data_refs): Likewise.
17639         (vect_supportable_dr_alignment): Likewise.
17640         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
17641         (vect_gen_prolog_loop_niters): Likewise.
17642         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
17643         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
17644         modify DR_STMT.
17645         (vect_recog_mask_conversion_pattern): Likewise.
17646         (vect_try_gather_scatter_pattern): Likewise.
17647         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
17648         to vect_get_store_cost.
17649         (vect_get_store_cost): Get stmt_info instead of DR.
17650         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
17651         (vect_get_load_cost): Get stmt_info instead of DR.
17653 2018-06-01  Richard Biener  <rguenther@suse.de>
17655         PR middle-end/86017
17656         * gimple-fold.c (var_decl_component_p): Also allow offsetted
17657         vars wrapped in MEM_REFs.
17659 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
17661         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17662         Fix subreg tests so that we only return a choice between
17663         GENERAL_REGS and FP_REGS if the original classes included both.
17665 2018-06-01  Richard Biener  <rguenther@suse.de>
17667         PR ipa/85960
17668         * tree-ssa-structalias.c (get_function_part_constraint):
17669         Handle NULL fi->decl.
17670         (find_func_aliases_for_call): Properly handle indirect
17671         fi from direct call.
17672         (find_func_clobbers): Likewise.
17673         (ipa_pta_execute): Likewise.
17674         (create_variable_info_for): For functions that are ifunc_resolver
17675         resolve to a varinfo that contains the result of the resolver call.
17676         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
17677         aliases.
17679 2018-05-31  Michael Collison  <michael.collison@arm.com>
17681         * config/aarch64/aarch64.md:
17682         (*fix_to_zero_extenddfdi2): New pattern.
17683         * gcc.target/aarch64/fix_extend1.c: New testcase.
17685 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
17687         PR middle-end/78809
17688         PR middle-end/83026
17689         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
17690         and BUILT_IN_STRNCMP_EQ.
17691         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
17692         BUILT_IN_STRNCMP_EQ.
17693         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
17694         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17695         (gimple_fold_builtin): Likewise.
17696         * tree-ssa-strlen.c (compute_string_length): New function.
17697         (determine_min_obsize): New function.
17698         (handle_builtin_string_cmp): New function to handle calls to
17699         string compare functions.
17700         (strlen_optimize_stmt): Add handling to builtin string compare
17701         calls.
17702         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
17703         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
17704         * tree.c (build_common_builtin_nodes): Add new defines of
17705         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
17707 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
17709         PR target/85984
17710         * bb-reorder.c (pass_partition_blocks::gate): Return false for
17711         functions with naked attribute.
17713 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
17715         * config/i386/sse.md (avx_vec_concat<mode>):
17716         Substitute concat_tg_mode mode attribute with xtg_mode.
17717         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
17718         (concat_tg_mode): Remove mode attribute.
17720 2018-05-31  Martin Sebor  <msebor@redhat.com>
17722         PR c/82063
17723         * calls.c (alloc_max_size): Correct a logic error/typo.
17724         Treat excessive arguments as infinite.  Warn for invalid arguments.
17725         * doc/invoke.texi (-Walloc-size-larger-than): Update.
17727 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
17729         PR target/85829
17730         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
17731         and movx for Haswell.
17733 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
17734             Cesar Philippidis  <cesar@codesourcery.com>
17736         PR middle-end/85879
17737         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
17738         when emitting error on private/firstprivate reductions.
17739         * omp-low.c (lower_omp_target): Avoid reference-type processing
17740         on pointers for firstprivate clause.
17742 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17744         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
17745         (st1x2): Likewise.
17746         (st1x3): Likewise.
17747         * config/aarch64/aarch64-simd.md
17748         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
17749         (aarch64_ld1_x3_<mode>): Likewise
17750         (aarch64_st1x2<VALLDIF:mode>): Likewise
17751         (aarch64_st1_x2_<mode>): Likewise
17752         (aarch64_st1x3<VALLDIF:mode>): Likewise
17753         (aarch64_st1_x3_<mode>): Likewise
17754         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
17755         (vld1_s8_x3): Likewise.
17756         (vld1_u16_x3): Likewise.
17757         (vld1_s16_x3): Likewise.
17758         (vld1_u32_x3): Likewise.
17759         (vld1_s32_x3): Likewise.
17760         (vld1_u64_x3): Likewise.
17761         (vld1_s64_x3): Likewise.
17762         (vld1_f16_x3): Likewise.
17763         (vld1_f32_x3): Likewise.
17764         (vld1_f64_x3): Likewise.
17765         (vld1_p8_x3): Likewise.
17766         (vld1_p16_x3): Likewise.
17767         (vld1_p64_x3): Likewise.
17768         (vld1q_u8_x3): Likewise.
17769         (vld1q_s8_x3): Likewise.
17770         (vld1q_u16_x3): Likewise.
17771         (vld1q_s16_x3): Likewise.
17772         (vld1q_u32_x3): Likewise.
17773         (vld1q_s32_x3): Likewise.
17774         (vld1q_u64_x3): Likewise.
17775         (vld1q_s64_x3): Likewise.
17776         (vld1q_f16_x3): Likewise.
17777         (vld1q_f32_x3): Likewise.
17778         (vld1q_f64_x3): Likewise.
17779         (vld1q_p8_x3): Likewise.
17780         (vld1q_p16_x3): Likewise.
17781         (vld1q_p64_x3): Likewise.
17782         (vst1_s64_x2): Likewise.
17783         (vst1_u64_x2): Likewise.
17784         (vst1_f64_x2): Likewise.
17785         (vst1_s8_x2): Likewise.
17786         (vst1_p8_x2): Likewise.
17787         (vst1_s16_x2): Likewise.
17788         (vst1_p16_x2): Likewise.
17789         (vst1_s32_x2): Likewise.
17790         (vst1_u8_x2): Likewise.
17791         (vst1_u16_x2): Likewise.
17792         (vst1_u32_x2): Likewise.
17793         (vst1_f16_x2): Likewise.
17794         (vst1_f32_x2): Likewise.
17795         (vst1_p64_x2): Likewise.
17796         (vst1q_s8_x2): Likewise.
17797         (vst1q_p8_x2): Likewise.
17798         (vst1q_s16_x2): Likewise.
17799         (vst1q_p16_x2): Likewise.
17800         (vst1q_s32_x2): Likewise.
17801         (vst1q_s64_x2): Likewise.
17802         (vst1q_u8_x2): Likewise.
17803         (vst1q_u16_x2): Likewise.
17804         (vst1q_u32_x2): Likewise.
17805         (vst1q_u64_x2): Likewise.
17806         (vst1q_f16_x2): Likewise.
17807         (vst1q_f32_x2): Likewise.
17808         (vst1q_f64_x2): Likewise.
17809         (vst1q_p64_x2): Likewise.
17810         (vst1_s64_x3): Likewise.
17811         (vst1_u64_x3): Likewise.
17812         (vst1_f64_x3): Likewise.
17813         (vst1_s8_x3): Likewise.
17814         (vst1_p8_x3): Likewise.
17815         (vst1_s16_x3): Likewise.
17816         (vst1_p16_x3): Likewise.
17817         (vst1_s32_x3): Likewise.
17818         (vst1_u8_x3): Likewise.
17819         (vst1_u16_x3): Likewise.
17820         (vst1_u32_x3): Likewise.
17821         (vst1_f16_x3): Likewise.
17822         (vst1_f32_x3): Likewise.
17823         (vst1_p64_x3): Likewise.
17824         (vst1q_s8_x3): Likewise.
17825         (vst1q_p8_x3): Likewise.
17826         (vst1q_s16_x3): Likewise.
17827         (vst1q_p16_x3): Likewise.
17828         (vst1q_s32_x3): Likewise.
17829         (vst1q_s64_x3): Likewise.
17830         (vst1q_u8_x3): Likewise.
17831         (vst1q_u16_x3): Likewise.
17832         (vst1q_u32_x3): Likewise.
17833         (vst1q_u64_x3): Likewise.
17834         (vst1q_f16_x3): Likewise.
17835         (vst1q_f32_x3): Likewise.
17836         (vst1q_f64_x3): Likewise.
17837         (vst1q_p64_x3): Likewise.
17839 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
17841         * config/msp430/msp430.c (msp430_output_labelref): Prepend
17842         user_label_prefix to name.
17844         * tree-core.h: Update comment about the format of NAME string
17845         passed to handler in attribute_spec.
17847         * config/msp430/msp430.md: Remove erroneous subreg expression from
17848         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
17849         zero_extend{q,h}isi2.
17851 2018-05-30  Borislav Petkov  <bp@suse.de>
17853         * doc/extend.texi: Document some architecture specific
17854         constraints and sort entries.
17856 2018-05-30  Martin Sebor  <msebor@redhat.com>
17858         PR middle-end/85369
17859         * builtins.c (expand_builtin_stpcpy_1): New function.
17860         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
17861         only if the former succeeds.
17863 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
17865         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
17866         in saphira.
17868 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17870         * doc/invoke.texi (-flinker-output): Document
17872 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17874         * passes.c (ipa_write_summaries): Only modify statements if body
17875         is in memory.
17876         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
17877         incrementally linking.
17878         (ipa_passes): Likewise.
17879         * lto-cgraph.c (lto_output_node): When incrementally linking do not
17880         pass down resolution info.
17881         * common.opt (flag_incremental_link): Update info.
17882         * gcc.c (plugin specs): Turn flinker-output=* to
17883         -plugin-opt=-linker-output-known
17884         * toplev.c (compile_file): Also cut compilation when doing incremental
17885         link.
17886         * flag-types. (enum lto_partition_model): Add
17887         LTO_LINKER_OUTPUT_NOLTOREL.
17888         (invoke.texi): Add -flinker-output docs.
17889         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
17890         link same way as WPA; do not stream in dead initializers.
17892         * dwarf2out.c (dwarf2out_die_ref_for_decl,
17893         darf2out_register_external_decl): Support incremental link.
17895 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17897         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
17899 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
17901         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
17902         it down to simple_object_copy_lto_debug_sections.
17903         (run_gcc): Determine incremental LTO link time and configure
17904         lto1 into non-wpa mode, disable renaming of debug sections.
17906 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17908         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
17909         descriptions of various incorrectly documented functions.
17911 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
17913         Revert:
17914         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
17915         address check not strict.
17917 2018-05-30  Richard Biener  <rguenther@suse.de>
17919         PR tree-optimization/85964
17920         * tracer.c (better_p): Drop initialized count check, we only
17921         call the function with initialized counts now.
17922         (find_best_successor): Do find a best edge if one
17923         has uninitialized count.
17924         (find_best_predecessor): Likewise.  Do BB frequency check only
17925         if count is initialized.
17927 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
17929         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
17930         (aarch64_ldrstr_offset_compare): New.
17931         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
17932         load/store orderings.
17933         (aarch64_gen_adjusted_ldpstp): Likewise.
17935 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
17937         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
17938         Check for subset of GENERAL_REGS and FP_REGS.
17939         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
17940         r=w alternative.
17942 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
17944         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
17945         and wi::to_poly_offset.  Add the current offset and then check
17946         whether the sum fits, rather than using an unchecked addition of
17947         a checked term.  Check for a shwi rather than a uhwi.
17948         * expr.c (get_bit_range): Use tree_to_poly_uint64.
17949         (store_constructor): Use poly_int_tree_p.
17950         (expand_expr_real_1): Likewise.
17951         * function.c (assign_temp): Likewise.
17952         * fold-const.c (const_binop): Use poly_int_tree_p and
17953         wi::to_poly_offset.
17954         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
17955         division.
17956         * ipa-icf-gimple.c (func_checker::compare_operand): Use
17957         to_poly_offset for MEM offsets.
17958         * ipa-icf.c (sem_variable::equals): Likewise.
17959         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
17960         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
17961         wi::to_poly_offset for BIT_FIELD_REF offsets.
17962         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
17963         wi::to_poly_offset.
17964         * var-tracking.c (emit_note_insn_var_location): Use
17965         tree_to_poly_uint64.
17967 2018-05-29  Jim Wilson  <jimw@sifive.com>
17969         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
17971 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
17973         PR target/85950
17974         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
17975         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
17976         sequence.
17977         (sse4_1_round<mode>2): Use nonimmediate_operand
17978         for operand 1 predicate.
17980 2018-05-29  Martin Sebor  <msebor@redhat.com>
17981             Richard Biener  <rguenther@suse.de>
17983         PR testsuite/85888
17984         * calls.c (get_size_range): Call determine_value_range instead
17985         of get_value_range..
17986         * tree-vrp.h (determine_value_range): Declared new function.
17987         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
17989 2018-05-29  Richard Biener  <rguenther@suse.de>
17991         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
17992         sure to use non-pattern stmts for get_earlier_stmt arguments.
17993         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
17994         called on pattern stmts.
17995         (get_later_stmt): Likewise.
17997 2018-05-29  Martin Liska  <mliska@suse.cz>
17999         PR gcov-profile/85759
18000         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
18001         env variables.
18003 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
18005         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
18006         VEC_UNPACK_*_EXPR.
18007         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
18008         VEC_PACK_*_EXPR.
18010         PR target/85918
18011         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
18012         VEC_PACK_FLOAT_EXPR): New tree codes.
18013         * tree-pretty-print.c (op_code_prio): Handle
18014         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
18015         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
18016         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
18017         * tree-inline.c (estimate_operator_cost): Likewise.
18018         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
18019         * fold-const.c (const_binop): Likewise.
18020         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
18021         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
18022         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
18023         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
18024         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
18025         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
18026         * expr.c (expand_expr_real_2): Likewise.
18027         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
18028         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
18029         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
18030         optabs.
18031         * optabs.c (expand_widen_pattern_expr): For
18032         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
18033         sign from result type rather than operand's type.
18034         (expand_binop_directly): For vec_packu_float_optab and
18035         vec_packs_float_optab allow result type to be different from operand's
18036         type.
18037         * optabs-tree.c (optab_for_tree_code): Handle
18038         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
18039         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
18040         * tree-vect-generic.c (expand_vector_operations_1):  Handle
18041         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
18042         VEC_PACK_FLOAT_EXPR.
18043         * tree-vect-stmts.c (supportable_widening_operation): Handle
18044         FIX_TRUNC_EXPR.
18045         (supportable_narrowing_operation): Handle FLOAT_EXPR.
18046         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
18047         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
18048         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
18049         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
18050         mode attributes.
18051         (vec_pack<floatprefix>_float_<mode>): New expander.
18052         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
18053         attributes.
18054         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
18055         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
18056         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
18057         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
18058         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
18059         Document.
18060         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
18061         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
18062         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
18063         VEC_PACK_FLOAT_EXPR): Document.
18065 2018-05-29  Richard Biener  <rguenther@suse.de>
18067         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
18068         member.
18069         (stmt_vec_info_vec): Make pointer.
18070         (init_stmt_vec_info_vec): Remove.
18071         (free_stmt_vec_info_vec): Likewise.
18072         (set_stmt_vec_info_vec): New function.
18073         (free_stmt_vec_infos): Likewise.
18074         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
18075         (set_vinfo_for_stmt): Likewise.
18076         (get_earlier_stmt): Likewise.
18077         (get_later_stmt): Likewise.
18078         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
18079         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
18080         (vec_info::~vec_info): Free stmt_vec_infos.
18081         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
18082         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
18083         (pass_slp_vectorize::execute): Likewise.
18084         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
18085         (free_stmt_vec_info_vec): Likewise.
18086         (set_stmt_vec_info_vec): New function.
18087         (free_stmt_vec_infos): Likewise.
18088         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
18089         the global stmt_vec_info_vec.
18090         * tree-parloops.c (gather_scalar_reductions): Use
18091         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
18092         vector.
18094 2018-05-29  Richard Biener  <rguenther@suse.de>
18096         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
18098 2018-05-29  Martin Liska  <mliska@suse.cz>
18099             David Malcolm  <dmalcolm@redhat.com>
18101         * vec.c (test_reverse): New.
18102         (vec_c_tests): Add new test.
18103         * vec.h (vl_ptr>::reverse): New function.
18105 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
18107         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
18109         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
18110         and later.
18112 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
18114         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
18116 2018-05-28  Richard Biener  <rguenther@suse.de>
18118         PR tree-optimization/85933
18119         * tree-vect-data-refs.c (vect_record_base_alignments): Only
18120         look at stmts marked as vectorizable.
18122 2018-05-28  Richard Biener  <rguenther@suse.de>
18124         PR tree-optimization/85934
18125         * tree-vect-generic.c (expand_vector_operations_1): Hoist
18126         vector boolean check before scalar optimization.
18128 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
18130         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
18131         for armv5te.
18133 2018-05-28  Mark Wielaard  <mark@klomp.org>
18135         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
18136         if it is an expression containing a minus sign.
18138 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
18140         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
18142 2018-05-27  Paul Koning  <ni1d@arrl.net>
18144         * config/pdp11/pdp11.md (truncsihi2): Remove.
18146 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
18147             Chung-Ju Wu  <jasonwucj@gmail.com>
18149         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
18150         implementation.
18151         (unaligned_store_dw): Ditto.
18152         * config/nds32/nds32-memory-manipulation.c
18153         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
18154         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
18155         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
18156         (emit_setmem_word_loop): Rename to ...
18157         (emit_setmem_doubleword_loop): ... this.
18158         (nds32_gen_dup_4_byte_to_word_value): New function.
18159         (nds32_gen_dup_8_byte_to_double_word_value): New function.
18160         (nds32_expand_setmem_loop): Refine implementation.
18161         (nds32_expand_setmem_loop_v3m): Ditto.
18162         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
18163         pattern.
18165 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18167         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
18169 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
18171         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
18172         (nds32_init_machine_status): Initialize machine->attr_naked_p and
18173         machine->attr_no_prologue_p.
18174         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
18175         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
18176         (nds32_expand_epilogue): Consider attr_naked_p.
18177         (nds32_expand_epilogue_v3pop): Likewise.
18178         (nds32_can_use_return_insn): Likewise.
18179         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
18180         attr_no_prologue_p fields.
18181         * config/nds32/nds32.opt (mret-in-naked-func): New option.
18183 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
18185         PR target/85918
18186         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
18187         attributes.
18188         * config/i386/sse.md
18189         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
18190         Rename to ...
18191         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
18192         ... this.
18193         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
18194         Rename to ...
18195         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
18196         ... this.
18197         (*<floatsuffix>floatv2div2sf2): Rename to ...
18198         (*float<floatunssuffix>v2div2sf2): ... this.
18199         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
18200         (float<floatunssuffix>v2div2sf2_mask): ... this.
18201         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
18202         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
18203         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
18204         to ...
18205         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
18206         ... this.
18207         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18208         Rename to ...
18209         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
18210         ... this.
18211         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18212         Rename to ...
18213         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
18214         ... this.
18215         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
18216         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
18217         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
18218         gen_ufix_truncv8dfv8si2.
18219         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
18220         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
18221         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
18222         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
18223         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
18224         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
18225         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
18226         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
18228 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18230         PR target/85900
18231         PR target/85345
18232         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
18234 2018-05-25  Jim Wilson  <jimw@sifive.com>
18236         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
18237         * config/riscv/riscv.c (struct machine_function): Add
18238         interrupt_handler_p and attribute_checked_p fields.
18239         (riscv_attribute_table): Add interrupt.
18240         (riscv_interrupt_type_p): New.
18241         (riscv_save_reg_p): Save extra regs for interrupt handler.
18242         (riscv_use_save_libcall): Return false  for interrupt handler.
18243         (riscv_first_stack_step): Add forward declaration.
18244         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
18245         for interrupt handler with large frame.  Use it for saved reg list.
18246         (riscv_expand_prologue): Move flag_stack_usage_info support to
18247         eliminate duplication.
18248         (riscv_expand_epilogue): Generate mret for interrupt handler.
18249         (riscv_epilogue_uses): New.
18250         (riscv_can_use_return_insn): Return false for interrupt handler.
18251         (riscv_function_ok_for_sibcall): Likewise.
18252         (riscv_set_current_function): Add interrupt handler support.
18253         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
18254         * config/riscv/riscv.md (UNSPECV_MRET): New.
18255         (GP_REGNUM): New.
18256         (riscv_frflags, riscv_fsflags): Use tab after opcode.
18257         (riscv_mret): New.
18258         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
18260 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
18262         PR tree-optimization/85712
18263         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
18264         this candidate has already been replaced in-situ by a copy.
18266 2018-05-25  Jason Merrill  <jason@redhat.com>
18268         PR c++/80485 - inline function non-zero address.
18269         * symtab.c (nonzero_address): Check DECL_COMDAT.
18271 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
18273         PR target/83628
18274         * config/alpha/alpha.md (ashlsi3): New insn pattern.
18275         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
18276         extension of SImode operation.  Use const123_operand predicate.
18277         (*saddsi_1): Remove.
18278         (*saddl_se_1): Ditto.
18279         (*ssubsi_1): Ditto.
18280         (*ssubl_se_1): Ditto.
18281         * config/alpha/predicates.md (const123_operand): New predicate.
18282         * config/alpha/constraints.md (P): Use IN_RANGE.
18284 2018-05-25  Richard Biener  <rguenther@suse.de>
18286         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
18287         defaulted to true.
18288         (ref_maybe_used_by_stmt_p): Likewise.
18289         (stmt_may_clobber_ref_p): Likewise.
18290         (stmt_may_clobber_ref_p_1): Likewise.
18291         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
18292         and pass it along.
18293         (ref_maybe_used_by_stmt_p): Likewise.
18294         (stmt_may_clobber_ref_p): Likewise.
18295         (stmt_may_clobber_ref_p_1): Likewise.
18296         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
18297         the alias oracle to disambiguate DRs with stmts DR analysis
18298         couldn't handle.
18299         (vect_analyze_data_refs): Do not give up on not analyzable
18300         DRs for BB vectorization.  Remove code truncating the dataref
18301         vector.
18303 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
18305         PR target/85832
18306         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
18307         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
18308         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
18310 2018-05-25  Richard Biener  <rguenther@suse.de>
18312         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
18313         function, combining stmt data ref gathering and fatal analysis
18314         parts.
18315         (vect_analyze_data_refs): Remove now redudnant code and simplify.
18316         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
18317         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
18318         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
18319         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
18321 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18323         PR tree-optimization/85720
18324         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
18325         SCC if all partitions are builtins.
18326         (version_loop_by_alias_check): New parameter.  Generate cancelable
18327         runtime alias check if all partitions are builtins.
18328         (distribute_loop): Update call to above function.
18330 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18332         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
18333         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
18334         (parm_default_def_partition_arg): Ditto.
18335         (set_parm_default_def_partition): Ditto.
18336         (get_parm_default_def_partitions): Ditto and make it static.
18337         (get_undefined_value_partitions): Ditto and make it static.
18338         (remove_ssa_form): Refactor call to init_var_map here.
18339         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
18340         computation for loop region.
18341         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
18342         (register_default_def): Delete.
18343         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
18344         (parm_default_def_partition_arg): Ditto.
18345         (set_parm_default_def_partition): Ditto.
18346         (get_parm_default_def_partitions): Ditto and make it static.
18347         (get_undefined_value_partitions): Ditto and make it static.
18348         (coalesce_with_default, coalesce_with_default): Update comment.
18349         (create_coalesce_list_for_region): New func factored out from
18350         create_outofssa_var_map.
18351         (populate_coalesce_list_for_outofssa): New func factored out from
18352         create_outofssa_var_map and coalesce_ssa_name.
18353         (create_outofssa_var_map): Delete.
18354         (coalesce_ssa_name): Refactor to support live range computation.
18355         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
18356         (get_parm_default_def_partitions): Delete.
18357         (get_undefined_value_partitions): Ditto.
18358         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
18359         computation for loop region.
18360         (new_tree_live_info, loe_visit_block): Ditto.
18361         (live_worklist, set_var_live_on_entry): Ditto.
18362         (calculate_live_on_exit, verify_live_on_entry): Ditto.
18363         * tree-ssa-live.h (struct _var_map): New fields.
18364         (init_var_map): Change decl.
18365         (region_contains_p): New.
18367 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
18369         * tree-ssa-live.h (live_merge_and_clear): Delete.
18371 2018-05-25  Richard Biener  <rguenther@suse.de>
18373         PR c++/85912
18374         * tree-dump.c (dequeue_and_dump): Remove access to removed
18375         operand 2 of a SWITCH_EXPR.
18377 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18379         * doc/sourcebuild.texi (vect_double_cond_arith): Include
18380         multiplication and division.
18381         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
18382         (cond_udiv@var{m}, cond_umod@var{m}): Document.
18383         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
18384         (cond_udiv_optab, cond_umod_optab): New optabs.
18385         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
18386         (IFN_COND_RDIV): New internal functions.
18387         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
18388         TRUNC_MOD_EXPR and RDIV_EXPR.
18389         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
18390         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
18391         New unspecs.
18392         (SVE_INT_BINARY): Include mult.
18393         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
18394         (optab, sve_int_op): Handle mult.
18395         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
18396         UNSPEC_COND_DIV.
18397         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
18398         for SVE_INT_BINARY_SD.
18400 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18402         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
18403         (optab, sve_int_op): Handle div and udiv.
18404         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
18405         for SVE_INT_BINARY_SD.
18406         (*<optab><mode>3): New insn for the same.
18408 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18410         * tree-vect-patterns.c: Include predict.h.
18411         (vect_recog_divmod_pattern): Restrict check for division support
18412         to when optimizing for size.
18414 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18416         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
18417         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
18418         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
18419         (gimple_match_op::set_op): Likewise.
18420         (gimple_resimplify4): Declare.
18421         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
18422         (expr::gen_transform): Likewise.
18423         (decision_tree::gen): Generate a simplification routine for 4 operands.
18424         * gimple-match-head.c (gimple_simplify): Add an overload for
18425         4 operands.  In the top-level function, handle up to 4 call
18426         arguments and call gimple_resimplify4.
18427         (gimple_resimplify4): New function.
18428         (build_call_internal): Pass a fourth operand.
18429         (maybe_push_to_seq): Likewise.
18430         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
18431         Fold VEC_COND_EXPRs of an operation and a default value into
18432         an IFN_COND_* function if possible.
18433         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
18434         New unspecs.
18435         (SVE_COND_FP_BINARY): Include them.
18436         (optab, sve_fp_op): Handle them.
18437         (SVE_INT_BINARY_REV): New code iterator.
18438         (SVE_COND_FP_BINARY_REV): New int iterator.
18439         (commutative): New int attribute.
18440         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
18441         Declare.
18442         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
18443         function.
18444         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
18445         (*cond_<optab><mode>): New patterns for reversed operands.
18447 2018-05-25  Richard Biener  <rguenther@suse.de>
18449         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
18450         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
18451         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
18452         (STMT_VINFO_GROUPED_ACCESS): Adjust.
18453         * tree-vect-data-refs.c (everywhere): Adjust users.
18454         * tree-vect-loop.c (everywhere): Likewise.
18455         * tree-vect-slp.c (everywhere): Likewise.
18456         * tree-vect-stmts.c (everywhere): Likewise.
18457         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
18459 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
18461         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
18462         Rename to...
18463         (gcc_cv_as_section_exclude): ... this.
18464         Try Solaris as #exclude syntax.
18465         * configure: Regenerate.
18466         * config.in: Regenerate.
18467         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
18468         SECTION_EXCLUDE.
18469         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
18470         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
18472         * varasm.c (default_elf_asm_named_section): Don't check if
18473         HAVE_GAS_SECTION_EXCLUDE is defined.
18475 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18477         * doc/md.texi: Update the documentation of the cond_* optabs
18478         to mention the new final operand.  Fix GET_MODE_NUNITS call.
18479         Describe the scalar case too.
18480         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
18481         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
18482         instead of 2.
18483         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
18484         (get_conditional_internal_fn): Update comment.
18485         * tree-vect-loop.c (vectorizable_reduction): Pass the original
18486         accumulator value as a final argument to conditional functions.
18487         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
18488         a define_expand and add an "else" operand.  Assert for now that
18489         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
18490         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
18491         (*cond_<optab><mode>): New patterns.
18492         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
18493         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
18494         (UNSPEC_COND_EOR): Delete.
18495         (optab): Remove associated mappings.
18496         (SVE_INT_BINARY): New code iterator.
18497         (sve_int_op): Remove int attribute and add "minus" to the code
18498         attribute.
18499         (SVE_COND_INT_OP): Delete.
18500         (SVE_COND_FP_OP): Rename to...
18501         (SVE_COND_FP_BINARY): ...this.
18503 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
18505         * optabs.c (can_reuse_operands_p): New function.
18506         (maybe_legitimize_operands): Try to reuse the results for
18507         earlier operands.
18509 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18511         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
18512         Add {q} suffix to insn mnemonic.
18514 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
18516         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
18517         (msp430_warn_func_return): New.
18519 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
18521         * fold-const.c (tree_nonzero_bits): New function.
18522         * fold-const.h (tree_nonzero_bits): Likewise.
18523         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
18524         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
18526 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
18528         PR target/85900
18529         PR target/85345
18530         * varasm.c (assemble_alias): Check ifunc_resolver only on
18531         FUNCTION_DECL.
18533 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
18535         PR target/85903
18536         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
18537         when memory input operand is handled.
18539 2018-05-24  Luis Machado  <luis.machado@linaro.org>
18541         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
18542         global.
18543         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
18545 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18547         * match.pd: Delay FMA folds until after vectorization.
18549 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
18551         PR target/83009
18552         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
18553         address check not strict.
18555 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
18557         * gimple-match.h (gimple_match_op): New class.
18558         (mprts_hook): Replace parameters with a gimple_match_op *.
18559         (maybe_build_generic_op): Likewise.
18560         (gimple_simplified_result_is_gimple_val): Replace parameters with
18561         a const gimple_match_op *.
18562         (gimple_simplify): Replace code_helper * and tree * parameters with
18563         a gimple_match_op * parameter.
18564         (gimple_resimplify1): Replace code_helper *, tree and tree *
18565         parameters with a gimple_match_op * parameter.
18566         (gimple_resimplify2): Likewise.
18567         (gimple_resimplify3): Likewise.
18568         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
18569         parameters with a gimple_match_op * parameter.
18570         * gimple-match-head.c (gimple_simplify): Change prototypes of
18571         auto-generated functions to take a gimple_match_op * instead of
18572         separate code_helper * and tree * parameters.  Make the same
18573         change in the top-level overload and update calls to the
18574         gimple_resimplify routines.  Update calls to the auto-generated
18575         functions and to maybe_push_res_to_seq in the publicly-facing
18576         operation-specific gimple_simplify overloads.
18577         (gimple_match_op::MAX_NUM_OPS): Define.
18578         (gimple_resimplify1): Replace rcode and ops with a single res_op
18579         parameter.  Update call to gimple_simplify.
18580         (gimple_resimplify2): Likewise.
18581         (gimple_resimplify3): Likewise.
18582         (mprts_hook): Replace parameters with a gimple_match_op *.
18583         (maybe_build_generic_op): Likewise.
18584         (build_call_internal): Replace type, nargs and ops with
18585         a gimple_match_op *.
18586         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
18587         with a single gimple_match_op *.  Update calls to mprts_hook,
18588         build_call_internal and gimple_simplified_result_is_gimple_val.
18589         Factor out code that is common to the tree_code and combined_fn cases.
18590         * genmatch.c (expr::gen_transform): Replace tem_code and
18591         tem_ops with a gimple_match_op called tem_op.  Update calls
18592         to the gimple_resimplify functions and maybe_push_res_to_seq.
18593         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
18594         res_ops.  Update call to the gimple_resimplify functions.
18595         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
18596         (decision_tree::gen): Make the functions take a gimple_match_op *
18597         called res_op instead of separate res_code and res_ops parameters.
18598         Update call accordingly.
18599         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
18600         and ops with a single res_op parameter.  Update calls to
18601         maybe_build_generic_op and maybe_push_res_to_seq.
18602         (fold_stmt_1): Update calls to gimple_simplify and
18603         replace_stmt_with_simplification.
18604         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
18605         and gimple_simplified_result_is_gimple_val.
18606         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
18607         gimple_simplify.
18608         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
18609         with a gimple_match_op *.
18610         (vn_nary_build_or_lookup): Likewise.  Update call to
18611         vn_nary_build_or_lookup_1.
18612         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
18613         gimple_match_op *.  Update calls to the gimple_resimplify routines
18614         and to gimple_simplified_result_is_gimple_val.
18615         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
18616         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
18617         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
18618         (visit_nary_op): Likewise.
18619         (visit_reference_op_load): Likewise.
18621 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18623         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
18624         modifier for printing the step amount.
18626 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
18628         PR target/78849
18629         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
18630         types.
18632 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
18634         * doc/sourcebuild.texi (Endianness): New subsubsection.
18636 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18638         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18639         <prefetch_dynamic_strides>: New const bool field.
18640         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18641         prefetch_dynamic_strides.
18642         (exynosm1_prefetch_tune): Likewise.
18643         (thunderxt88_prefetch_tune): Likewise.
18644         (thunderx_prefetch_tune): Likewise.
18645         (thunderx2t99_prefetch_tune): Likewise.
18646         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
18647         false.
18648         (aarch64_override_options_internal): Update to set
18649         PARAM_PREFETCH_DYNAMIC_STRIDES.
18650         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
18651         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
18652         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
18653         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
18654         prefetch-dynamic-strides setting.
18656 2018-05-23  Luis Machado  <luis.machado@linaro.org>
18658         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
18659         <minimum_stride>: New const int field.
18660         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
18661         minimum_stride field defaulting to -1.
18662         (exynosm1_prefetch_tune): Likewise.
18663         (thunderxt88_prefetch_tune): Likewise.
18664         (thunderx_prefetch_tune): Likewise.
18665         (thunderx2t99_prefetch_tune): Likewise.
18666         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
18667         <default_opt_level>: Set to 3.
18668         (aarch64_override_options_internal): Update to set
18669         PARAM_PREFETCH_MINIMUM_STRIDE.
18670         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
18671         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
18672         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
18673         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
18674         stride is constant and is below the minimum stride threshold.
18676 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18678         * config/arm/arm-cpus.in (mode26): Delete.
18679         (armv4): Delete mode26 reference.
18680         * config/arm/arm.c (arm_configure_build_target): Delete use of
18681         isa_bit_mode26.
18683 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18685         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
18686         New insn pattern.
18687         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
18688         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
18689         for non-SSE modes.
18690         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
18691         (floatunsdidf2): Ditto.
18693 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
18695         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
18696         (fixuns_trunc<mode>si2_avx512f): Ditto.
18697         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
18698         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
18699         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
18701 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18703         PR rtl-optimization/79985
18704         * df-scan.c (df_insn_refs_collect): Remove special case for
18705         global registers and asm statements.
18707 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
18709         * extend.texi (Global Register Variables): Rewrite the bullet list.
18710         Note that the register is available for allocation. Note that access
18711         via inline asm must use constraints. Add note about async-signal
18712         handlers. Remove paragraph about automagic register selection.
18714 2018-05-23  Richard Biener  <rguenther@suse.de>
18716         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
18717         of fixed offset from memset VN.
18719 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
18721         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
18722         first_interp field.
18723         (alloc_cand_and_find_basis): Initialize first_interp field.
18724         (slsr_process_mul): Modify first_interp field.
18725         (slsr_process_add): Likewise.
18726         (slsr_process_cast): Modify first_interp field for each new
18727         interpretation.
18728         (slsr_process_copy): Likewise.
18729         (dump_candidate): Dump first_interp field.
18730         (replace_mult_candidate): Process all interpretations, not just
18731         subsequent ones.
18732         (replace_rhs_if_not_dup): Likewise.
18733         (replace_one_candidate): Likewise.
18735 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
18737         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
18738         Add new boolean.
18739         (aarch64_needs_frame_chain): New function.
18740         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
18742 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
18744         PR target/84882
18745         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
18746         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
18747         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
18748         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
18749         as true for strict-align.
18750         (aarch64_can_inline_p): Perform checks even when callee has no
18751         attributes to check for strict alignment.
18752         * doc/extend.texi (AArch64 Function Attributes): Document
18753         no-strict-align.
18754         * doc/invoke.texi: (AArch64 Options): Likewise.
18756 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
18758         PR tree-optimization/85853
18759         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
18760         the handling of the root of the node to...
18761         (vect_slp_analyze_node_operations_1): ...this new function,
18762         and run the whole thing with the child nodes' def types
18763         set according to their SLP node's def type.
18765 2018-05-23  Richard Biener  <rguenther@suse.de>
18767         PR middle-end/85874
18768         * tree-data-ref.c (create_runtime_alias_checks): Defer
18769         and ignore overflow warnings.
18771 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
18773         PR tree-optimization/85822
18774         * tree-vrp.c (is_masked_range_test): Fix handling of negative
18775         constants.
18777 2018-05-23  Richard Biener  <rguenther@suse.de>
18779         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
18780         memset constants via native_interpret_expr.
18782 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
18784         PR target/85345
18785         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
18786         attribute.
18787         (cgraph_node::create_alias): Likewise.
18788         (cgraph_node::get_availability): Check ifunc_resolver instead
18789         of looking up ifunc attribute.
18790         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
18791         * varasm.c (do_assemble_alias): Likewise.
18792         (assemble_alias): Likewise.
18793         (default_binds_local_p_3): Likewise.
18794         * cgraph.h (cgraph_node): Add ifunc_resolver.
18795         (cgraph_node::only_called_directly_or_aliased_p): Return false
18796         for IFUNC resolver.
18797         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
18798         attribute.
18799         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
18800         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
18801         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
18802         instead of looking up ifunc attribute.
18804 2018-05-22  Luis Machado  <luis.machado@linaro.org>
18806         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
18808 2018-05-22  Martin Sebor  <msebor@redhat.com>
18810         PR middle-end/85359
18811         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
18812         only when expasion succeeds.
18813         (expand_builtin_strcmp): Same.
18814         (expand_builtin_strncmp): Same.
18816 2018-05-22  Martin Sebor  <msebor@redhat.com>
18818         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
18820 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18821             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18823         * config/aarch64/aarch64-ldpstp.md: Replace uses of
18824         aarch64_mem_pair_operand with memory_operand and delete operand swapping
18825         code.
18826         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18827         Add check for legitimate_address.
18828         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
18829         (aarch64_swap_ldrstr_operands): New.
18830         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
18831         Define prototype.
18833 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
18834             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18836         * config/aarch64/aarch64.md: New patterns to generate stp
18837         and ldp.
18838         (store_pair_sw, store_pair_dw): New patterns to generate stp for
18839         single words and double words.
18840         (load_pair_sw, load_pair_dw): Likewise.
18841         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
18842         Delete.
18843         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
18844         Delete.
18845         * config/aarch64/aarch64-ldpstp.md: Modify peephole
18846         for different mode ldpstp and add peephole for merged zero stores.
18847         Likewise for loads.
18848         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
18849         Add size check.
18850         (aarch64_gen_store_pair): Rename calls to match new patterns.
18851         (aarch64_gen_load_pair): Rename calls to match new patterns.
18852         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
18853         (load_pair<DREG:mode><DREG2:mode>): ... This.
18854         (store_pair<mode>): Rename to...
18855         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
18856         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
18857         New mode iterators.
18858         (V_INT_EQUIV): Handle SImode.
18859         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
18860         New predicate.
18862 2018-05-22  Martin Sebor  <msebor@redhat.com>
18864         PR c/85623
18865         * calls.c (maybe_warn_nonstring_arg): Use string length to set
18866         or ajust the presumed bound on an operation to avoid unnecessary
18867         warnings.
18869 2018-05-22  Martin Sebor  <msebor@redhat.com>
18871         PR tree-optimization/85826
18872         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
18873         assuming that a DECL necesarily has a constant size.
18875 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
18877         PR middle-end/85862
18878         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
18880 2018-05-22  Richard Biener  <rguenther@suse.de>
18882         PR tree-optimization/85834
18883         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
18884         non-constant and non-zero memset arguments.
18886 2018-05-22  Martin Liska  <mliska@suse.cz>
18888         PR ipa/85607
18889         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
18891 2018-05-22  Richard Biener  <rguenther@suse.de>
18893         PR tree-optimization/85863
18894         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
18895         comparisons when vectype is specified.
18896         (vectorizable_condition): Do not specify vectype for
18897         vect_is_simple_cond when SLP vectorizing.
18899 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
18901         PR target/85657
18902         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
18903         define __ibm128 as long double.
18904         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
18905         as a distinct type when IEEE 128-bit support is enabled.
18906         (init_float128_ieee): Fix up conversions between IFmode and IEEE
18907         128-bit types to use the correct functions.
18908         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
18909         convert between 128-bit floating point types that have different
18910         modes but the same representation, instead of using gen_lowpart to
18911         makean alias.
18912         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
18913         KFmode.
18914         (IFKF_reg): New attributes to give the register constraints for
18915         IFmode and KFmode.
18916         (extend<mode>tf2_internal): New insns to mark an explicit
18917         conversion between 128-bit floating point types that have a
18918         different mode but share the same representation.
18920 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
18922         PR tree-optimization/85814
18923         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
18924         a null return from get_strinfo when unsharing the next
18925         strinfo in the chain.
18927 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
18929         PR gcc/84923
18930         * varasm.c (weak_finish): Clean up weak_decls.
18932 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18934         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
18935         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
18936         UNSPEC_UADALP values.
18937         * config/aarch64/iterators.md (ABAL): New int iterator.
18938         (ABDL2): Likewise.
18939         (ADALP): Likewise.
18940         (sur): Add mappings for the above.
18941         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
18942         New define_insn.
18943         (aarch64_<sur>abal<mode>_4): Likewise.
18944         (aarch64_<sur>adalp<mode>_3): Likewise.
18945         (<sur>sadv16qi): New define_expand.
18947 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
18949         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
18950         (*movdf_internal): Ditto.
18951         (*rcpsf2_sse): Ditto.
18952         (*rsqrtsf2_sse): Ditto.
18953         (*sqrt<mode>2_sse): Ditto.
18955 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
18957         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
18958         eor3q<mode>4.
18959         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
18960         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
18961         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
18962         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
18963         vbcaxq_s64): New.
18964         * config/aarch64/arm_neon.h: Likewise.
18965         * config/aarch64/iterators.md (VQ_I): New.
18967 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
18969         * config.gcc: Add arc/t-multilib-linux to tmake_file for
18970         arc*-*-linux*.
18971         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
18972         MULTILIB_DIRNAMES
18974 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
18976         * config/nds32/constraints.md (S): New constraint.
18977         * config/nds32/nds32.md (call_internal): Use constraint S.
18978         (call_value_internal): Likewise.
18979         (sibcall_internal): Likewise.
18980         (sibcall_value_internal): Likewise.
18982 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18983             Chung-Ju Wu  <jasonwucj@gmail.com>
18985         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
18986         into consideration.
18988 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
18989             Chung-Ju Wu  <jasonwucj@gmail.com>
18991         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
18992         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
18993         (nds32_rtx_costs_impl): Simplify.
18994         (nds32_address_cost_impl): Simplify.
18995         (nds32_init_rtx_costs): New function.
18996         (nds32_rtx_costs_speed_prefer): Likewise.
18997         (nds32_rtx_costs_size_prefer): Likewise.
18998         (nds32_address_cost_speed_prefer): Likewise.
18999         (nds32_address_cost_speed_fwprop): Likewise.
19000         (nds32_address_cost_size_prefer): Likewise.
19001         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
19002         * config/nds32/nds32.c (nds32_option_override): Use
19003         nds32_init_rtx_costs function.
19005 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
19007         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
19008         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
19009         (TARGET_PIPELINE_N8): Likewise.
19010         (TARGET_PIPELINE_N10): Likewise.
19011         (TARGET_PIPELINE_N13): Likewise.
19012         (TARGET_PIPELINE_GRAYWOLF): Likewise.
19014 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
19016         * config/nds32/nds32-fpu.md: Update copyright year.
19018 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19020         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
19022 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19024         * config/nds32/nds32.c
19025         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
19026         * config/nds32/nds32.opt (minline-asm-r15): New option.
19028 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19030         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
19031         MASK_HW_ABS.
19032         * config/nds32/nds32.md (abssi2): New pattern.
19034 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
19036         * config/i386/i386.md (rex64namesuffix): New mode attribute.
19037         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
19038         Merge insn pattern from sse_cvtsi2ss<round_name> and
19039         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
19040         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
19041         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
19042         using SWI48 mode iterator.
19043         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
19044         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
19045         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
19046         pattern from sse_cvttss2si<round_saeonly_name>
19047         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
19048         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
19049         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
19050         using SWI48 mode iterator.
19051         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
19052         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
19053         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
19054         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
19055         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
19056         using SWI48 mode iterator.
19057         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
19058         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
19059         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
19060         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
19061         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
19062         SWI48 mode iterator.
19063         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
19064         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
19065         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
19066         pattern from sse_cvttsd2si<round_saeonly_name>
19067         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
19069 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19071         * config/nds32/nds32-md-auxiliary.c
19072         (nds32_valid_smw_lwm_base_p): Refine.
19073         (nds32_output_smw_single_word): Refine.
19074         (nds32_output_smw_double_word): New.
19075         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
19077 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19079         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
19080         (nds32_output_stack_pop): Refine.
19081         (nds32_expand_unaligned_load): Refine.
19082         (nds32_expand_unaligned_store): Refine.
19084 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
19085             Chung-Ju Wu  <jasonwucj@gmail.com>
19087         * config/nds32/constants.md: Add TP_REGNUM constant.
19088         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
19089         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
19090         UNSPEC_ADD32.
19091         * config/nds32/nds32-doubleword.md: Consider flag_pic.
19092         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
19093         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
19094         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
19095         and PIC code generation.
19096         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
19097         code generation.
19098         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
19099         optimization.
19100         * config/nds32/nds32.md: Support TLS and PIC.
19101         * config/nds32/nds32.c: Support TLS and PIC.
19102         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
19103         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
19104         predicate.
19106 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
19108         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
19109         mode with E_ prefix.
19111 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
19112             Chung-Ju Wu  <jasonwucj@gmail.com>
19114         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
19115         * config/nds32/nds32-md-auxiliary.c
19116         (symbolic_reference_mentioned_p): New.
19117         (nds32_legitimize_ict_address): New.
19118         (nds32_expand_ict_move): New.
19119         (nds32_indirect_call_referenced_p): New.
19120         (nds32_symbol_binds_local_p): Delete.
19121         (nds32_long_call_p): Modify.
19122         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
19123         * config/nds32/nds32-protos.h
19124         (symbolic_reference_mentioned_p): Declare.
19125         (nds32_legitimize_ict_address): Declare.
19126         (nds32_expand_ict_move): Declare.
19127         (nds32_indirect_call_referenced_p): Declare.
19128         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
19129         (nds32_relax_group): Use nds32_ict_const_p as condition.
19130         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
19131         (nds32_asm_file_start): Output ict_model directive in asm code.
19132         (nds32_legitimate_address_p): Consider indirect call.
19133         (nds32_print_operand): Consider indirect call.
19134         (nds32_print_operand_address): Consider indirect call.
19135         (nds32_insert_attributes): Handle "indirect_call" attribute.
19136         (TARGET_LEGITIMATE_ADDRESS_P): Define.
19137         (TARGET_LEGITIMATE_CONSTANT_P): Define.
19138         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
19139         (TARGET_DELEGITIMIZE_ADDRESS): Define.
19140         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
19141         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
19142         (TARGET_ICT_MODEL_SMALL): Define.
19143         (TARGET_ICT_MODEL_LARGE): Define.
19144         * config/nds32/nds32.md (movsi): Consider ict model.
19145         (call, call_value): Consider ict model.
19146         (sibcall, sibcall_value): Consider ict model.
19147         * config/nds32/nds32.opt (mict-model): New option.
19148         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
19149         model.
19151 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
19152             Monk Chiang  <sh.chiang04@gmail.com>
19153             Jim Wilson <jimw@sifive.com>
19155         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
19156         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
19157         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
19158         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
19159         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
19160         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
19161         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
19162         compute save_libcall_adjustment properly.
19163         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
19164         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
19165         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
19166         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
19167         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
19168         (ABI_SPEC): Handle mabi=ilp32e.
19169         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
19170         (RVE): Add RVE mask.
19171         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
19172         <-march>: Add rv32e as an example.
19174 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
19176         PR c++/82899
19177         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
19178         (intra_create_variable_infos): Handle C++ constructors.
19180 2018-05-18  Martin Liska  <mliska@suse.cz>
19182         * passes.def: Remove a redundant pass.
19184 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
19186         PR bootstrap/85838
19187         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
19189 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19191         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
19192         (ARMv4): Update.
19193         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
19194         (ARMv6m): Update.
19195         (armv2, armv2a, armv3, armv3m): Delete architectures.
19196         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
19197         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
19198         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
19199         Delete cpus.
19200         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
19201         (*mulsidi3adddi): Likewise.
19202         (mulsidi3): Likewise.
19203         (*mulsidi3_nov6): Likewise.
19204         (umulsidi3): Likewise.
19205         (umulsidi3_nov6): Likewise.
19206         (umaddsidi4): Likewise.
19207         (*umulsidi3adddi): Likewise.
19208         (smulsi3_highpart): Likewise.
19209         (*smulsi3_highpart_nov6): Likewise.
19210         (umulsi3_highpart): Likewise.
19211         (*umulsi3_highpart_nov6): Likewise.
19212         * config/arm/arm.h (arm_arch3m): Delete.
19213         * config/arm/arm.c (arm_arch3m): Delete.
19214         (arm_option_override_internal): Update armv3-related comment.
19215         (arm_configure_build_target): Delete use of isa_bit_mode32.
19216         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
19217         (arm_rtx_costs_internal): Delete check of arm_arch3m.
19218         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
19219         (mulsa3): Likewise.
19220         (mulusa3): Likewise.
19221         * config/arm/arm-protos.h (arm_arch3m): Delete.
19222         * config/arm/arm-tables.opt: Regenerate.
19223         * config/arm/arm-tune.md: Likewise.
19224         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
19225         deleted architectures.
19227 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19229         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
19230         (armv5t, armv5te): New features.
19231         (ARMv5, ARMv5e): Delete fgroups.
19232         (ARMv5t, ARMv5te): Adjust for above changes.
19233         (ARMv6m): Likewise.
19234         (armv5, armv5e): Delete arches.
19235         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
19236         arm_arch5.
19237         (*call_reg_arm): Likewise.
19238         (*call_value_reg_armv5): Likewise.
19239         (*call_value_reg_arm): Likewise.
19240         (*call_symbol): Likewise.
19241         (*call_value_symbol): Likewise.
19242         (*sibcall_insn): Likewise.
19243         (*sibcall_value_insn): Likewise.
19244         (clzsi2): Likewise.
19245         (prefetch): Likewise.
19246         (define_split and define_peephole2 dependent on arm_arch5):
19247         Likewise.
19248         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
19249         arm_arch5e.
19250         (TARGET_ARM_QBIT): Likewise.
19251         (TARGET_DSP_MULTIPLY): Likewise.
19252         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
19253         (arm_arch5, arm_arch5e): Delete.
19254         (arm_arch5t, arm_arch5te): Declare.
19255         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
19256         (arm_arch5t): Declare.
19257         (arm_option_reconfigure_globals): Update for the above.
19258         (arm_options_perform_arch_sanity_checks): Update comment, replace
19259         use of arm_arch5 with arm_arch5t.
19260         (use_return_insn): Likewise.
19261         (arm_emit_call_insn): Likewise.
19262         (output_return_instruction): Likewise.
19263         (arm_final_prescan_insn): Likewise.
19264         (arm_coproc_builtin_available): Likewise.
19265         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
19266         arm_arch5e with arm_arch5t and arm_arch5te.
19267         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
19268         (arm_arch5t, arm_arch5te): Declare.
19269         * config/arm/arm-tables.opt: Regenerate.
19270         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
19271         * config/arm/t-multilib: Likewise.
19272         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
19273         instead of arm_arch5.
19274         (*call_reg_thumb1): Likewise.
19275         (*call_value_reg_thumb1_v5): Likewise.
19276         (*call_value_reg_thumb1): Likewise.
19277         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
19278         unreachable path.
19279         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
19281 2018-05-18  Martin Liska  <mliska@suse.cz>
19283         PR gcov-profile/84846
19284         * doc/gcov.texi: Document -t option of gcov tool.
19286 2018-05-18  Martin Liska  <mliska@suse.cz>
19288         PR gcov-profile/84846
19289         * gcov.c (print_usage): Add new -t option.
19290         (process_args): Handle the option.
19291         (generate_results): Use stdout as output when requested by
19292         the option.
19294 2018-05-18  Martin Liska  <mliska@suse.cz>
19296         PR gcov-profile/84846
19297         * coverage.c (coverage_init): Write PWD to .gcno file.
19298         * doc/gcov.texi: Document how working directory is printed.
19299         * gcov-dump.c (dump_gcov_file): Print PWD.
19300         * gcov.c (output_intermediate_file): Likewise.
19301         (read_graph_file): Read PWD string.
19302         (output_lines): Print PWD.
19304 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19306         PR middle-end/85817
19307         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
19308         for retval and return false if all args to phi are zero.
19310 2018-05-18  Richard Biener  <rguenther@suse.de>
19312         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
19313         method.
19314         (evrp_dom_walker::before_dom_children): Call it.
19316 2018-05-18  Richard Biener  <rguenther@suse.de>
19318         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
19319         results when processing array refs with variable index.
19321 2018-05-18  Toon Moene  <toon@moene.org>
19323         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
19324         directly after that of -floop-interchange. Indicate that both
19325         options are enabled by default when specifying -O3.
19327 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19329         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
19330         iterator.  Delete separate integer-mode vec_set<mode> expander.
19331         (aarch64_simd_vec_setv2di): Delete.
19332         (vec_setv2di): Delete.
19333         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
19334         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
19335         the "w, r" alternative.
19337 2018-05-18  Martin Liska  <mliska@suse.cz>
19339         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
19340         * tree-pass.h (make_pass_lower_switch_O0): New function.
19341         * tree-switch-conversion.c (node_has_low_bound): Remove.
19342         (node_has_high_bound): Likewise.
19343         (node_is_bounded): Likewise.
19344         (class pass_lower_switch): Make it a template type and create
19345         two instances.
19346         (pass_lower_switch::execute): Add template argument.
19347         (make_pass_lower_switch): New function.
19348         (make_pass_lower_switch_O0): New function.
19349         (do_jump_if_equal): Remove.
19350         (emit_case_nodes): Simplify to just handle all 3 cases and leave
19351         all the hard work to tree optimization passes.
19353 2018-05-18  Martin Liska  <mliska@suse.cz>
19355         * dbgcnt.c (limit_low): Renamed from limit.
19356         (limit_high): New variable.
19357         (dbg_cnt_is_enabled): Check for upper limit.
19358         (dbg_cnt): Adjust dumping.
19359         (dbg_cnt_set_limit_by_index): Add new argument for high
19360         value.
19361         (dbg_cnt_set_limit_by_name): Likewise.
19362         (dbg_cnt_process_single_pair): Parse new format.
19363         (dbg_cnt_process_opt): Use strtok.
19364         (dbg_cnt_list_all_counters): Remove 'value' and add
19365         'limit_high'.
19366         * doc/invoke.texi: Document changes.
19368 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
19370         * doc/sourcebuild.texi (scalar_all_fma): Document.
19371         * tree.def (FMA_EXPR): Delete.
19372         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
19373         * internal-fn.c (ternary_direct): New macro.
19374         (expand_ternary_optab_fn): Likewise.
19375         (direct_ternary_optab_supported_p): Likewise.
19376         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
19377         * builtins.c (fold_builtin_fma): Delete.
19378         (fold_builtin_3): Don't call it.
19379         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
19380         * expr.c (expand_expr_real_2): Likewise.
19381         * fold-const.c (operand_equal_p): Likewise.
19382         (fold_ternary_loc): Likewise.
19383         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
19384         * gimple.c (DEFTREECODE): Likewise.
19385         * gimplify.c (gimplify_expr): Likewise.
19386         * optabs-tree.c (optab_for_tree_code): Likewise.
19387         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19388         * tree-eh.c (operation_could_trap_p): Likewise.
19389         (stmt_could_throw_1_p): Likewise.
19390         * tree-inline.c (estimate_operator_cost): Likewise.
19391         * tree-pretty-print.c (dump_generic_node): Likewise.
19392         (op_code_prio): Likewise.
19393         * tree-ssa-loop-im.c (stmt_cost): Likewise.
19394         * tree-ssa-operands.c (get_expr_operands): Likewise.
19395         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
19396         * fold-const-call.h (fold_fma): Delete.
19397         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
19398         CFN_FNMA and CFN_FNMS.
19399         (fold_fma): Delete.
19400         * genmatch.c (combined_fn): New enum.
19401         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
19402         (commutative_op): New function.
19403         (commutate): Use it.  Handle more than 2 operands.
19404         (dt_operand::gen_gimple_expr): Use commutative_op.
19405         (parser::parse_expr): Allow :c to be used with non-binary
19406         operators if the commutative operand is known.
19407         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
19408         CFN_FMS, CFN_FNMA and CFN_FNMS.
19409         (backprop::process_assign_use): Remove FMA_EXPR handling.
19410         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
19411         (gen_hsa_fma): New function.
19412         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
19413         IFN_FNMA and IFN_FNMS.
19414         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
19415         * gimple-fold.h (follow_all_ssa_edges): Declare.
19416         * gimple-fold.c (follow_all_ssa_edges): New function.
19417         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
19418         gimple_build interface and use follow_all_ssa_edges to fold the result.
19419         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
19420         instead of checking for optabs directly.
19421         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
19422         rather than FMA_EXPRs.
19423         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
19424         call to IFN_FMA instead of an FMA_EXPR.
19426 2018-05-17  Jim Wilson  <jimw@sifive.com>
19428         * expr.c (do_tablejump): When converting index to Pmode, if we have a
19429         sign extended promoted subreg, and the range does not have the sign bit
19430         set, then do a sign extend.
19432         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
19433         test, check for sign extended subreg and/or constant operands, and
19434         do a sign extend in that case.
19436 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
19438         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
19439         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
19440         Add untyped.
19441         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
19442         Change logics_shift_reg to logics_shift_imm.
19443         (thunderx2t99_fp_loadpair_basic): Delete.
19444         (thunderx2t99_fp_storepair_basic): Delete.
19445         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
19446         (thunderx2t99_asimd_polynomial): Delete.
19447         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
19448         and neon_fp_mul_d_scalar_q.
19449         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
19450         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
19451         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
19452         (thunderx2t99_asimd_lut): Add missing tbl types.
19453         (thunderx2t99_asimd_ext): Delete.
19454         (thunderx2t99_asimd_load1_1_mult): Delete.
19455         (thunderx2t99_asimd_load1_2_mult): Delete.
19456         (thunderx2t99_asimd_load1_ldp): New.
19457         (thunderx2t99_asimd_load1): New.
19458         (thunderx2t99_asimd_load2): Add missing *load2* types.
19459         (thunderx2t99_asimd_load3): New.
19460         (thunderx2t99_asimd_load4): New.
19461         (thunderx2t99_asimd_store1_1_mult): Delete.
19462         (thunderx2t99_asimd_store1_2_mult): Delete.
19463         (thunderx2t99_asimd_store2_mult): Delete.
19464         (thunderx2t99_asimd_store2_onelane): Delete.
19465         (thunderx2t99_asimd_store_stp): New.
19466         (thunderx2t99_asimd_store1): New.
19467         (thunderx2t99_asimd_store2): New.
19468         (thunderx2t99_asimd_store3): New.
19469         (thunderx2t99_asimd_store4): New.
19471 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
19473         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
19474         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
19476 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
19477             Segher Boessenkool  <segher@kernel.crashing.org>
19479         PR target/85698
19480         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
19481         operand.
19483 2018-05-17  Richard Biener  <rguenther@suse.de>
19485         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
19486         for pruning loop and prune defs feeding only already visited PHIs.
19488 2018-05-17  Richard Biener  <rguenther@suse.de>
19490         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
19492 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
19493             Richard Biener  <rguenther@suse.de>
19495         PR tree-optimization/85793
19496         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
19497         for VMAT_ELEMENTWISE.
19499 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19501         * internal-fn.h (lookup_internal_fn): Declare
19502         * internal-fn.c (lookup_internal_fn): New function.
19503         * gimple.c (gimple_build_call_from_tree): Handle calls to
19504         internal functions.
19505         * gimple-pretty-print.c (dump_gimple_call): Print "." before
19506         internal function names.
19507         * tree-pretty-print.c (dump_generic_node): Likewise.
19508         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
19510 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
19512         * gimple-fold.h (gimple_build): Make the function forms take
19513         combined_fn rather than built_in_function.
19514         (gimple_simplify): Likewise.
19515         * gimple-match-head.c (gimple_simplify): Likewise.
19516         * gimple-fold.c (gimple_build): Likewise.
19517         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
19518         rather than gimple_build_call_internal.
19519         (get_initial_defs_for_reduction): Likewise.
19520         (vect_create_epilog_for_reduction): Likewise.
19521         (vectorizable_live_operation): Likewise.
19523 2018-05-17  Martin Liska  <mliska@suse.cz>
19525         * gimple-ssa-sprintf.c (format_directive): Do not use
19526         space in between 'G_' and '('.
19528 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
19530         PR target/85323
19531         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
19532         even if the mask is not all ones.
19534         PR target/85323
19535         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
19536         vector.
19537         (ix86_gimple_fold_builtin): Likewise.
19539         PR target/85323
19540         * config/i386/i386.c: Include tree-vector-builder.h.
19541         (ix86_vector_shift_count): New function.
19542         (ix86_fold_builtin): Fold shift builtins by scalar count.
19543         (ix86_gimple_fold_builtin): Likewise.
19545         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
19546         _mm512_setzero): New intrinsics.
19548 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
19549             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19551         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
19552         code generation for cases where splatting a value is not useful.
19553         * simplify-rtx.c (simplify_ternary_operation): Simplify
19554         vec_merge across a vec_duplicate and a paradoxical subreg forming
19555         a vector mode to a vec_concat.
19557 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
19559         * config.gcc: Support "goldmont-plus".
19560         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19561         "goldmont-plus".
19562         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19563         PROCESSOR_GOLDMONT_PLUS.
19564         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
19565         (processor_target_table): Add "goldmont-plus".
19566         (PTA_GOLDMONT_PLUS): Define.
19567         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
19568         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
19569         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
19570         (fold_builtin_cpu): Add "goldmont-plus".
19571         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
19572         (ix86_option_override_internal): Add "goldmont-plus".
19573         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
19574         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
19575         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
19576         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
19578 2018-05-17  Richard Biener  <rguenther@suse.de>
19580         PR tree-optimization/85757
19581         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
19582         remove defs that only feed that PHI from further processing.
19584 2018-05-16  Jim Wilson  <jimw@sifive.com>
19586         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
19587         asterisk to name.
19588         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
19590 2018-05-16  Mark Wielaard  <mark@klomp.org>
19592         * dwarf2out.c (count_index_strings): New function.
19593         (output_indirect_strings): Call count_index_strings and generate
19594         header for dwarf_version >= 5.
19596 2018-05-16  Mark Wielaard  <mark@klomp.org>
19598         * dwarf2out.c (dwarf_FORM): New function.
19599         (set_indirect_string): Use dwarf_FORM.
19600         (reset_indirect_string): Likewise.
19601         (size_of_die): Likewise.
19602         (value_format): Likewise.
19603         (output_die): Likewise.
19604         (add_skeleton_AT_string): Likewise.
19605         (output_macinfo_op): Likewise.
19606         (index_string): Likewise.
19607         (output_index_string_offset): Likewise.
19608         (output_index_string): Likewise.
19609         (count_index_strings): Likewise.
19611 2018-05-16  Carl Love  <cel@us.ibm.com>
19613         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
19614         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
19616 2018-05-16  Martin Jambor  <mjambor@suse.cz>
19618         * ipa-prop.c (ipa_free_all_edge_args): Remove.
19619         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
19621 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
19623         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
19624         (fnma<mode>4): Likewise.
19625         (fms<mode>4): Likewise.
19626         (fnms<mode>4): Likewise.
19627         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
19628         (aarch64_fnma<mode>4): Likewise.
19629         (aarch64_fms<mode>4): Likewise.
19630         (aarch64_fnms<mode>4): Likewise.
19631         (aarch64_fnmadd<mode>4): Likewise.
19633 2018-05-16  Jason Merrill  <jason@redhat.com>
19635         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
19637 2018-05-16  Richard Biener  <rguenther@suse.de>
19639         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
19640         (dump_stmt_cost): Declare.
19641         (add_stmt_cost): Dump cost we add.
19642         (add_stmt_costs): New function.
19643         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
19644         No longer exported.
19645         (vect_analyze_stmt): Adjust prototype.
19646         (vectorizable_condition): Likewise.
19647         (vectorizable_live_operation): Likewise.
19648         (vectorizable_reduction): Likewise.
19649         (vectorizable_induction): Likewise.
19650         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
19651         cost vector to pass to vectorizable_ and record afterwards.
19652         (vect_model_reduction_cost): Take cost vector argument and adjust.
19653         (vect_model_induction_cost): Likewise.
19654         (vectorizable_reduction): Likewise.
19655         (vectorizable_induction): Likewise.
19656         (vectorizable_live_operation): Likewise.
19657         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
19658         SLP_TREE_NUMBER_OF_VEC_STMTS.
19659         (vect_analyze_slp_cost_1): Remove.
19660         (vect_analyze_slp_cost): Likewise.
19661         (vect_slp_analyze_node_operations): Take visited args and
19662         a target cost vector.  Avoid processing already visited stmt sets.
19663         (vect_slp_analyze_operations): Use a local cost vector to gather
19664         costs and register those of non-discarded instances.
19665         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
19666         (vect_schedule_slp_instance): Remove copying of
19667         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
19668         zero.
19669         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
19670         adding cost.  Record cost entry location.
19671         (vect_prologue_cost_for_slp_op): Function to compute cost of
19672         a constant or invariant generated for SLP vect in the prologue,
19673         split out from vect_analyze_slp_cost_1.
19674         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
19675         (vect_model_promotion_demotion_cost): Likewise.
19676         (vect_model_store_cost): Likewise, make static.
19677         (vect_model_load_cost): Likewise.
19678         (vectorizable_bswap): Add cost vector arg and adjust.
19679         (vectorizable_call): Likewise.
19680         (vectorizable_simd_clone_call): Likewise.
19681         (vectorizable_conversion): Likewise.
19682         (vectorizable_assignment): Likewise.
19683         (vectorizable_shift): Likewise.
19684         (vectorizable_operation): Likewise.
19685         (vectorizable_store): Likewise.
19686         (vectorizable_load): Likewise.
19687         (vectorizable_condition): Likewise.
19688         (vectorizable_comparison): Likewise.
19689         (can_vectorize_live_stmts): Likewise.
19690         (vect_analyze_stmt): Likewise.
19691         (vect_transform_stmt): Adjust calls to vectorizable_*.
19692         * tree-vectorizer.c: Include gimple-pretty-print.h.
19693         (dump_stmt_cost): New function.
19695 2018-05-16  Richard Biener  <rguenther@suse.de>
19697         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
19698         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
19699         * tree-ssa-dse.c: Include tree-ssa-loop.h.
19700         (check_name): New callback.
19701         (dse_classify_store): Track cycles via a visited bitmap of PHI
19702         defs and simplify handling of in-loop and across loop dead stores
19703         and properly fail for loop-variant refs.  Handle byte-tracking with
19704         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
19705         limiting the walk.
19707 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
19709         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
19710         (vect_get_mask_type_for_stmt): Likewise.
19711         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
19712         split out from...
19713         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
19714         to determine the statement's vector type and the vector type that
19715         should be used for calculating nunits.  Deal with cases in which
19716         the type has to be deferred.
19717         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
19718         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
19719         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
19720         (vect_determine_vf_for_stmt): New functions, split out from...
19721         (vect_determine_vectorization_factor): ...here.
19722         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
19723         (vect_get_mask_type_for_stmt): New functions, split out from
19724         vect_determine_vectorization_factor.
19726 2018-05-16  Richard Biener  <rguenther@suse.de>
19728         * tree-cfg.c (verify_gimple_assign_ternary): Properly
19729         verify the [VEC_]COND_EXPR embedded comparison.
19731 2018-05-15  Martin Sebor  <msebor@redhat.com>
19733         PR tree-optimization/85753
19734         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
19735         RECORD_TYPE in addition to ARRAY_TYPE.
19737 2018-05-15  Martin Sebor  <msebor@redhat.com>
19739         PR middle-end/85643
19740         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
19742 2018-05-15  Richard Biener  <rguenther@suse.de>
19744         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
19745         add by_clobber_p one.  Change algorithm to collect all defs
19746         representing uses we need to walk and try reducing them to
19747         a single one before failing.
19748         (dse_dom_walker::dse_optimize_stmt): Adjust.
19750 2018-05-13  Mark Wielaard  <mark@klomp.org>
19752         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
19753         (size_of_loc_descr): Likewise.
19754         (output_loc_operands): Likewise.
19755         (output_loc_operands_raw): Likewise.
19756         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
19757         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
19758         (hash_loc_operands): Likewise.
19759         (compare_loc_operands): Likewise.
19761 2018-05-14  Mark Wielaard  <mark@klomp.org>
19763         * dwarf2out.c (count_index_addrs): New function.
19764         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
19766 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19768         PR tree-optimization/83648
19769         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
19770         return value as malloc candidate.
19772 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
19774         PR ipa/85734
19775         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
19776         param as true in call to suggest_attribute.
19778 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
19780         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
19781         -mreadonly-in-sdata.
19783 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
19785         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
19786         New pattern.
19787         (aarch64_crypto_aesd_fused): Likewise.
19789 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
19791         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
19792         (movsi_aarch64): Likewise.
19793         (load_pairsi): Likewise.
19794         (load_pairdi): Likewise.
19795         (store_pairsi): Likewise.
19796         (store_pairdi): Likewise.
19797         (load_pairsf): Likewise.
19798         (load_pairdf): Likewise.
19799         (store_pairsf): Likewise.
19800         (store_pairdf): Likewise.
19801         (zero_extend): Likewise.
19802         (trunc): Swap alternatives.
19803         (fcvt_target): Add '?' to prefer w over r.
19805 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
19807         PR target/85756
19808         * config/i386/i386.md: Disallow non-commutative arithmetics in
19809         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
19810         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
19811         in the peephole2 before it.
19813 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
19815         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
19816         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
19817         (ix86_handle_option): Handle -mcldemote.
19818         * config.gcc: New header.
19819         * config/i386/cldemoteintrin.h: New file.
19820         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
19821         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
19822         -mcldemote.
19823         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19824         OPTION_MASK_ISA_CLDEMOTE.
19825         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
19826         (ix86_valid_target_attribute_inner_p): Ditto.
19827         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
19828         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
19829         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
19830         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
19831         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
19832         (cldemote): New.
19833         * config/i386/i386.opt: Add -mcldemote.
19834         * config/i386/x86intrin.h: New header.
19835         * doc/invoke.texi: Add -mcldemote.
19837 2018-05-14  Richard Biener  <rguenther@suse.de>
19839         * doc/match-and-simplify.texi: Adjust :s documentation.
19841 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
19843         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
19844         intended memcpy size.
19845         (REORDER_45): Likewise.
19847 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
19849         * sort.cc: New file.
19850         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
19851         * vec.c (qsort_chk): Use gcc_qsort.
19852         * Makefile.in (OBJS-libcommon): Add sort.o.
19853         (build/sort.o): New target.  Use it...
19854         (BUILD_RTL): ... here, and...
19855         (build/gencfn-macros): ... here, and...
19856         (build/genmatch): ... here.
19858 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19859             Chung-Ju Wu  <jasonwucj@gmail.com>
19861         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
19862         * config/nds32/nds32-graywolf.md: New file.
19863         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
19864         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
19865         pipeline.
19866         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
19867         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
19868         * config/nds32/nds32.md (pipeline_model): Add graywolf.
19869         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
19870         * config/nds32/pipelines.md: Include n15 settings.
19872 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19873             Chung-Ju Wu  <jasonwucj@gmail.com>
19875         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
19876         * config/nds32/nds32-n13.md: New file.
19877         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
19878         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
19879         pipeline.
19880         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
19881         * config/nds32/nds32.md (pipeline_model): Add n13.
19882         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
19883         * config/nds32/pipelines.md: Include n13 settings.
19885 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
19886             Chung-Ju Wu  <jasonwucj@gmail.com>
19888         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
19889         * config/nds32/nds32-n10.md: New file.
19890         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
19891         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
19892         pipeline.
19893         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
19894         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
19895         * config/nds32/nds32.md (pipeline_model): Add n10.
19896         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
19897         * config/nds32/pipelines.md: Include n10 settings.
19899 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
19900             Kito Cheng  <kito.cheng@gmail.com>
19901             Chung-Ju Wu  <jasonwucj@gmail.com>
19903         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
19904         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
19905         Add enum values for DSP extension instructions.
19906         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
19907         New constraints.
19908         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
19909         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
19910         New code iterators.
19911         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
19912         * config/nds32/nds32-dspext.md: New file for DSP implementation.
19913         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
19914         * config/nds32/nds32-intrinsic.md: Likewise.
19915         * config/nds32/nds32_intrinsic.h: Likewise.
19916         * config/nds32/nds32-md-auxiliary.c: Likewise.
19917         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
19918         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
19919         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
19920         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
19921         * config/nds32/nds32-protos.h: New declarations for DSP extension.
19922         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
19923         TYPE_DMAC in switch statement.
19924         * config/nds32/nds32.c: New checking and implementation for DSP
19925         extension instructions.
19926         * config/nds32/nds32.h: Likewise.
19927         * config/nds32/nds32.md: Likewise.
19928         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
19929         * config/nds32/predicates.md: Implement new predicates for DSP
19930         extension.
19932 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
19934         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
19935         Reformat alternatives and attributes so it is easier to identify
19936         which constraints/attributes go with which instruction.
19937         (mov<mode>_hardfloat32, FMOVE64): Likewise.
19938         (mov<mode>_softfloat32, FMOVE64): Likewise.
19939         (mov<mode>_hardfloat64, FMOVE64): Likewise.
19940         (mov<mode>_softfloat64, FMOVE64): Likewise.
19942 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
19944         * doc/extend.texi (PowerPC Built-in Functions): Rename this
19945         subsection.
19946         (Basic PowerPC Built-in Functions): The new name of the
19947         subsection previously known as "PowerPC Built-in Functions".
19948         (Basic PowerPC Built-in Functions Available on all Configurations):
19949         New subsubsection.
19950         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
19951         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
19952         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
19953         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
19955 2018-05-11  Martin Jambor  <mjambor@suse.cz>
19957         PR ipa/85655
19958         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
19959         single const.
19961 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19963         PR target/85733
19964         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
19966 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
19968         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
19969         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
19970         (ix86_handle_option): Handle -mwaitpkg.
19971         * config.gcc: New header.
19972         * config/i386/cpuid.h (bit_WAITPKG): New bit.
19973         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
19974         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
19975         function type.
19976         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
19977         OPTION_MASK_ISA_WAITPKG.
19978         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
19979         (ix86_option_override_internal): Add PTA_WAITPKG.
19980         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
19981         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
19982         IX86_BUILTIN_TPAUSE.
19983         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
19984         __builtin_ia32_umwait and __builtin_ia32_tpause.
19985         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
19986         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
19987         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
19988         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
19989         UNSPECV_TPAUSE): New.
19990         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
19991         * config/i386/i386.opt: Add -mwaitpkg.
19992         * config/i386/waitpkgintrin.h: New file.
19993         * config/i386/x86intrin.h: New header.
19994         * doc/invoke.texi: Add -mwaitpkg.
19996 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
19998         PR target/85606
19999         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
20000         equivalent.
20001         (cortex-m0): Use armv6s-m isa.
20002         (cortex-m0plus): Likewise.
20003         (cortex-m1): Likewise.
20004         (cortex-m0.small-multiply): Likewise.
20005         (cortex-m0plus.small-multiply): Likewise.
20006         (cortex-m1.small-multiply): Likewise.
20008 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
20009             Jakub Jelinek  <jakub@redhat.com>
20011         PR tree-optimization/85692
20012         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
20013         source permute as well.
20015 2018-05-11  Martin Liska  <mliska@suse.cz>
20017         PR sanitizer/85556
20018         * doc/extend.texi: Document LLVM style format for no_sanitize
20019         attribute.
20021 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
20023         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
20024         mode_supports_vsx_dform_quad to mode_supports_dq_form.
20025         (mode_supports_vsx_dform_quad): Likewise.
20026         (mode_supports_vmx_dform): Move these functions to be next to the
20027         other mode_supports functions.
20028         (mode_supports_dq_form): Likewise.
20029         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
20030         mode_supports_dq_form.
20031         (reg_offset_addressing_ok_p): Likewise.
20032         (offsettable_ok_by_alignment): Likewise.
20033         (rs6000_legitimate_offset_address_p): Likewise.
20034         (legitimate_lo_sum_address_p): Likewise.
20035         (rs6000_legitimize_address): Likewise.
20036         (rs6000_legitimize_reload_address): Likewise.
20037         (rs6000_secondary_reload_inner): Likewise.
20038         (rs6000_preferred_reload_class): Likewise.
20039         (rs6000_output_move_128bit): Likewise.
20041 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
20043         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
20044         Generate SImode target register for null target.
20045         <case IX86_BUILTIN_XGETBV>: Ditto.
20046         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
20047         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
20049 2018-05-10  Carl Love  <cel@us.ibm.com>
20051         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
20052         dcbtt and dcbtstt if operands[2] is 0.
20054 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
20056         PR target/85693
20057         * config/i386/sse.md (usadv64qi): New expander.
20059 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
20061         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
20062         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
20063         -maltivec=be support.
20064         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
20065         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
20066         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
20067         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
20068         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
20069         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
20070         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
20071         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
20072         altivec_vsumsws): Adjust.
20073         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
20074         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
20075         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
20076         support.
20077         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
20078         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
20079         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
20080         (altivec_lve<VI_char>x): Delete expand.
20081         (*altivec_lve<VI_char>x_internal): Rename to...
20082         (altivec_lve<VI_char>x): ... this.
20083         (altivec_lvxl_<mode>): Delete expand.
20084         (*altivec_lvxl_<mode>_internal): Rename to ...
20085         (altivec_lvxl_<mode>): ... this.
20086         (altivec_stvxl_<mode>): Delete expand.
20087         (*altivec_stvxl_<mode>_internal): Rename to ...
20088         (altivec_stvxl_<mode>): ... this.
20089         (altivec_stve<VI_char>x): Delete expand.
20090         (*altivec_stve<VI_char>x_internal): Rename to ...
20091         (altivec_stve<VI_char>x): ... this.
20092         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
20093         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
20094         reduc_plus_scal_<mode>): Adjust.
20095         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
20096         comment.
20097         (rs6000_cpu_cpp_builtins): Adjust.
20098         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
20099         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
20100         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
20101         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
20102         -maltivec=be support.
20103         (rs6000_split_vec_extract_var): Adjust.
20104         (rs6000_split_v4si_init): Adjust.
20105         (swap_selector_for_mode): Delete.
20106         (altivec_expand_lvx_be, altivec_expand_stvx_be,
20107         altivec_expand_stvex_be): Delete.
20108         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
20109         -maltivec=be support.
20110         (rs6000_gimple_fold_builtin): Ditto.
20111         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
20112         Adjust.
20113         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
20114         (TARGET_DIRECT_MOVE_64BIT): Adjust.
20115         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
20116         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
20117         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
20118         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
20119         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
20120         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
20121         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
20122         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
20123         anonymous split): Adjust.
20124         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
20125         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
20127 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
20129         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
20130         when --with-gxx-include-dir is also specified.
20131         * configure: Regenerate.
20133 2018-05-09  Jim Wilson  <jimw@sifive.com>
20135         PR target/84797
20136         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
20137         * config/riscv/t-withmultilib: New.
20138         * config/riscv/withmultilib.h: New.
20139         * doc/install.texi: Document RISC-V --with-multilib-list support.
20141 2018-05-09  Richard Biener  <rguenther@suse.de>
20143         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
20144         vector.
20145         (vect_bb_vectorization_profitable_p): Adjust.  Compute
20146         actual scalar cost using the cost vector and the add_stmt_cost
20147         machinery.
20149 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20151         PR rtl-optimization/85645
20152         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
20153         in the REG_CFA_REGISTER note for LR, don't leave it empty.
20155 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20157         PR rtl-optimization/85645
20158         * shrink-wrap.c (spread_components): Return a boolean saying if
20159         anything was changed.
20160         (try_shrink_wrapping_separate): Iterate spread_components until
20161         nothing changes anymore.
20163 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20165         PR rtl-optimization/85645
20166         * regrename.c (build_def_use): Also kill the chains that include the
20167         destination of a REG_CFA_REGISTER note.
20169 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
20171         PR rtl-optimization/85645
20172         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
20173         insn that has a REG_CFA_REGISTER note.
20175 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
20177         * cfgexpand.c (expand_clobber): New function.
20178         (expand_gimple_stmt_1): Use it.
20179         * tree-vect-stmts.c (vect_clobber_variable): New function,
20180         split out from...
20181         (vectorizable_simd_clone_call): ...here.
20182         (vectorizable_store): Emit a clobber either side of an
20183         IFN_STORE_LANES sequence.
20184         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
20186 2018-05-09  Tom de Vries  <tom@codesourcery.com>
20188         PR target/85626
20189         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
20190         (define_insn "trap_if_false"): Add exit after trap.
20192 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
20194         PR rtl-optimization/85638
20195         * bb-reorder.c: Include common/common-target.h.
20196         (create_forwarder_block): New function extracted from...
20197         (fix_up_crossing_landing_pad): ...here.  Rename into...
20198         (dw2_fix_up_crossing_landing_pad): ...this.
20199         (sjlj_fix_up_crossing_landing_pad): New function.
20200         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
20201         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
20202         from both partitions and exit the loop after one iteration.
20204 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20206         Revert:
20207         * doc/extend.texi (PowerPC Built-in Functions): Rename this
20208         subsection.
20209         (Basic PowerPC Built-in Functions): The new name of the
20210         subsection previously known as "PowerPC Built-in Functions".
20211         (Basic PowerPC Built-in Functions Available on all Configurations):
20212         New subsubsection.
20213         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20214         subsubsection.
20215         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20216         subsubsection.
20217         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20218         subsubsection.
20219         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20220         subsubsection.
20222 2018-05-08  Jim Wilson  <jimw@sifive.com>
20224         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
20225         (LD_EMUL_SUFFIX): New.
20226         (LINK_SPEC): Use it.
20228 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20230         * doc/extend.texi (PowerPC Built-in Functions): Rename this
20231         subsection.
20232         (Basic PowerPC Built-in Functions): The new name of the
20233         subsection previously known as "PowerPC Built-in Functions".
20234         (Basic PowerPC Built-in Functions Available on all Configurations):
20235         New subsubsection.
20236         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
20237         subsubsection.
20238         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
20239         subsubsection.
20240         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
20241         subsubsection.
20242         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
20243         subsubsection.
20245 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
20247         PR target/85683
20248         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
20249         after cmpelim optimization.
20251 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
20253         * config.gcc: Support "goldmont".
20254         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
20255         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
20256         PROCESSOR_GOLDMONT.
20257         * config/i386/i386.c (m_GOLDMONT): Define.
20258         (processor_target_table): Add "goldmont".
20259         (PTA_GOLDMONT): Define.
20260         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
20261         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
20262         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
20263         (fold_builtin_cpu): Add "goldmont".
20264         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
20265         (ix86_option_override_internal): Add "goldmont".
20266         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
20267         (processor_type): Add PROCESSOR_GOLDMONT.
20268         * config/i386/i386.md: Add CPU "glm".
20269         * config/i386/glm.md: New file.
20270         * config/i386/x86-tune.def: Add m_GOLDMONT.
20271         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
20273 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
20275         PR target/85572
20276         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
20277         E_V4DImode.
20278         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
20279         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
20280         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
20282         PR target/85317
20283         * config/i386/i386.c (ix86_fold_builtin): Handle
20284         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
20286         PR target/85480
20287         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
20288         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
20290 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
20292         PR target/85658
20293         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
20294         (check_arch): Likewise.
20295         (check_fpu): Return the result rather than printing it.
20296         (end arch): Fix operator precedence.
20297         (end cpu): Likewise.
20298         (END): Print the result from check_fpu.
20300 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20301             Alan Hayward  <alan.hayward@arm.com>
20302             David Sherwood  <david.sherwood@arm.com>
20304         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
20305         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
20306         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
20307         (*fcmuo<mode>_and): New patterns.
20309 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
20311         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
20312         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
20313         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
20314         (cmp_op, sve_imm_con): New code attributes.
20315         (SVE_COND_INT_CMP, imm_con): Delete.
20316         (cmp_op): Remove above unspecs from int attribute.
20317         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
20318         to...
20319         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
20320         comparison-specific unspecs.
20321         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
20322         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
20323         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
20324         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
20325         (*vec_fcm<cmp_op><mode>): Rename to...
20326         (*fcm<cmp_op><mode>): ...this and adjust likewise.
20327         (*vec_fcmuo<mode>): Rename to...
20328         (*fcmuo<mode>): ...this and adjust likewise.
20329         (*pred_fcm<cmp_op><mode>): New pattern.
20330         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
20331         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
20332         functions.
20333         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
20334         and UNORDERED.
20335         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
20336         (aarch64_emit_sve_predicated_cond): New function.
20337         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
20338         (aarch64_emit_unspec_cond_or): Replace with...
20339         (aarch64_emit_sve_or_conds): ...this new function.  Use
20340         aarch64_emit_sve_ptrue_op for the individual comparisons and
20341         aarch64_emit_binop to OR them together.
20342         (aarch64_emit_inverted_unspec_cond): Replace with...
20343         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
20344         aarch64_emit_sve_ptrue_op for the comparison and
20345         aarch64_emit_unop to invert the result.
20346         (aarch64_expand_sve_vec_cmp_float): Update after the above
20347         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
20349 2018-05-07  Nathan Sidwell  <nathan@acm.org>
20351         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
20352         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
20353         (Backwards Compatibility): Likewise.
20355 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20357         PR bootstrap/85681
20358         Revert:
20359         2018-05-07  Luis Machado  <luis.machado@linaro.org>
20361         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20362         <prefetch_dynamic_strides>: New const bool field.
20363         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20364         prefetch_dynamic_strides.
20365         (exynosm1_prefetch_tune): Likewise.
20366         (thunderxt88_prefetch_tune): Likewise.
20367         (thunderx_prefetch_tune): Likewise.
20368         (thunderx2t99_prefetch_tune): Likewise.
20369         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20370         to false.
20371         (aarch64_override_options_internal): Update to set
20372         PARAM_PREFETCH_DYNAMIC_STRIDES.
20373         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20374         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20375         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20376         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20377         prefetch-dynamic-strides setting.
20379         2018-05-07  Luis Machado  <luis.machado@linaro.org>
20381         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20382         <minimum_stride>: New const int field.
20383         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20384         minimum_stride field.
20385         (exynosm1_prefetch_tune): Likewise.
20386         (thunderxt88_prefetch_tune): Likewise.
20387         (thunderx_prefetch_tune): Likewise.
20388         (thunderx2t99_prefetch_tune): Likewise.
20389         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20390         (aarch64_override_options_internal): Update to set
20391         PARAM_PREFETCH_MINIMUM_STRIDE.
20392         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20393         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20394         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20395         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20396         stride is constant and is below the minimum stride threshold.
20398 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20400         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
20401         to 512.
20403 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20405         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20406         <prefetch_dynamic_strides>: New const bool field.
20407         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20408         prefetch_dynamic_strides.
20409         (exynosm1_prefetch_tune): Likewise.
20410         (thunderxt88_prefetch_tune): Likewise.
20411         (thunderx_prefetch_tune): Likewise.
20412         (thunderx2t99_prefetch_tune): Likewise.
20413         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
20414         to false.
20415         (aarch64_override_options_internal): Update to set
20416         PARAM_PREFETCH_DYNAMIC_STRIDES.
20417         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
20418         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
20419         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
20420         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
20421         prefetch-dynamic-strides setting.
20423 2018-05-07  Luis Machado  <luis.machado@linaro.org>
20425         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
20426         <minimum_stride>: New const int field.
20427         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
20428         minimum_stride field.
20429         (exynosm1_prefetch_tune): Likewise.
20430         (thunderxt88_prefetch_tune): Likewise.
20431         (thunderx_prefetch_tune): Likewise.
20432         (thunderx2t99_prefetch_tune): Likewise.
20433         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
20434         (aarch64_override_options_internal): Update to set
20435         PARAM_PREFETCH_MINIMUM_STRIDE.
20436         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
20437         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
20438         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
20439         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
20440         stride is constant and is below the minimum stride threshold.
20442 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
20444         PR c++/85659
20445         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
20446         the type is addressable.  Don't force op into register if it has
20447         BLKmode.
20449 2018-05-05  Roland McGrath  <mcgrathr@google.com>
20451         PR other/77609
20452         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
20453         any section for which we don't know a specific type it should have,
20454         regardless of name.  Previously this was done only for the exact
20455         names ".init_array", ".fini_array", and ".preinit_array".
20456         (default_elf_asm_named_section): Add comment about
20457         relationship with default_section_type_flags and SECTION_NOTYPE.
20458         (get_section): Don't consider it a type conflict if one side has
20459         SECTION_NOTYPE and the other doesn't, as long as neither has the
20460         SECTION_BSS et al used in the default_section_type_flags logic.
20462 2018-05-05  Tom de Vries  <tom@codesourcery.com>
20464         PR target/85653
20465         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
20466         (workaround_barsyncs): New function.
20467         (nvptx_reorg): Use workaround_barsyncs.
20468         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
20469         (define_expand "nvptx_membar_cta"): New define_expand.
20470         (define_insn "*nvptx_membar_cta"): New insn.
20472 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
20474         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
20475         To improve optimization opportunities.
20476         * builtin-types.def: The new needed builtin types for the above.
20478 2018-05-04  Richard Biener  <rguenther@suse.de>
20480         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
20481         * gimple-ssa-store-merging.c
20482         (imm_store_chain_info::output_merged_store): Remove redundant create,
20483         release split_store vector contents on failure.
20484         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
20485         scalar stmt vector on cache hit.
20487 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
20489         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
20490         Xilinx FP support.
20491         * config.gcc (powerpc-xilinx-eabi*): Remove.
20492         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
20493         support.
20494         (fusion_addis_mem_combo_load): Ditto.
20495         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
20496         FP support.
20497         (rs6000_cpu_cpp_builtins): Ditto.
20498         * config/rs6000/rs6000-linux.c
20499         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
20500         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
20501         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
20502         support.
20503         (rs6000_setup_reg_addr_masks): Ditto.
20504         (rs6000_init_hard_regno_mode_ok): Ditto.
20505         (rs6000_option_override_internal): Ditto.
20506         (legitimate_lo_sum_address_p): Ditto.
20507         (rs6000_legitimize_address): Ditto.
20508         (rs6000_legitimize_reload_address): Ditto.
20509         (rs6000_legitimate_address_p): Ditto.
20510         (abi_v4_pass_in_fpr): Ditto.
20511         (setup_incoming_varargs): Ditto.
20512         (rs6000_gimplify_va_arg): Ditto.
20513         (rs6000_split_multireg_move): Ditto.
20514         (rs6000_savres_strategy): Ditto.
20515         (rs6000_emit_prologue_components): Ditto.
20516         (rs6000_emit_epilogue_components): Ditto.
20517         (rs6000_emit_prologue): Ditto.
20518         (rs6000_emit_epilogue): Ditto.
20519         (rs6000_elf_file_end): Ditto.
20520         (rs6000_function_value): Ditto.
20521         (rs6000_libcall_value): Ditto.
20522         * config/rs6000/rs6000.h: Ditto.
20523         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
20524         (TARGET_MINMAX): ... this.  New.
20525         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
20526         * config/rs6000/rs6000.md: Remove Xilinx FP support.
20527         (*movsi_internal1_single): Delete.
20528         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
20529         mfpu=, mxilinx-fpu): Delete.
20530         * config/rs6000/singlefp.h: Delete.
20531         * config/rs6000/sysv4.h: Remove Xilinx FP support.
20532         * config/rs6000/t-rs6000: Ditto.
20533         * config/rs6000/t-xilinx: Delete.
20534         * config/rs6000/titan.md: Adjust for fp_type removal.
20535         * config/rs6000/vsx.md: Remove Xilinx FP support.
20536         (VStype_simple): Delete.
20537         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
20538         * config/rs6000/xfpu.h: Delete.
20539         * config/rs6000/xfpu.md: Delete.
20540         * config/rs6000/xilinx.h: Delete.
20541         * config/rs6000/xilinx.opt: Delete.
20542         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
20543         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
20545 2018-05-04  Tom de Vries  <tom@codesourcery.com>
20547         PR libgomp/85639
20548         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
20549         if ignore == 0.
20551 2018-05-04  Richard Biener  <rguenther@suse.de>
20553         PR middle-end/85627
20554         * tree-complex.c (update_complex_assignment): We are always in SSA form.
20555         (expand_complex_div_wide): Likewise.
20556         (expand_complex_operations_1): Likewise.
20557         (expand_complex_libcall): Preserve EH info of the original stmt.
20558         (tree_lower_complex): Handle removed blocks.
20559         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
20560         on complex multiplication and division libcall builtins.
20562 2018-05-04  Richard Biener  <rguenther@suse.de>
20564         PR middle-end/85574
20565         * fold-const.c (negate_expr_p): Restrict negation of operand
20566         zero of a division to when we know that can happen without
20567         overflow.
20568         (fold_negate_expr_1): Likewise.
20570 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
20572         PR libstdc++/85466
20573         * real.h (real_nextafter): Declare.
20574         * real.c (real_nextafter): New function.
20575         * fold-const-call.c (fold_const_nextafter): New function.
20576         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
20577         CASE_CFN_NEXTTOWARD.
20578         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
20579         even when arg1_mode is different from arg0_mode.
20581 2018-05-03  Nathan Sidwell  <nathan@acm.org>
20583         * doc/extend.texi (Deprecated Features): Remove
20584         -ffriend-injection.
20585         (Backwards Compatibility): Likewise.
20586         * doc/invoke.texi (C++ Language Options): Likewise.
20587         (C++ Dialect Options): Likewise.
20589 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
20591         PR target/85530
20592         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
20593         _mm512_mask_mullox_epi64): New intrinsics.
20595 2018-05-03  Tom de Vries  <tom@codesourcery.com>
20597         PR testsuite/85106
20598         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20599         dump files): Add offload-tree.
20601 2018-05-03  Richard Biener  <rguenther@suse.de>
20603         PR tree-optimization/85615
20604         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
20605         to loops not nested in BBs loop father to avoid creating multi-entry
20606         loops.
20608 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
20610         PR tree-optimization/70291
20611         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
20612         arguments.  Change return type to tree.  Emit libcall as a new
20613         statement rather than replacing existing one when inplace_p is true.
20614         (expand_complex_multiplication_components): New function.
20615         (expand_complex_multiplication): Expand floating-point complex
20616         multiplication using the above.
20617         (expand_complex_division): Rename inner_type parameter to type.
20618         Update expand_complex_libcall call-site.
20619         (expand_complex_operations_1): Update expand_complex_multiplication
20620         and expand_complex_division call-sites.
20622 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20624         PR target/85582
20625         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20626         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20627         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
20628         the highest significant bit of the shift count mask is clear.  In
20629         check whether and[sq]i3 is needed verify that all significant bits
20630         of the shift count other than the highest are set.
20632 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20634         PR libgomp/82428
20635         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
20636         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
20637         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
20638         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
20639         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
20640         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
20641         __builtin_goacc_parlevel_size.
20643 2018-05-02  Richard Biener  <rguenther@suse.de>
20645         PR tree-optimization/85597
20646         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
20647         do not use split vect_get_vec_defs call but call vect_get_slp_defs
20648         directly.
20650 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20652         PR testsuite/85106
20653         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20654         dump files): Add ltrans-tree.
20656 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20658         PR testsuite/85106
20659         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
20660         dump files): Add wpa-ipa.
20662 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
20664         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
20665         powerpc*-*-linux*paired* target.
20666         * config/rs6000/750cl.h: Delete.
20667         * config/rs6000/paired.h: Delete.
20668         * config/rs6000/paired.md: Delete.
20669         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
20670         float support.
20671         * config/rs6000/rs6000-builtin.def: Remove paired float support.
20672         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20673         comment.  Remove paired float support.
20674         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
20675         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
20676         VECTOR_PAIRED.
20677         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
20678         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
20679         declarations.
20680         * config/rs6000/rs6000.c: Remove paired float support.
20681         (paired_expand_vector_init, paired_expand_vector_move,
20682         paired_emit_vector_compare, paired_emit_vector_cond_expr,
20683         (paired_expand_lv_builtin, paired_expand_stv_builtin,
20684         paired_expand_builtin, paired_expand_predicate_builtin,
20685         paired_init_builtins): Delete.
20686         * config/rs6000/rs6000.h: Remove paired float support.
20687         * config/rs6000/rs6000.md: Remove paired float support.
20688         (move_from_CR_ov_bit): Delete.
20689         * config/rs6000/rs6000.opt (mpaired): Delete.
20690         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
20691         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
20693 2018-05-02  Richard Biener  <rguenther@suse.de>
20695         PR middle-end/85567
20696         * gimplify.c (gimplify_save_expr): When in SSA form allow
20697         SAVE_EXPRs to compute to SSA vars.
20699 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
20701         PR target/85582
20702         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
20703         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
20704         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
20705         clobber operands[2], instead use a new pseudo.  Formatting fixes.
20707 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
20709         PR tree-optimization/85586
20710         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
20711         exit early for statements in the same group if the accesses are
20712         not strided.
20714 2018-05-02  Tom de Vries  <tom@codesourcery.com>
20716         PR lto/85451
20717         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
20718         error message.
20720 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
20722         PR tree-optimization/85143
20723         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
20725 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20727         PR lto/85451
20728         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
20729         not found" error message.
20731 2018-05-01  Tom de Vries  <tom@codesourcery.com>
20733         PR other/83786
20734         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
20735         * vec.c (test_ordered_remove_if): New function.
20736         (vec_c_tests): Call test_ordered_remove_if.
20737         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
20738         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
20739         * tree-vect-patterns.c (vect_pattern_recog_1): Use
20740         VEC_ORDERED_REMOVE_IF.
20742 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
20744         PR tree-optimization/82665
20745         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
20746         pointer subtraction where arguments come from a memchr call.
20748 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
20750         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
20751         --push-state --as-needed and --pop-state instead of --as-needed and
20752         --no-as-needed if ld supports it.
20753         * configure: Regenerated.
20755         PR web/85578
20756         * doc/install.texi2html: Replace _002d with - and _002a with * in
20757         generated html files using sed.
20759 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20761         PR c++/85523
20762         * gcc-rich-location.c (blank_line_before_p): New function.
20763         (use_new_line): New function.
20764         (gcc_rich_location::add_fixit_insert_formatted): New function.
20765         * gcc-rich-location.h
20766         (gcc_rich_location::add_fixit_insert_formatted): New function.
20768 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20770         * selftest.c (assert_streq): Rename "expected" and "actual" to
20771         "val1" and "val2".  Extend NULL-handling to cover both inputs
20772         symmetrically, while still requiring both to be non-NULL for a pass.
20773         * selftest.h (assert_streq): Rename "expected" and "actual" to
20774         "val1" and "val2".
20775         (ASSERT_EQ): Likewise.
20776         (ASSERT_EQ_AT): Likewise.
20777         (ASSERT_KNOWN_EQ): Likewise.
20778         (ASSERT_KNOWN_EQ_AT): Likewise.
20779         (ASSERT_NE): Likewise.
20780         (ASSERT_MAYBE_NE): Likewise.
20781         (ASSERT_MAYBE_NE_AT): Likewise.
20782         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
20783         the assertion to pass.
20784         (ASSERT_STREQ_AT): Likewise.
20786 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20788         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
20789         interaction with -pie.
20791 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20793         * selftest.h: Fix alphabetization of per-source-file selftest
20794         declarations.
20796 2018-04-30  Jason Merrill  <jason@redhat.com>
20798         PR c++/61982 - dead stores to destroyed objects.
20799         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
20800         of clobber.
20802 2018-04-30  Jason Merrill  <jason@redhat.com>
20804         * tree.c (build_clobber): New.
20805         * tree.h: Declare it.
20806         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
20808 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20810         * diagnostic-show-locus.c (layout::layout): Update for
20811         location_get_source_line returning a char_span.
20812         (struct char_span): Move to input.h.
20813         (struct correction): Update for fields in char_span becoming
20814         private.
20815         (struct source_line): Update for location_get_source_line
20816         returning a char_span.
20817         (layout::print_line): Likewise.
20818         * edit-context.c (edited_file::print_content): Likewise.
20819         (edited_file::print_diff_hunk): Likewise.
20820         (edited_file::print_run_of_changed_lines): Likewise.
20821         (edited_file::get_num_lines): Likewise.
20822         (edited_line::edited_line): Likewise.
20823         * final.c (asm_show_source): Likewise.
20824         * input.c (location_get_source_line): Convert return type
20825         from const char * to char_span, losing the final "line_len"
20826         param.
20827         (dump_location_info): Update for the above.
20828         (get_substring_ranges_for_loc): Likewise.  Use a char_span
20829         when handling the literal within the line.
20830         (test_reading_source_line): Update for location_get_source_line
20831         returning a char_span.
20832         * input.h (class char_span): Move here from
20833         diagnostic-show-locus.c, converting from a struct to a class.
20834         Make data members private.
20835         (char_span::operator bool): New.
20836         (char_span::length): New.
20837         (char_span::get_buffer): New.
20838         (char_span::operator[]): New.
20839         (char_span::subspan): Make const.
20840         (char_span::xstrdup): New.
20841         (location_get_source_line): Convert return type from const char *
20842         to char_span, losing the final "line_size" param.
20844 2018-04-30  Jan Hubicka  <jh@suse.cz>
20846         * lto-wrapper.c (ltrans_priorities): New static var.
20847         (cmp_priority): New.
20848         (run_gcc): Read priorities and if doing parallel build order
20849         the Makefile by them.
20851 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
20853         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
20855 2018-04-30  Richard Biener  <rguenther@suse.de>
20857         * tree-cfg.c (verify_address): Remove base argument, add
20858         flag whether to check TREE_ADDRESSABLE and do that.
20859         (verify_expr): Remove.
20860         (verify_types_in_gimple_reference): Add pieces from verify_expr.
20861         (verify_gimple_assign_single): Likewise.
20862         (verify_gimple_switch): Likewise.
20863         (verify_expr_location_1): Dereference tp once.  Add (disabled)
20864         piece from verify_expr.
20865         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
20867 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20869         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
20871 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20873         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
20874         (small_data_pattern): Likewise.
20875         (arc_rewrite_small_data): Likewise.
20876         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
20877         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
20878         (get_symbol_alignment): New function.
20879         (legitimate_small_data_address_p): Likewise.
20880         (legitimate_scaled_address): Update, call
20881         legitimate_small_data_address_p.
20882         (output_sdata): New static variable.
20883         (arc_print_operand): Update how we handle small data operands.
20884         (arc_print_operand_address): Likewise.
20885         (arc_legitimate_address_p): Update, use
20886         legitimate_small_data_address_p.
20887         (arc_rewrite_small_data_p): Remove.
20888         (arc_rewrite_small_data_1): Likewise.
20889         (arc_rewrite_small_data): Likewise.
20890         (small_data_pattern): Likewise.
20891         (compact_sda_memory_operand): Update to use
20892         legitimate_small_data_address_p and get_symbol_alignment.
20893         (prepare_move_operands): Don't rewite sdata pattern.
20894         (prepare_extend_operands): Remove.
20895         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
20896         pattern.
20897         (zero_extendqisi2): Likewise.
20898         (zero_extendhisi2): Likewise.
20899         (extendqihi2): Likewise.
20900         (extendqisi2): Likewise.
20901         (extendhisi2): Likewise.
20902         (addsi3): Likewise.
20903         (subsi3): Likewise.
20904         (andsi3): Likewise.
20905         * config/arc/constraints.md (Usd): Change it to memory constraint.
20907 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
20909         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
20910         as source of std instructions.
20911         * config/arc/arc.md (movsi_insn): Update pattern predicate to
20912         allow 6-bit constants as source for store instructions.
20913         (movdi_insn): Update instruction pattern to allow 6-bit constants
20914         as source for store instructions.
20916 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
20918         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
20920 2018-04-30  Nathan Sidwell  <nathan@acm.org>
20921             Sandra Loosemore <sandra@codesourcery.com>
20923         * dumpfile.c (dump_open): Allow '-' for stdout.
20924         * doc/invoke.texi (Developer Options): Document dump filename
20925         determination early.  Document stdin/stdout selection.
20927 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
20929         Microblaze Target: PIC data text relative
20931         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
20932         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
20933         Add declaration.
20934         * config/microblaze/microblaze.h (microblaze_constant_address_p):
20935         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
20936         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
20937         New addressing mode for data-text relative position indepenedent code.
20938         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
20939         'ADDRESS_SYMBOLIC_TXT_REL'.
20940         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
20941         (microblaze_legitimate_pic_operand): Exclude function calls from
20942         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
20943         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
20944         addresses cases.
20945         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20946         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
20947         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
20948         for 'address + offset'.
20949         (microblaze_expand_prologue): Add new function prologue call for
20950         'r20' assignation.
20951         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
20952         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
20953         table in case of TARGET_PIC_DATA_TEXT_REL.
20954         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
20955         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
20956         Add new macros 'UNSPEC_TEXT',
20957         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
20958         + exclude function calls from 'UNSPEC_PLT' in case of data text
20959         relative mode.
20960         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20961         new target hook for generating address diff vector tables in case of
20962         flag_pic.
20963         * doc/tm.texi : Regenerate.
20964         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
20965         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
20966         of addr diff vector generation.
20967         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
20968         target hook definition.
20969         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
20970         Add default function for generate_pic_addr_diff_vec -> flag_pic.
20971         * doc/invoke.texi (Add new pic option): Add new microblaze pic
20972         option for data text relative.
20974 2018-04-30  Richard Biener  <rguenther@suse.de>
20976         * tree-chrec.h (evolution_function_is_constant_p): Remove
20977         redundant check.
20978         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
20980 2018-04-30  Richard Biener  <rguenther@suse.de>
20982         PR bootstrap/85571
20983         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
20985 2018-04-30  Richard Biener  <rguenther@suse.de>
20987         PR tree-optimization/28364
20988         PR tree-optimization/85275
20989         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
20990         copying first exit test.
20992 2018-04-28  Mark Wielaard  <mark@klomp.org>
20994         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
20995         dwarf_version >= 5.
20996         (dwarf_AT): Handle DW_AT_addr_base.
20997         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
20999 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
21001         PR target/84431
21002         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
21003         (*ashl<dwi>3_doubleword_mask_1): Ditto.
21004         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
21005         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
21007 2018-04-28  Richard Biener  <rguenther@suse.de>
21009         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
21010         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
21011         to reflect use.  Only add interesting stmts.
21013 2018-04-27  Martin Jambor  <mjambor@suse.cz>
21015         PR ipa/85549
21016         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
21017         the jump function allows for passing through aggregate values.
21019 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
21021         * input.h (in_system_header_at): Convert from macro to inline
21022         function.
21023         (from_macro_expansion_at): Likewise.
21024         (from_macro_definition_at): Likewise.
21026 2018-04-27  Jeff Law  <law@redhat.com>
21028         * config.gcc: Mark tile* targets as deprecated/obsolete.
21030 2018-04-27  Richard Biener  <rguenther@suse.de>
21032         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
21033         fix for ILP32.
21035 2018-04-27  Richard Biener  <rguenther@suse.de>
21037         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
21039 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
21041         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
21042         with Yd constraint. Set "preferred_for_speed" attribute from
21043         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
21044         with Yd constraint.
21045         (*movdi_internal): Ditto.
21046         (movti_interunit splitters): Remove
21047         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
21048         (movdi_interunit splitters): Ditto.
21049         * config/i386/constraints.md (Ye): Remove.
21050         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
21052 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21054         PR target/85512
21055         * config/aarch64/constraints.md (Usg): Limit to 31.
21056         (Usj): Limit to 63.
21058 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
21060         PR tree-optimization/85529
21061         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
21062         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
21063         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
21064         zero extension or masking of the MSB bit.
21065         (optimize_range_tests): Add FIRST_BB argument, pass it through
21066         to optimize_range_tests_var_bound.
21067         (maybe_optimize_range_tests, reassociate_bb): Adjust
21068         optimize_range_tests callers.
21070 2018-04-26  Richard Biener  <rguenther@suse.de>
21071             Jakub Jelinek  <jakub@redhat.com>
21073         * cgraph.h (symbol_table): Just declare debug method here.
21074         * symtab.c (symbol_table::debug): Define.
21076 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
21078         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
21080 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
21082         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
21083         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
21084         (*movdi_internal): Substitute Yi and Yj constraint with x
21085         and Ym and Yn constraint with y constraint.  Update "isa"
21086         attribute and set "preferred_for_speed" attribute from
21087         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
21088         (*movsi_internal): Ditto.
21089         (*movdf_internal): Ditto.
21090         (*movsf_internal): Ditto.
21091         (*zero_extendsidi2): Ditto.
21092         * config/i386/sse.md (vec_set<mode>_0): Ditto.
21093         (sse2_loadld): Ditto.
21094         (*vec_extract<ssevecmodelower>_0): Ditto.
21095         (*vec_extractv4si_0_zext_sse4): Ditto.
21096         (vec_concatv2di): Ditto.
21097         (*vec_dup<mode>): Ditto.
21098         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
21099         * config/i386/constraints.md (Yi): Remove.
21100         (Yj): Remove.
21101         (Ym): Remove.
21102         (Yn): Remove.
21104 2018-04-26  Nathan Sidwell  <nathan@acm.org>
21106         * dumpfile.c (dump_open): New.
21107         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
21108         (dump_finish): Detect stdio/stderr by value not name.
21110 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
21112         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
21114 2018-04-26  Tom de Vries  <tom@codesourcery.com>
21116         PR target/84952
21117         * config/nvptx/nvptx.c (verify_neutering_jumps)
21118         (verify_neutering_labels): New function
21119         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
21121 2018-04-26  Tom de Vries  <tom@codesourcery.com>
21123         PR target/84025
21124         * config/nvptx/nvptx.c (needs_neutering_p): New function.
21125         (nvptx_single): Use needs_neutering_p to skip over insns that do not
21126         need neutering.
21128 2018-04-26  Richard Biener <rguenther@suse.de>
21129             Tom de Vries  <tom@codesourcery.com>
21131         PR lto/85422
21132         * lto-streamer-out.c (output_function): Fixup loops if required to match
21133         discovery done in the reader.
21135 2018-04-26  Richard Biener  <rguenther@suse.de>
21137         PR tree-optimization/85116
21138         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
21139         have a loop exit from the single latch predecessor.  Remove
21140         case of header with just condition.
21141         (ch_base::copy_headers): Exclude infinite loops from any
21142         processing.
21143         (pass_ch::execute): Record exits.
21145 2018-04-26  Richard Biener  <rguenther@suse.de>
21147         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
21148         prologue cost vector and pass it to vect_get_load_cost.
21149         (vect_get_peeling_costs_all_drs): Likewise.
21150         (vect_peeling_hash_get_lowest_cost): Likewise.
21151         (vect_enhance_data_refs_alignment): Likewise.
21153 2018-04-26  Richard Biener  <rguenther@suse.de>
21155         PR middle-end/85450
21156         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
21157         checking of integer<->pointer conversions.
21158         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
21159         sign-/zero-extending pointer types.
21160         (expand_omp_for_static_chunk): Likewise.
21162 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
21163             Jean Lee  <xiaoyur347@gmail.com>
21165         * config/mips/mips.c (mips_asan_shadow_offset): New function.
21166         (TARGET_ASAN_SHADOW_OFFSET): Define.
21167         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
21168         true for -fsanitize=address.
21170 2018-04-25  Mark Wielaard  <mark@klomp.org>
21172         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
21173         shorter ones.
21175 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
21177         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
21178         than "alu", remove explicit "memory" and "imm_disp" attributes.
21179         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
21181         PR middle-end/85414
21182         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
21183         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
21184         gen_lowpart_no_emit.
21186 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
21188         PR target/85473
21189         * config/i386/i386.c (ix86_expand_builtin): Change memory
21190         operand to XI, extend p0 to Pmode.
21191         * config/i386/i386.md: Change unspec volatile and operand
21192         1 mode to XI, change operand 0 mode to P.
21194 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21196         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
21197         GET_MODE_MASK before any checking.
21198         (nds32_can_use_bset_p): Likewise.
21199         (nds32_can_use_btgl_p): Likewise.
21201 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21203         * config/nds32/nds32-doubleword.md: New define_split pattern for
21204         illegal register number.
21206 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21208         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
21210 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
21212         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
21214 2018-04-25  Richard Biener  <rguenther@suse.de>
21216         * lto-streamer.h (LTO_major_version): Bump to 8.
21218 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
21220         * BASE-VER: Set to 9.0.0.
21222 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
21224         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
21225         in __abskf2 and __powikf2.
21227 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21229         PR target/85512
21230         * config/aarch64/constraints.md (Usg, Usj): New constraints.
21231         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
21232         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
21233         Use the above on operand 2.  Reindent.
21234         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
21236 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
21238         PR target/85485
21239         * common/config/i386/i386-common.c (ix86_handle_option): Don't
21240         handle OPT_mcet.
21241         * config/i386/i386.opt (mcet): Removed.
21242         * doc/install.texi: Remove -mcet documentation.
21243         * doc/invoke.texi: Likewise.
21245 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
21247         PR target/85485
21248         * doc/install.texi: Remove -mcet from bootstrap-cet.
21250 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21252         PR target/85511
21253         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
21254         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
21255         if TARGET_64BIT.
21257         PR target/85503
21258         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
21259         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
21260         containing a CONST_VECTOR.
21262 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
21264         * doc/install.texi: Update newlib dependency for nvptx.
21266 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21268         PR target/85508
21269         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
21270         instead of INTVAL when shifting x left.
21272 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
21274         PR tree-optimization/85478
21275         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
21276         vect_grouped_store_supported for single element vectors.
21278 2018-04-24  Richard Biener  <rguenther@suse.de>
21280         PR target/85491
21281         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
21282         load cost increase to the case of non-constant step.
21284 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
21286         PR target/84828
21287         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
21288         destination if any_malformed_asm.
21290 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
21292         PR middle-end/85496
21293         * expr.c (store_field): In the bitfield case, if the value comes from
21294         a function call and is returned in registers by means of a PARALLEL,
21295         do not change the mode of the temporary unless BLKmode and VOIDmode.
21297 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
21299         PR rtl-optimization/85423
21300         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
21301         dependencies to debug insns when the previous insn is non-debug.
21303 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21305         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
21306         enums into a single definition.
21307         (fls): Fix predicates and printing.
21308         (seti): Likewise.
21310 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
21312         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
21313         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
21314         and short u6 immediate.
21315         (check_if_valid_sleep_operand): Remove.
21316         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
21318 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21320         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
21321         flag_always_save_lp condition.
21322         * config/nds32/nds32.opt (malways-save-lp): New option.
21324 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21326         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
21327         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
21328         * config/nds32/nds32.h
21329         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
21330         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
21332 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21334         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
21335         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
21337 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
21338             Chung-Ju Wu  <jasonwucj@gmail.com>
21340         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
21341         Declare.
21342         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
21343         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
21345 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21347         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
21349 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21351         * config/nds32/nds32-protos.h (nds32_data_alignment,
21352         nds32_local_alignment): Declare.
21353         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
21354         nds32_local_alignment): New functions.
21355         (TARGET_CONSTANT_ALIGNMENT): Define.
21356         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
21358 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21360         * config/nds32/nds32.c
21361         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
21362         (TARGET_MODES_TIEABLE_P): Likewise.
21364 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
21366         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
21367         level Ofast and Og.
21369 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
21370             Chung-Ju Wu  <jasonwucj@gmail.com>
21372         * config/nds32/constants.md (unspec_volatile_element): Add enum values
21373         for unaligned access.
21374         * config/nds32/nds32-intrinsic.c: Implementation of expanding
21375         unaligned access.
21376         * config/nds32/nds32-intrinsic.md: Likewise.
21377         * config/nds32/nds32_intrinsic.h: Likewise.
21378         * config/nds32/nds32.h (nds32_builtins): Likewise.
21379         * config/nds32/nds32.opt (munaligned-access): New option.
21380         * config/nds32/nds32.c (nds32_asm_file_start): Display
21381         flag_unaligned_access status.
21383 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
21385         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
21386         -mno-relax is present.
21387         * config/riscv/linux.h (LINK_SPEC): Ditto.
21389 2018-04-20  Martin Sebor  <msebor@redhat.com>
21391         PR c/85365
21392         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
21393         for null pointers.
21394         (gimple_fold_builtin_stxcpy_chk): Same.
21395         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
21397 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
21399         PR target/85456
21400         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
21401         __powikf2 when long double is IEEE 128-bit.
21403 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
21405         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
21406         step to make sure stack always aligned.
21408 2018-04-20  Carl Love  <cel@us.ibm.com>
21410         PR target/83402
21411         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
21412         size check for arg0.
21414 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
21415             Tom de Vries  <tom@codesourcery.com>
21417         PR target/85445
21418         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
21419         Emit insns for calls too.
21420         (nvptx_find_par): Always look for worker-level predecessor insn.
21421         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
21422         calls.
21423         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
21424         (nvptx_process_pars): Propagate frames for calls.
21426 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
21428         PR target/85469
21429         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
21430         Removed.
21431         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
21432         (ix86_handle_option): Don't handle OPT_mibt.
21433         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
21434         __SHSTK__.
21435         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
21436         has_ibt and ibt.
21437         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
21438         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
21439         (ix86_target_macros): Define __CET__ with flag_cf_protection
21440         for -fcf-protection.
21441         * config/i386/i386.c (isa2_opts): Remove -mibt.
21442         * config/i386/i386.h (TARGET_IBT): Removed.
21443         (TARGET_IBT_P): Likewise.
21444         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
21445         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
21446         * config/i386/i386.opt (mcet): Update help message.
21447         (mshstk): Likewise.
21448         (mibt): Removed.
21449         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
21450         -mcet as an alias for -mshstk.
21452 2018-04-20  Richard Biener <rguenther@suse.de>
21454         PR middle-end/85475
21455         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
21456         complexity by forcing a single use of the multiply operand.
21458 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21460         ipa/85449
21461         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
21462         recursion dependency to only apply to non-clones.
21464 2018-04-20  Martin Jambor  <mjambor@suse.cz>
21466         ipa/85447
21467         * ipa-cp.c (create_specialized_node): Check that clones of
21468         self-recursive edges exist during IPA-CP.
21470 2018-04-19  Toon Moene  <toon@moene.org>
21472         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
21473         by -O3.
21475 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21477         PR tree-optimization/85467
21478         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
21479         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
21480         VECTOR_CST element to type.
21482 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21484         PR target/85397
21485         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
21486         * config/i386/i386.md (builtin_setjmp_setup): Removed.
21487         (builtin_longjmp): Likewise.
21488         (save_stack_nonlocal): New pattern.
21489         (restore_stack_nonlocal): Likewise.
21491 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21493         PR target/85404
21494         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21495         Replace ASM_OUTPUT_LABEL with fprintf.
21497 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
21499         PR target/85417
21500         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
21501         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
21502         * config/i386/i386-c.c (ix86_target_macros_internal): Also
21503         define __IBT__ and __SHSTK__ for -fcf-protection.
21504         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
21505         TARGET_IBT.
21506         (ix86_trampoline_init): Likewise.
21507         (x86_output_mi_thunk): Likewise.
21508         (ix86_notrack_prefixed_insn_p): Likewise.
21509         (ix86_option_override_internal): Don't disallow -fcf-protection.
21510         * config/i386/i386.md (rdssp<mode>): Also enable for
21511         -fcf-protection.
21512         (incssp<mode>): Likewise.
21513         (nop_endbr): Likewise.
21514         * config/i386/i386.opt (mcet): Change help message to built-in
21515         functions only.
21516         (mibt): Likewise.
21517         (mshstk): Likewise.
21518         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
21519         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
21520         enable CET built-in functions.
21522 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
21524         * common/config/i386/i386-common.c
21525         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
21526         OPTION_MASK_ISA_MOVDIRI_UNSET,
21527         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
21528         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
21529         * config.gcc (movdirintrin.h): New header.
21530         * config/i386/cpuid.h (bit_MOVDIRI,
21531         bit_MOVDIR64B): New bits.
21532         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
21533         and -mmvodir64b.
21534         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
21535         (VOID, PVOID, PCVOID)): New function types.
21536         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
21537         __builtin_ia32_directstoreu_u64,
21538         __builtin_ia32_movdir64b): New builtins.
21539         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
21540         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
21541         and -mmovdiri.
21542         (ix86_valid_target_attribute_inner_p): Ditto.
21543         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
21544         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
21545         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
21546         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
21547         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
21548         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
21549         (movdiri<mode>, movdir64b_<mode>): New.
21550         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
21551         * config/i386/immintrin.h: Include movdirintrin.h.
21552         * config/i386/movdirintrin.h: New file.
21553         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
21555 2018-04-19  Richard Biener  <rguenther@suse.de>
21557         PR middle-end/85455
21558         * cfg.c (clear_bb_flags): When loop state says we have
21559         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
21561 2018-04-19  Richard Biener  <rguenther@suse.de>
21563         PR tree-optimization/84737
21564         * tree-vect-data-refs.c (vect_copy_ref_info): New function
21565         copying restrict info.
21566         (vect_setup_realignment): Use it.
21567         * tree-vectorizer.h (vect_copy_ref_info): Declare.
21568         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
21569         the first DR to all generated stores.
21570         (vectorizable_load): Likewise for loads.
21572 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
21574         PR tree-optimization/85446
21575         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
21576         the integral and pointer types to have the same precision.
21578         * doc/install.texi: Document --disable-cet being the default and
21579         --enable-cet=auto.
21581 2018-04-18  Martin Liska  <mliska@suse.cz>
21583         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
21584         style.
21586 2018-04-18  Martin Liska  <mliska@suse.cz>
21588         Revert
21589         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
21591         PR ipa/83983
21592         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
21593         arguments if they are comparable.
21595 2018-04-18  Martin Liska  <mliska@suse.cz>
21597         Revert
21598         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
21600         PR lto/84805
21601         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
21602         incomplete types.
21604 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
21606         PR target/85388
21607         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
21608         ENDBR after calling __morestack.
21610 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
21612         PR jit/85384
21613         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
21614         by using gcc_base_ver to generate a gcc_driver_version, and use
21615         it when generating GCC_DRIVER_NAME.
21616         * configure: Regenerate.
21618 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
21620         PR target/81084
21621         * config.gcc: Obsolete powerpc*-*-*spe*.
21623 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21625         PR debug/84637
21626         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
21627         (stabstr_D): Change type of unum from unsigned int to
21628         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
21629         type.
21631 2018-04-17  Jim Wilson  <jimw@sifive.com>
21633         PR 84856
21634         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
21635         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
21636         Set arg_pointer_offset after using pretend_args_size.
21638 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21640         PR rtl-optimization/85431
21641         * dse.c (record_store): Ignore zero width stores.
21643         PR sanitizer/85230
21644         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
21645         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
21646         __builtin_stack_restore rather than after it.
21647         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
21648         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
21649         argument instead of virtual_dynamic_stack_rtx.
21651 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
21653         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
21654         New prototype.
21655         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21656         Add note to error message to explain internal mapping of overloaded
21657         built-in function name to non-overloaded built-in function name.
21658         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
21659         function.
21661 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
21663         PR target/85424
21664         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
21665         where the inputs overlap with the output.
21667 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21669         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
21670         (=v, v) alternative and explicit "memory" attribute.
21671         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
21672         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21673         attributes.
21674         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
21675         "sselog1" type instead of "sselog".
21676         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
21677         "sselog".  Remove explicit "memory" attribute.
21678         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
21679         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21680         attributes.
21681         (vec_extract_hi_v32hi): Merge all alternatives into one, use
21682         "sselog1" type instead of "sselog".  Remove explicit "memory"
21683         attribute.
21684         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
21685         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21686         attribute.
21687         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
21688         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
21689         attributes.
21690         (vec_extract_hi_v64qi): Merge all alternatives into one, use
21691         "sselog1" type instead of "sselog".  Remove explicit "memory"
21692         attribute.
21693         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
21694         use "sselog1" type instead of "sselog".  Remove explicit "memory"
21695         attribute.
21697         PR target/85430
21698         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
21700         PR middle-end/85414
21701         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
21702         on a SUBREG.
21704 2018-04-17  Martin Jambor  <mjambor@suse.cz>
21706         PR ipa/85421
21707         * ipa-cp.c (create_specialized_node): Call
21708         expand_all_artificial_thunks if necessary.
21710 2018-04-17  Martin Liska  <mliska@suse.cz>
21712         PR lto/85405
21713         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
21714         in message, remote space in between '_G' and '('.
21716 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
21718         PR target/85281
21719         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
21720         avx512f_vmcmp<mode>3<round_saeonly_name>,
21721         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
21722         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
21723         avx512f_rndscale<mode><round_saeonly_name>,
21724         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
21725         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
21726         Use %<iptr>2 instead of %2 for -masm=intel.
21727         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
21728         avx512f_vcvttss2usi<round_saeonly_name>,
21729         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
21730         -masm=intel.
21731         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
21732         avx512f_vcvttsd2usi<round_saeonly_name>,
21733         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
21734         Use %q1 instead of %1 for -masm=intel.
21735         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
21736         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
21737         of %3 for -masm=intel.
21738         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
21739         -masm=intel.
21740         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
21741         -masm=intel.
21742         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
21743         -masm=intel.
21744         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
21745         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
21746         %g1.
21747         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
21748         -masm=intel.
21749         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
21750         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
21751         %g1 and one with %0 and %1.
21752         (avx512er_vmrcp28<mode><round_saeonly_name>,
21753         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
21754         %1 for -masm=intel.
21755         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
21756         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
21757         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
21758         of %0 and %{%4%} for -masm=intel.
21759         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
21760         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
21761         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
21762         order of %0 and %{%5%}%{z%} for -masm=intel.
21764 2018-04-17  Jan Hubicka  <jh@suse.cz>
21766         PR lto/85405
21767         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
21769 2018-04-17  Martin Liska  <mliska@suse.cz>
21771         PR ipa/85329
21772         * multiple_target.c (create_dispatcher_calls): Set apostrophes
21773         for target_clone error message.  Make default implementation
21774         clone to be a local declaration.
21775         (separate_attrs): Add new argument and check for an empty
21776         string.
21777         (expand_target_clones): Handle it.
21778         (ipa_target_clone): Make redirection just for target_clones
21779         functions.
21781 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
21782             Tom de Vries  <tom@codesourcery.com>
21784         PR middle-end/84955
21785         * omp-expand.c (expand_oacc_for): Add dummy false branch for
21786         tiled basic blocks without omp continue statements.
21788 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
21790         PR target/83660
21791         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
21792         vec_extract expression as having side effects to make sure it gets
21793         a cleanup point.
21795 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
21797         PR target/85403
21798         * config/i386/i386.c (get_builtin_code_for_version): Check
21799         error_mark_node.
21801 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
21803         PR target/84331
21804         * config.gcc: Support "skylake".
21805         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
21806         PROCESSOR_SKYLAKE.
21807         * config/i386/i386.c (m_SKYLAKE): Define.
21808         (processor_target_table): Add "skylake".
21809         (ix86_option_override_internal): Add "skylake".
21810         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
21811         PROCESSOR_CANNONLAKE.
21812         (get_builtin_code_for_version): Fix priority for
21813         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
21814         PROCESSOR_SKYLAKE-AVX512.
21815         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
21816         (processor_type): Add PROCESSOR_SKYLAKE.
21818 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
21819             Jason Merrill  <jason@redhat.com>
21821         PR c++/85112
21822         * convert.c (convert_to_integer_1): Use direct recursion for
21823         enumeral types and types with a precision less than the number
21824         of bits in their mode.
21826 2018-04-16  Julia Koval  <julia.koval@intel.com>
21828         PR target/84413
21829         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
21830         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
21832 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
21834         PR target/85293
21835         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
21836         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
21837         and -mno-direct-move.
21839 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
21841         PR target/83402
21842         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
21843         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
21844         Ensure negative shifts result in {0}.
21846 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
21848         PR rtl-optimization/79916
21849         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
21850         regs (if any) to define how to gnerate SD moves when LRA is in
21851         progress.
21853 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
21855         PR rtl-optimization/85393
21856         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
21857         * except.c (expand_dw2_landing_pad_for_region): Make static.
21858         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
21859         a label and unconditional jump to old_bb, rather than
21860         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
21861         basic block.
21863         PR rtl-optimization/85376
21864         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
21865         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
21866         instead of a specific value.
21868 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21869             Bin Cheng  <bin.cheng@arm.com>
21871         PR tree-optimization/82965
21872         PR tree-optimization/83991
21873         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
21874         by_profile_only parameter.
21875         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
21876         information if the loop was predicted to iterate too many times.
21877         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
21879 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
21881         PR lto/71991
21882         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
21883         always inline.
21885 2018-04-13  Martin Liska  <mliska@suse.cz>
21886             Jakub Jelinek  <jakub@redhat.com>
21888         PR middle-end/81657
21889         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
21890         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
21891         * builtins.c (expand_builtin_memory_copy_args): Use
21892         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
21893         handle dest_addr == pc_rtx.
21895 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
21897         PR target/85291
21898         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
21899         asked to not generate direct moves.
21900         (fix_trunc<mode>si2_stfiwx): Similar.
21901         (fix_trunc<mode>si2_internal): Similar.
21903 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21905         PR debug/83157
21906         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
21907         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
21908         lookup if dest in some wider mode is known to be const0_rtx and
21909         if so, record permanent equivalence for it to be ZERO_EXTEND of
21910         the narrower mode destination.
21912 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21914         * lto-streamer-out.c (output_function): Revert 259346.
21915         * omp-expand.c (expand_oacc_for): Likewise.
21917 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
21919         PR rtl-optimization/85354
21920         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
21921         * sel-sched.c (sel_global_init): ... here.
21923 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
21925         PR target/85238
21926         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
21927         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
21928         mode for PE-COFF targets.
21929         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
21930         (i386_pe_asm_lto_end): Likewise.
21931         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
21932         (TARGET_ASM_LTO_END): Likewise.
21933         * config/i386/winnt.c (saved_debug_info_level): New static variable.
21934         (i386_pe_asm_lto_start): New function.
21935         (i386_pe_asm_lto_end): Likewise.
21937 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
21938             Richard Biener  <rguenther@suse.de>
21940         PR middle-end/84955
21941         * lto-streamer-out.c (output_function): Fix CFG loop state before
21942         streaming out.
21943         * omp-expand.c (expand_oacc_for): Handle calls to internal
21944         functions like regular functions.
21946 2018-04-12  Richard Biener  <rguenther@suse.de>
21948         PR lto/85371
21949         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
21950         for the early LTO debug to properly generate references to it
21951         during DIE emission.  Do not re-use that for the skeleton for
21952         split-dwarf.
21953         (dwarf2out_early_finish): Likewise.
21955 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21957         PR target/85328
21958         * config/i386/sse.md
21959         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
21960         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
21961         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
21962         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
21963         and output is a reg, avoid creating invalid lowpart subreg, but
21964         instead split into a 512-bit move.  Don't split if not AVX512VL,
21965         input is xmm16+ reg and output is a mem.
21966         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
21967         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
21968         xmm16+ reg and output is a mem.
21970 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
21972         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
21973         also for flag_dwarf2_cfi_asm.
21975 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
21977         PR rtl-optimization/85342
21978         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
21979         a bool scalar var inside of the loop instead.  Don't try to update
21980         recog_data.operand after failed apply_change_group.
21982 2018-04-12  Tom de Vries  <tom@codesourcery.com>
21984         PR target/85296
21985         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
21986         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
21987         array with flexible array member as array without given dimension.
21988         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
21989         argument for undefined param to true.
21991 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
21993         PR target/85321
21994         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
21995         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
21996         from PowerPC section.
21997         * config/rs6000/sysv4.opt (mcall-): Improve help text.
21998         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
21999         help text that is too long.
22000         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
22001         help text that is too long.
22002         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
22003         help text that is too long.
22005 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
22007         * config/alpha/alpha.md (stack_probe_internal): Rename
22008         from "probe_stack".  Update all callers.
22010 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
22012         PR rtl-optimization/84566
22013         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
22014         sched_macro_fuse_insns.
22016 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
22018         PR target/84301
22019         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
22020         (compute_block_dependences): ... from here.
22022 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22024         PR tree-optimization/85331
22025         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
22026         from int to HOST_WIDE_INT.
22028 2018-04-11  Martin Jambor  <mjambor@suse.cz>
22030         PR ipa/84149
22031         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
22032         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
22033         not the same as the source val.
22034         (cgraph_edge_brings_value_p): New parameter.
22035         (gather_edges_for_value): Pass destination value to
22036         cgraph_edge_brings_value_p.
22037         (perhaps_add_new_callers): Likewise.
22038         (get_info_about_necessary_edges): Likewise and exclude values brought
22039         only by self-recursive edges.
22040         (create_specialized_node): Redirect only clones of self-calling edges.
22041         (+self_recursive_pass_through_p): New function.
22042         (find_more_scalar_values_for_callers_subset): Use it.
22043         (find_aggregate_values_for_callers_subset): Likewise.
22044         (known_aggs_to_agg_replacement_list): Removed.
22045         (decide_whether_version_node): Re-calculate known constants for all
22046         remaining context clones.
22048 2018-04-11  Richard Biener  <rguenther@suse.de>
22050         PR lto/85339
22051         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
22052         from early DWARF output.
22053         (dwarf2out_early_finish): Output line info unconditionally into
22054         early DWARF and add reference to it.
22056 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22058         PR target/85281
22059         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
22060         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
22061         other than V2DFmode using iptr mode attribute.
22062         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
22064 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
22066         PR rtl-optimization/84659
22067         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
22069 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
22071         PR debug/85302
22072         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
22073         SIZEP is NULL.
22074         (output_loc_list): Pass address of a dummy size variable even in the
22075         locview handling loop.
22076         (index_location_lists): Add comment on why skip_loc_list_entry can't
22077         call size_of_locs.
22079 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22081         PR target/85261
22082         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
22083         into register.
22085 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
22087         PR target/85321
22088         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
22089         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
22090         and -mstring-compare-inline-limit.
22092 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22094         PR target/85287
22095         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
22096         for stack clash protection in a register whenever we need it to be in
22097         a register.
22099 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22101         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
22102         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
22104 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
22106         PR target/85321
22107         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
22108         the help text.
22109         (mlong-double-): Ditto.
22110         * config/rs6000/sysv4.opt (msdata=): Ditto.
22111         (mtls-size=): Ditto.
22113 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22115         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
22116         erroneous entries for
22117         "vector int vec_ldl (int, long int *)", and
22118         "vector unsigned int vec_ldl (int, unsigned long int *)".
22119         Add comments and entries for
22120         "vector bool char vec_ldl (int, bool char *)",
22121         "vector bool short vec_ldl (int, bool short *)",
22122         "vector bool int vec_ldl (int, bool int *)",
22123         "vector bool long long vec_ldl (int, bool long long *)",
22124         "vector pixel vec_ldl (int, pixel *)",
22125         "vector long long vec_ldl (int, long long *)",
22126         "vector unsigned long long vec_ldl (int, unsigned long long *)".
22127         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
22128         type tree bool_long_long_type_node and correct definition of
22129         bool_V2DI_type_node to make reference to this new type tree.
22130         (rs6000_mangle_type): Replace erroneous reference to
22131         bool_long_type_node with bool_long_long_type_node.
22132         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
22133         comments to emphasize sign distinctions for char and int types and
22134         replace RS6000_BTI_bool_long constant with
22135         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
22136         use of RS6000_BTI_pixel.
22137         (bool_long_type_node): Remove this macro definition.
22138         (bool_long_long_type_node): New macro definition
22140 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
22142         PR rtl-optimization/85300
22143         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
22144         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
22145         simplify_unary_operation fails.
22147 2018-04-10  Martin Liska  <mliska@suse.cz>
22149         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
22150         cgraph_edge and ipa_ref.
22152 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
22154         PR target/85177
22155         PR target/85255
22156         * config/i386/sse.md
22157         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
22158         computation of the VEC_MERGE selector from mask.
22159         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
22160         Fix decoding of the VEC_MERGE selector into mask.
22162 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
22164         PR tree-optimization/85286
22165         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
22167 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
22169         * final.c (final_1): Set insn_last_address as well as
22170         insn_current_address.
22172 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22174         PR target/85173
22175         * explow.c (emit_stack_probe): Call validize_mem on memory location
22176         before passing it to gen_probe_stack.  Create address operand and
22177         legitimize it for the probe_stack_address case.
22179 2018-04-09  Jan Hubicka  <jh@suse.cz>
22181         PR lto/85078
22182         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
22183         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
22184         * tree.c (free_lang_data_in_type): Fix handling of binfos;
22185         walk basetypes.
22186         (free_lang_data): Rebuild type inheritance graph.
22188 2018-04-09  Martin Sebor  <msebor@redhat.com>
22190         * invoke.texi (-finline-small-functions): Mention other optimization
22191         options.
22192         (-findirect-inlining, -fpartial-inlining): Same.
22193         (-finline-functions-called-once): Same.
22194         (-freorder-blocks-and-partition): Same.
22196 2018-04-09  Jan Hubicka  <jh@suse.cz>
22198         PR rtl/84058
22199         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
22200         jumps; choose last target that matches the criteria (i.e.
22201         no partition changes for non-crossing jumps).
22202         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
22203         support for redirecting crossing jumps to non-crossing.
22205 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
22207         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
22208         also for naked functions.
22210 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
22212         * config/arc/arc.md (add_shift): New pattern.
22213         (add_shift2): Likewise.
22214         (sub_shift): Likewise.
22215         (sub_shift_cmp0_noout): Likewise.
22216         (compare_si_ashiftsi): Likewise.
22217         (xbfu_cmp0_noout): New combine pattern.
22218         (xbfu_cmp0"): Likewise.
22219         (movsi_set_cc_insn): Place the predicable variant first.
22220         (commutative_binary_cmp0_noout): Remove clobber.
22221         (commutative_binary_cmp0): New pattern.
22222         (noncommutative_binary_cmp0): Likewise.
22223         (noncommutative_binary_cmp0_noout): Likewise.
22224         (noncommutative_binary_comparison_result_used): Removed.
22225         (rsub_cmp0): New pattern.
22226         (rsub_cmp0_noout): Likewise.
22227         (extzvsi): Changed, keep only meaningful variants.
22228         (SQH, SEZ): New iterators.
22229         (SQH_postfix): New mode attribute.
22230         (SEZ_prefix): New code attribute.
22231         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
22232         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
22233         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
22234         of numerical value.
22235         (noncommutative_operator): Check the availability of barrel
22236         shifter option.
22238 2018-04-09  Richard Biener  <rguenther@suse.de>
22240         PR tree-optimization/85284
22241         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
22242         Only use the niter constraining form of simple_iv when the exit
22243         is always executed.
22245 2018-04-09  Tom de Vries  <tom@codesourcery.com>
22247         PR target/84041
22248         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
22249         (define_expand "*memory_barrier"): New define_expand.
22250         (define_insn "memory_barrier"): New insn.
22252 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22254         PR rtl-optimization/80463
22255         PR rtl-optimization/83972
22256         PR rtl-optimization/83480
22258         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
22259         correct producer for the insn.
22260         (tidy_control_flow): Fixup seqnos in case of debug insns.
22262 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22264         PR rtl-optimization/83913
22266         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
22267         different sched-times when merging exprs.
22269 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22271         PR rtl-optimization/83962
22273         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
22274         tidy_fallthru_edge and tidy_control_flow.
22276 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
22278         PR rtl-optimization/83530
22280         * sel-sched.c (force_next_insn): New global variable.
22281         (remove_insn_for_debug): When force_next_insn is true, also leave only
22282         next insn in the ready list.
22283         (sel_sched_region): When the region wasn't scheduled, make another pass
22284         over it with force_next_insn set to 1.
22286 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
22288         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
22289         into tm_file.
22290         * config/nds32/constants.md (unspec_volatile_element): Add enum values
22291         for interrupt control.
22292         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22293         functions for interrupt control.
22294         * config/nds32/nds32-intrinsic.md: Likewise.
22295         * config/nds32/nds32_intrinsic.h: Likewise.
22296         * config/nds32/nds32.h (nds32_builtins): Likewise.
22298 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
22300         * config/nds32/nds32.c (nds32_init_machine_status,
22301         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
22302         strict_aligned_p field.
22303         (nds32_expand_to_rtl_hook): New function.
22304         (TARGET_EXPAND_TO_RTL_HOOK): Define.
22305         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
22307 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22308             Chung-Ju Wu  <jasonwucj@gmail.com>
22310         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
22311         * config/nds32/nds32-n7.md: New file.
22312         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
22313         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
22314         pipeline.
22315         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
22316         * config/nds32/nds32.md (pipeline_model): Add n7.
22317         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
22318         * config/nds32/pipelines.md: Include n7 settings.
22320 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22321             Chung-Ju Wu  <jasonwucj@gmail.com>
22323         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
22324         * config/nds32/nds32-e8.md: New file.
22325         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
22326         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
22327         pipeline.
22328         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
22329         * config/nds32/nds32.md (pipeline_model): Add e8.
22330         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
22331         * config/nds32/pipelines.md: Include e8 settings.
22333 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22334             Chung-Ju Wu  <jasonwucj@gmail.com>
22336         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
22337         * config/nds32/nds32-n8.md: New file.
22338         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
22339         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
22340         pipeline.
22341         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
22342         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
22343         * config/nds32/nds32.md (pipeline_model): Add n8.
22344         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
22345         * config/nds32/pipelines.md: Include n8 settings.
22347 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
22348             Chung-Ju Wu  <jasonwucj@gmail.com>
22350         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
22351         * config/nds32/nds32-n9-2r1w.md: New file.
22352         * config/nds32/nds32-n9-3r2w.md: New file.
22353         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
22354         nds32_register_ports): New or modify for cpu n9.
22355         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
22356         pipeline.
22357         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
22358         * config/nds32/nds32-utils.c: New file.
22359         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
22360         TARGET_MUL_SLOW): Define.
22361         * config/nds32/nds32.md (pipeline_model): New attribute.
22362         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
22363         New options that support cpu n9.
22364         * config/nds32/pipelines.md: Include n9 settings.
22365         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
22367 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
22369         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
22370         information if necessary.
22371         (output_cond_branch_compare_zero): Likewise.
22372         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
22373         (nds32_target_alignment): Refine for alignment.
22374         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
22375         (FUNCTION_BOUNDARY): Modify.
22376         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
22377         align case.
22378         * config/nds32/nds32.opt (malways-align, malign-functions): New.
22380 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
22382         * config/nds32/constants.md (unspec_volatile_element): Add values for
22383         TLB operation and data prefetch.
22384         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
22385         functions for TLB operation and data prefetch.
22386         * config/nds32/nds32-intrinsic.md: Likewise.
22387         * config/nds32/nds32_intrinsic.h: Likewise.
22388         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
22389         (nds32_print_operand): Likewise.
22390         * config/nds32/nds32.h (nds32_builtins): Likewise.
22392 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22393         Andrew Pinski <pinsika@gcc.gnu.org>
22395         PR middle-end/82976
22396         * match.pd: Use constant_boolean_node of correct type instead of
22397         boolean_true_node or boolean_false_node for simplifying
22398         pointer comparisons to zero.
22400 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
22402         PR tree-optimization/80021
22403         * tree.c (verify_type_variant): Make error call in verify_variant_match
22404         translatable and remove final full stop.
22406 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22408         * config/nds32/constants.md (unspec_volatile_element): Add
22409         UNSPEC_VOLATILE_EH_RETURN.
22410         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
22411         nds32_output_stack_pop): Support dwarf exception handling process.
22412         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
22413         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
22414         exception handling process.
22415         (nds32_compute_stack_frame): Likewise.
22416         (nds32_return_addr_rtx): Likewise.
22417         (nds32_initial_elimination_offset): Likewise.
22418         (nds32_expand_prologue): Likewise.
22419         (nds32_expand_epilogue): Likewise.
22420         (nds32_dynamic_chain_address): New function.
22421         * config/nds32/nds32.h (machine_function): Add fields for dwarf
22422         exception handling.
22423         (DYNAMIC_CHAIN_ADDRESS): Define.
22424         (EH_RETURN_DATA_REGNO): Define.
22425         (EH_RETURN_STACKADJ_RTX): Define.
22426         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
22427         patterns for dwarf exception handling.
22429 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22431         * config/nds32/nds32.h: Clean up obsolete macros.
22433 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22435         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22436         Add enum values for particular instructions.
22437         * config/nds32/nds32-intrinsic.c: Implementation of expanding
22438         particular intrinsic functions.
22439         * config/nds32/nds32-intrinsic.md: Likewise.
22440         * config/nds32/nds32_intrinsic.h: Likewise.
22441         * config/nds32/nds32.h (nds32_builtins): Likewise.
22442         * config/nds32/nds32.md (type): Add pbsad and pbsada.
22443         (btst, ave): New patterns for particular instructions.
22445 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22447         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
22448         Add enum values for atomic load/store and memory sync.
22449         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
22450         and memory sync.
22451         * config/nds32/nds32-intrinsic.md: Likewise.
22452         * config/nds32/nds32_intrinsic.h: Likewise.
22453         * config/nds32/nds32.h (nds32_builtins): Likewise.
22455 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
22457         PR tree-optimization/85257
22458         * fold-const.c (native_encode_vector): If not all elts could fit
22459         and off is -1, return 0 rather than offset.
22460         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
22461         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
22462         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
22463         adjust buffer in native_interpret_expr call.
22465 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22467         * config/nds32/constants.md (unspec_volatile_element): Add cache
22468         control enum values.
22469         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
22470         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
22471         * config/nds32/nds32.c (nds32_cctl_names): New.
22472         (nds32_print_operand): Handle cache control register names.
22473         * config/nds32/nds32.h (nds32_builtins): New enum values.
22474         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
22475         macros.
22476         * config/nds32/nds32.md (type): Add mmu.
22477         * config/nds32/pipelines.md (simple_insn): Add mmu.
22479 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22481         * config/nds32/nds32.md (type): Remove call.
22482         * config/nds32/pipelines.md (simple_insn): Likewise.
22484 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22486         * config/nds32/constants.md (unspec_volatile_element): Add
22487         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
22488         UNSPEC_VOLATILE_FMFCFG.
22489         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
22490         description for fmfcfg and fmfcsr.
22491         (bdesc_1arg): Add fmtcsr.
22492         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
22493         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
22494         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
22495         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
22496         unspec_fmfcfg): New patterns.
22497         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
22498         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
22499         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
22500         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
22501         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
22502         __nds32__fmfcfg): Define.
22504 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
22506         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
22507         intrinsic register names.
22508         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
22509         intrinsic register enum values and macros.
22511 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
22513         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
22514         for load/store addressing form.
22515         (nds32_print_operand_address): Likewise.
22517 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
22519         PR target/85196
22520         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
22521         based on LABEL_REF.  Remove useless assertion.
22522         (pic_address_needs_scratch): Fix formatting.
22523         (sparc_legitimize_pic_address): Minor tweaks.
22524         (sparc_delegitimize_address): Adjust assertion accordingly.
22525         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
22526         into symbolic_operand.
22527         (movsi_high_pic_label_ref): Likewise.
22528         (movsi_lo_sum_pic_label_ref): Likewise.
22529         (movdi_pic_label_ref): Likewise.
22530         (movdi_high_pic_label_ref): Likewise.
22531         (movdi_lo_sum_pic_label_ref): Likewise.
22533 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
22535         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
22536         custom LIB_SPEC setup.
22538 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
22539             Kito Cheng  <kito.cheng@gmail.com>
22541         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
22542         * config/riscv/freebsd.h: New.
22544 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22546         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
22547         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
22548         file.
22550 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22551             Kito Cheng  <kito.cheng@gmail.com>
22553         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
22554         nds32_output_call, nds32_symbol_binds_local_p): New functions.
22555         * config/nds32/nds32-protos.h (nds32_output_call,
22556         nds32_output_return): Declare.
22557         * config/nds32/nds32.md: Refine all the call and return patterns.
22559 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22561         PR debug/85252
22562         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
22563         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
22565         PR rtl-optimization/84872
22566         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
22567         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
22568         EDGE_CROSSING edge.
22570 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
22572         * expr.c (copy_blkmode_to_reg): Revert 254862.
22573         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
22575 2018-04-06  Richard Biener  <rguenther@suse.de>
22577         PR middle-end/85244
22578         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
22579         after seeing a component reference with an adjacent field.  Treat
22580         refs to arrays at struct end of external decls similar to
22581         refs to unconstrained commons.
22583 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
22585         PR sanitizer/85213
22586         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
22587         look through SAVE_EXPRs with non-side-effects argument.  Adjust
22588         recursive calls.
22589         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
22590         save_p here.
22592 2018-04-06  Richard Biener  <rguenther@suse.de>
22594         PR middle-end/85180
22595         * alias.c (find_base_term): New wrapper around find_base_term
22596         unwinding CSELIB_VAL_PTR changes.
22597         (find_base_term): Do not restore CSELIB_VAL_PTR during the
22598         recursion.
22600 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22602         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
22603         instructions.
22604         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
22605         constant definitions.
22606         ("nop"): lr 0,0 -> nopr r0
22607         ("nop_lr0", "nop_lr1"): New insn definitions.
22609 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
22611         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
22612         NDS32_V3PUSH_AVAILABLE_P macro.
22614 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
22615             Chung-Ju Wu  <jasonwucj@gmail.com>
22617         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
22618         (nds32*-*-*): Add float and fpu_config into supported_defaults.
22619         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
22620         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
22621         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
22622         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
22623         * config/nds32/constraints.md: New constraints and checking for hard
22624         float configuration.
22625         * config/nds32/iterators.md: New mode iterator and attribute for hard
22626         float configuration.
22627         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
22628         patterns.
22629         * config/nds32/nds32-fpu.md: New file.
22630         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
22631         deal with hard float code generation.
22632         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
22633         ARCH_V3S.
22634         (abi_type, float_reg_number): New enum type.
22635         * config/nds32/nds32-predicates.c: New predicates for hard float.
22636         * config/nds32/nds32-protos.h: Declare functions for hard float.
22637         * config/nds32/nds32.c: Implementation for hard float configuration.
22638         * config/nds32/nds32.h: Definitions for hard float configuration.
22639         * config/nds32/nds32.md: Include hard float machine description and
22640         modify patterns for hard float configuration.
22641         * config/nds32/nds32.opt: New options for hard float configuration.
22642         * config/nds32/predicates.md: New predicates for hard float
22643         configuration.
22645 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22647         * common/config/nds32/nds32-common.c
22648         (nds32_option_optimization_table): Enable -mreleax-hint by default.
22650 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
22652         PR middle-end/85195
22653         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
22654         CONSTRUCTOR_ELT (ctor, ...)->value.
22656 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
22658         PR target/85193
22659         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
22661 2018-04-05  Tom de Vries  <tom@codesourcery.com>
22663         PR target/85204
22664         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
22665         cond jump.
22667 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22668             Kito Cheng  <kito.cheng@gmail.com>
22670         * config/nds32/constraints.md (U33): Fine-tune checking condition.
22671         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
22672         * config/nds32/nds32.h (nds32_16bit_address_type): Add
22673         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
22675 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
22676             Kito Cheng  <kito.cheng@gmail.com>
22678         * config/nds32/constraints.md (Ufe): New memory constraint.
22679         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
22680         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
22681         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
22682         operands.
22683         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
22684         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
22686 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22688         * config/nds32/nds32.md: Use optimize_size in the condition for
22689         alu-shift instructions.
22691 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22693         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
22695 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22697         * config/nds32/nds32.md (negsi2): Refine pattern.
22699 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22700             Chung-Ju Wu  <jasonwucj@gmail.com>
22702         * config/nds32/iterators.md (shift_rotate): New code iterator.
22703         (shift): New code attribute.
22704         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
22705         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
22706         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
22707         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
22708         bit-wise operations.
22709         (andsi3, *andsi3): Ditto.
22710         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
22711         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
22712         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
22713         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
22714         nds32_ior_operand, nds32_xor_operand): New predicates.
22716 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22718         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
22719         (addsi3, subsi3): ... this.
22721 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22723         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
22725 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22727         * config/nds32/nds32.md: Adjust indention.
22729 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
22731         * config/nds32/nds32.md (feature): New attribute.
22733 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
22735         * config/nds32/nds32.md (subtype): New attribute.
22737 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
22739         PR target/85203
22740         * config/arm/arm-builtins.c (arm_expand_builtin): Change
22741         expansion to perform a bitwise AND of the argument followed by a
22742         boolean negation of the result.
22744 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
22746         PR rtl-optimization/84878
22747         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
22748         the basic block.  Assert the use reference is not artificial and that
22749         it has an associated insn.
22751 2018-04-04  Michael Matz  <matz@suse.de>
22753         * builtins.c (compute_objsize): Pass correct operand
22754         to array_at_struct_end_p.
22756 2018-04-04  Richard Biener  <rguenther@suse.de>
22758         PR lto/85176
22759         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
22760         from contexts for DINFO_LEVEL_TERSE and below.
22762 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22764         * config/nds32/nds32-doubleword.md (move_<mode>): Require
22765         resiter_operand condition.
22766         * config/nds32/nds32.md (*move<mode>): Ditto.
22768 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22769             Monk Chiang  <sh.chiang04@gmail.com>
22771         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
22773 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22775         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
22777 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22778             Kito Cheng  <kito.cheng@gmail.com>
22780         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
22781         nds32_cond_code_str, output_cond_branch,
22782         output_cond_branch_compare_zero, nds32_expand_cbranch,
22783         nds32_expand_cstore, nds32_expand_movcc,
22784         nds32_output_cbranchsi4_equality_zero,
22785         nds32_output_cbranchsi4_equality_reg,
22786         nds32_output_cbranchsi4_equality_reg_or_const_int,
22787         nds32_output_cbranchsi4_greater_less_zero: New functions.
22788         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
22789         nds32_expand_cstore, nds32_expand_movcc,
22790         nds32_output_cbranchsi4_equality_zero,
22791         nds32_output_cbranchsi4_equality_reg,
22792         nds32_output_cbranchsi4_equality_reg_or_const_int,
22793         nds32_output_cbranchsi4_greater_less_zero): Declare.
22794         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
22795         nds32_rimm11s_operand): New predicates.
22796         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
22797         * config/nds32/nds32.md: Rewrite all the branch and conditional move
22798         patterns.
22800 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
22802         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
22803         * config/nds32/nds32.md: Ditto.
22804         * config/nds32/pipelines.md: Ditto.
22806 2018-04-04  Richard Biener  <rguenther@suse.de>
22808         PR tree-optimization/85168
22809         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
22810         propagating abnormals.
22812 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22814         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
22816 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
22817             Kito Cheng  <kito.cheng@gmail.com>
22819         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
22820         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
22821         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
22822         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
22823         * config/nds32/nds32.md (sibcall_internal): New.
22824         (sibcall_register): Remove.
22825         (sibcall_immediate): Remove.
22826         (sibcall_value_internal): New.
22827         (sibcall_value_register): Remove.
22828         (sibcall_value_immediate): Remove.
22829         * config/nds32/predicates.md (nds32_general_register_operand): New.
22830         (nds32_call_address_operand): New.
22832 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22834         PR rtl-optimization/85167
22835         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
22836         bb_defs if *split_p, instead preinitialize it to NULL.
22838         PR tree-optimization/85156
22839         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
22840         evaluating the argument multiple times.
22842 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
22844         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
22845         than vector.
22846         (_mm_cvtpd_ps): Likewise.
22847         (_mm_cvttpd_epi32): Likewise.
22848         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
22849         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
22850         vector, pixel, and bool following altivec.h include.
22852 2018-04-03  Martin Sebor  <msebor@redhat.com>
22854         * doc/extend.texi (Common Function Attributes): Clarify.
22855         (const attribute): Likewise.
22856         (pure attribute): Likewise.
22858 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
22860         PR target/85169
22861         * config/i386/i386.c (ix86_expand_vector_set): Use
22862         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
22864 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
22866         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
22867         instructions when changing rounding bits to preserve precision bits
22868         in the x87 control word.
22870 2018-04-03  Martin Liska  <mliska@suse.cz>
22872         PR tree-optimization/82491
22873         * rtl.h (strip_offset_and_add): Replace += suboffset with
22874         poly_uint64 () + suboffset.
22876 2018-03-29  Martin Liska  <mliska@suse.cz>
22877             Martin Jambor  <mjambor@suse.cz>
22879         PR ipa/84947
22880         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
22881         param_type is not an integral or pointer type.
22883 2018-04-03  Richard Biener  <rguenther@suse.de>
22885         * sese.h (recompute_all_dominators): Remove.
22887 2018-04-02  Martin Sebor  <msebor@redhat.com>
22889         * doc/invoke.texi (-Wrestrict): Fix typos.
22891 2018-04-02  Jim Wilson  <jimw@sifive.com>
22893         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
22894         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
22895         (<optab>di3, <optab>si3_extend): Likewise.
22896         (<optab>si3_mask, <optab>si3_mask_1): New.
22897         (<optab>di3_mask, <optab>di3_mask_1): New.
22898         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
22899         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
22900         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
22902 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
22904         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
22905         example.
22907 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
22909         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
22910         (nds32_canonicalize_comparison): New function.
22912 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22913             Kito Cheng  <kito.cheng@gmail.com>
22914             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
22916         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
22917         * config/nds32/constants.md (unspec_volatile_element): Add
22918         UNSPEC_VOLATILE_RELAX_GROUP.
22919         * config/nds32/nds32-relax-opt.c: New file.
22920         * config/nds32/nds32-predicates.c
22921         (nds32_symbol_load_store_p): New function.
22922         * config/nds32/nds32-protos.h
22923         (nds32_symbol_load_store_p): Declare function.
22924         (make_pass_nds32_relax_opt): Declare new rtl pass function.
22925         * config/nds32/nds32.c
22926         (nds32_register_pass): New function to register pass.
22927         (nds32_register_passes): New function to register passes.
22928         * config/nds32/nds32.md (relax_group): New pattern.
22929         * config/nds32/nds32.opt (mrelax-hint): New option.
22930         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
22932 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
22934         * config/nds32/t-nds32: Modify files dependency.
22936 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22938         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
22939         (PROFILE_HOOK): Define its implementation.
22941 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
22943         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
22944         type and 32-bit size.
22946 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
22948         PR middle-end/85090
22949         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
22950         (V_128_256): New mode iterator.
22951         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
22952         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
22953         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
22954         of V.
22955         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
22956         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
22958 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
22960         PR target/83315
22961         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
22962         NaN inputs correctly.
22964 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
22966         PR target/80546
22967         * config/rs6000/vsx.md (??r): New mode attribute.
22968         (*vsx_mov<mode>_64bit): Use it.
22969         (*vsx_mov<mode>_32bit): Likewise.
22971 2018-03-30  Martin Sebor  <msebor@redhat.com>
22973         PR tree-optimization/84818
22974         * builtins.c (check_access): Use warning_n.
22976 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
22978         PR target/83822
22979         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
22980         condition.
22981         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
22982         condition.
22984 2018-03-30  Julia Koval  <julia.koval@intel.com>
22986         PR target/84413
22987         * x86-tune.def (movx, partial_reg_dependency): Enable for
22988         m_SKYLAKE_AVX512.
22990 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
22992         PR inline-asm/84985
22993         * lra-constraints.c (process_alt_operands): Move setting
22994         this_alternative_matches below.
22996 2018-03-29  Martin Liska  <mliska@suse.cz>
22998         PR lto/84995.
22999         * doc/invoke.texi: Document how LTO works with debug info.
23000         Describe auto-load support of binutils.  Mention 'x86-64'
23001         as valid option value of -march option.
23003 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
23005         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
23007         PR c/85094
23008         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
23009         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
23010         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
23011         checking.
23013 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
23015         PR target/84912
23016         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
23017         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
23018         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
23019         for RS6000_BTM_POWERPC64.
23020         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
23021         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
23022         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
23023         definition.
23024         (DIVDE): Use it.
23025         (DIVDEU): Likewise.
23027 2018-03-28  Carl Love  <cel@us.ibm.com>
23029         Revert
23030         2017-09-27  Carl Love  <cel@us.ibm.com>
23032         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
23033         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
23034         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
23035         fctiw instruction.
23037 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23039         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
23040         instead of __vector bool.
23041         (_mm_max_pu8): Likewise.
23042         (_mm_min_pi16): Likewise.
23044 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
23046         PR target/84912
23047         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
23048         (DIVWEUO): Likewise.
23049         (DIVDEO): Likewise.
23050         (DIVDEUO): Likewise.
23051         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
23052         DIVWEUO and DIVDEUO.
23053         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
23054         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
23055         (div_extend): Likewise.
23056         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
23057         builtin function.
23058         (__builtin_divweuo): Likewise.
23059         (__builtin_divdeo): Likewise.
23060         (__builtin_divdeuo): Likewise.
23062 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
23064         PR target/85095
23065         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
23066         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
23068         PR tree-optimization/82004
23069         * gimple-match-head.c (optimize_pow_to_exp): New function.
23070         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
23071         Don't fold to exp if optimize_pow_to_exp is false.
23073 2018-03-28  Martin Liska  <mliska@suse.cz>
23075         PR other/84819
23076         * calls.c (initialize_argument_information): Fix trailing space.
23077         * common.opt: Fix typo and provide better explanation for
23078         -fsanitize-coverage option.
23079         * config/i386/i386.opt: Fix typo.
23081 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
23082             Martin Liska  <mliska@suse.cz>
23084         PR sanitizer/85081
23085         * gimplify.c (asan_poison_variable): Don't do the check for
23086         gimplify_omp_ctxp here.
23087         (gimplify_decl_expr): Do it here.
23088         (gimplify_target_expr): Likewise.
23090 2018-03-28  Martin Liska  <mliska@suse.cz>
23092         PR target/84988
23093         * config/i386/i386.c (ix86_function_arg_advance): Do not call
23094         chkp_type_bounds_count if MPX is not enabled.
23096 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
23098         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
23100 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
23102         PR target/84914
23103         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
23104         function to create the function decl for complex long double
23105         multiply and divide for -mabi=ieeelongdouble.
23106         (init_float128_ieee): Call it.
23108 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
23110         PR target/85044
23111         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
23112         -fcf-protection=branch -mibt.
23113         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
23115 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
23117         PR target/81863
23118         * config/arm/arm.c (arm_valid_symbolic_address): Handle
23119         arm_word_relocations.
23121 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
23123         PR target/85056
23124         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
23125         extern array declarations.
23127 2018-03-27  Richard Biener  <rguenther@suse.de>
23129         PR middle-end/84067
23130         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
23131         explicit single_use checks.
23133 2018-03-27  Richard Biener  <rguenther@suse.de>
23135         PR tree-optimization/85082
23136         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
23137         Valueize the VUSE.
23139 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
23141         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
23142         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
23143         Turn on fasynchronous-unwind-tables and funwind-tables.
23145 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
23147         PR target/85073
23148         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
23149         (*bmi_blsr_<mode>_ccz): Ditto.
23151 2018-03-26  Tom de Vries  <tom@codesourcery.com>
23153         PR tree-optimization/85063
23154         * omp-general.c (offloading_function_p): New function.  Factor out
23155         of ...
23156         * omp-offload.c (pass_omp_target_link::gate): ... here.
23157         * omp-general.h (offloading_function_p): Declare.
23158         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
23159         with attribute omp declare target for offloading functions.
23161 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
23163         PR tree-optimization/84005
23164         * tree-data-ref.h (get_base_for_alignment): Declare.
23165         * tree-data-ref.c (get_base_for_alignment_1): New function.
23166         (get_base_for_alignment): Likewise.
23167         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
23168         get_base_for_alignment to find a suitable base object, instead
23169         of always using drb->base_address.
23171 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
23173         PR inline-asm/85022
23174         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
23175         known size by default.
23177 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
23179         PR inline-asm/85030
23180         * lra-constraints.c (process_alt_operands): Don't match BLKmode
23181         and non BLKmode operands.
23183 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23185         PR target/85026
23186         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
23187         Clean up attributes.
23189 2018-03-23  Richard Biener  <rguenther@suse.de>
23191         PR debug/85020
23192         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
23193         we are going to emit early debug for LTO.
23195 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
23197         PR inline-asm/85034
23198         * function.c (match_asm_constraints_1): Don't optimize if input
23199         doesn't satisfy general_operand predicate for output's mode.
23201         PR inline-asm/85022
23202         * alias.c (write_dependence_p): Don't require for x_canonicalized
23203         non-VOIDmode if x has VOIDmode.
23205         PR sanitizer/85029
23206         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
23207         just don't try to optimize it rather than assert it never happens.
23209 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23211         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
23212         macro expansions for definition of ST_INTERNAL_<mode> and
23213         LD_INTERNAL_<mode> builtins.
23214         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
23215         Remove prototype.
23216         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
23217         function.
23218         (altivec_expand_st_builtin): Likewise.
23219         (altivec_expand_builtin): Remove calls to deleted functions.
23220         (rs6000_address_for_altivec): Delete this function.
23221         * config/rs6000/vector.md: Remove expands for
23222         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
23224 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
23226         PR target/84826
23227         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
23228         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
23229         re-computing once computed.
23230         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
23231         (arm_init_machine_status): Initialize
23232         machine->static_chain_stack_bytes.
23234 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
23236         PR target/84760
23237         * doc/extend.texi: Add four new prototypes for vec_ld.
23238         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
23239         definitions for more logical presentation.
23240         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
23241         entries for V1TI variants of __builtin_altivec_ld builtin.
23242         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
23243         handling of V1TI variant of LVX icode pattern.
23244         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
23245         (rs6000_gimple_fold_builtin): Likewise.
23246         (altivec_init_builtins): Add code to define
23247         __builtin_altivec_lvx_v1ti function.
23249 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
23251         PR inline-asm/84941
23252         * function.c (match_asm_constraints_1): Don't do the optimization
23253         if input isn't a REG, SUBREG, MEM or constant.
23255 2018-03-22  Tom de Vries  <tom@codesourcery.com>
23257         PR tree-optimization/84956
23258         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
23259         bb_has_abnormal_pred.
23261 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
23263         PR sanitizer/85018
23264         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
23265         DECL_INITIAL (decl) to decl at the end.
23266         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
23267         adjust the comment.
23269 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
23271         * doc/extend.texi (__builtin_tgmath): Document when complex
23272         integer types are treated as _Complex _Float64.
23274 2018-03-21  Tom de Vries  <tom@codesourcery.com>
23276         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
23278 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23280         PR tree-optimization/84960
23281         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
23282         if it is ENTRY block, move them into single succ of ENTRY in that case.
23284 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
23286         PR tree-optimization/84811
23287         * poly-int.h (poly_span_traits): Remove the T3 parameter and
23288         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
23289         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
23290         (known_subrange_p): Update accordingly.  Cast each value involved
23291         in the size comparison, rather than casting the result of the
23292         subtraction.
23294 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23296         PR tree-optimization/84982
23297         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
23298         by flipping the least significant bit rather than all bits from
23299         bitpos to bitpos + bitsize - 1.
23301 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23303         * doc/extend.texi (Deprecated Features): Remove mention of
23304         long-deleted deprecations.
23306 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
23308         PR jit/84288
23309         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
23310         * configure: Regenerate.
23312 2018-03-21  Tom de Vries  <tom@codesourcery.com>
23314         PR tree-optimization/83126
23315         * tree-parloops.c (num_phis): New function.
23316         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
23318 2018-03-21  Nathan Sidwell  <nathan@acm.org>
23320         * doc/extend.texi (Deprecated Features): Update deprecated flags,
23321         mention anon-struct/union members and trailing attributes.
23323 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
23325         PR tree-optimization/84969
23326         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
23327         builtin memset partitions if they set different rhs values.
23329 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
23331         PR rtl-optimization/84989
23332         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
23333         VEC_DUPLICATE with scalar result mode.
23335 2018-03-21  Martin Liska  <mliska@suse.cz>
23337         PR ipa/84963
23338         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
23339         not intended return statement.
23341 2018-03-21  Martin Liska  <mliska@suse.cz>
23343         PR target/84988
23344         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
23345         (chkp_find_bound_slots_1): Limit number of iterations.
23347 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
23349         PR target/84838
23350         * Minor grammar fixes for x86 options.
23352 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23354         PR debug/84875
23355         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
23356         holding REG_CFA_RESTORE notes, instead turn them into a USE.
23358 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
23360         PR target/83789
23361         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
23362         (altivec_lvx_<mode>_1op): Likewise.
23363         (altivec_stvx_<mode>_2op): Likewise.
23364         (altivec_stvx_<mode>_1op): Likewise.
23365         (altivec_lvx_<VM2:mode>): New define_expand.
23366         (altivec_stvx_<VM2:mode>): Likewise.
23367         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
23368         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23369         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
23370         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
23371         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
23372         (rs6000_gen_lvx): Likewise.
23373         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
23374         (altivec_expand_stv_builtin): Likewise.
23375         (altivec_expand_builtin): Likewise.
23376         * config/rs6000/vector.md: Likewise.
23378 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23380         PR target/82518
23381         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
23382         BYTES_BIG_ENDIAN.
23384 2018-03-20  Richard Biener  <rguenther@suse.de>
23386         PR target/84986
23387         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
23388         sign-conversions as zero, fall back to standard scalar_stmt
23389         cost for the rest.
23391 2018-03-20  Martin Liska  <mliska@suse.cz>
23393         PR ipa/84825
23394         * predict.c (rebuild_frequencies): Handle case when we have
23395         PROFILE_ABSENT, but flag_guess_branch_prob is false.
23397 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23399         PR target/84990
23400         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
23401         flag_section_anchors.
23402         * varasm.c (use_blocks_for_decl_p): Remove hack for
23403         dw2_force_const_mem.
23405         PR target/84845
23406         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
23407         to ...
23408         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
23409         be created, use lowpart_subreg of operands[0] rather than operands[0]
23410         itself.
23411         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
23412         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
23413         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
23414         and n constraint instead of aarch64_shift_imm_di and Usd.
23415         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
23416         (*aarch64_<optab>_reg_minus<mode>3): ... this.
23418 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
23420         PR target/82989
23421         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
23422         to favor GPR over NEON registers.
23423         (<shift>di3_neon): Likewise.
23425 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23427         PR target/84952
23428         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
23429         (nvptx_process_pars): Emit bar.sync asap and alap.
23431 2018-03-20  Tom de Vries  <tom@codesourcery.com>
23433         PR target/84954
23434         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
23435         seen_label if seen_label is already set.
23437 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
23439         PR target/84945
23440         * config/i386/i386.c (fold_builtin_cpu): For features above 31
23441         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
23442         Use 1U instead of 1.  Formatting fixes.
23444         PR c/84953
23445         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
23446         instead of TREE_TYPE (s1) for the return value.
23448 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
23450         PR tree-optimization/84946
23451         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
23452         bitsize + bitsize in poly_uint64 rather than poly_int64.
23454         PR sanitizer/78651
23455         * dwarf2asm.c: Include fold-const.c.
23456         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
23457         of decl rather than decl itself.
23459         PR rtl-optimization/84643
23460         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
23462 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
23464         PR sanitizer/78651
23465         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
23466         calling assemble_variable.
23468 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
23470         PR target/81647
23471         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
23472         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
23474 2018-03-19  Jim Wilson  <jimw@sifive.com>
23476         PR bootstrap/84856
23477         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
23478         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
23479         (riscv_first_stack_step): Likewise.
23480         (riscv_option_override): Use STACK_BOUNDARY instead of
23481         MIN_STACK_BOUNDARY.
23482         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
23483         MIN_STACK_BOUNDARY.
23484         (BIGGEST_ALIGNMENT): Set to 128.
23485         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
23486         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
23487         STACK_BOUNDARY.
23489 2018-03-19  Richard Biener  <rguenther@suse.de>
23491         PR tree-optimization/84933
23492         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
23493         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
23495 2018-03-19  Richard Biener  <rguenther@suse.de>
23497         PR tree-optimization/84859
23498         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
23499         (cond_if_else_store_replacement): Perform sinking operation on
23500         single-store BBs regardless of MAX_STORES_TO_SINK setting.
23501         Generalize what a BB with a single eligible store is.
23503 2018-03-19  Richard Biener  <rguenther@suse.de>
23505         PR tree-optimization/84929
23506         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
23507         chrec_is_positive against non-chrec arg.
23509 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
23511         PR target/84711
23512         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
23514 2018-03-18  Martin Liska  <mliska@suse.cz>
23516         PR rtl-optimization/84635
23517         * regrename.c (build_def_use): Use matches_mode only when
23518         matches >= 0.
23520 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
23522         PR tree-optimization/84913
23523         * tree-vect-loop.c (vectorizable_reduction): Don't try to
23524         vectorize chains of COND_EXPRs.
23526 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23528         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
23530 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23532         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
23534 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
23536         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
23538 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23539             Kito Cheng  <kito.cheng@gmail.com>
23541         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
23542         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
23543         (nds32_adjust_reg_alloc_order): New function.
23544         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
23546 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
23548         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
23549         nds32_print_operand, nds32_print_operand_address): Use
23550         HOST_WIDE_INT_PRINT_DEC instead.
23552 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
23554         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
23556 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
23558         PR target/84902
23559         * config/i386/i386.c (initial_ix86_tune_features,
23560         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
23561         unsigned long long.
23562         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
23563         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
23564         rather than 1u << ix86_tune.  Formatting fix.
23565         (ix86_option_override_internal): Change ix86_arch_mask from
23566         unsigned int to unsigned HOST_WIDE_INT, initialize to
23567         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
23568         (ix86_function_specific_restore): Likewise.
23570 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23572         PR target/84899
23573         * postreload.c (reload_combine_recognize_pattern): Perform
23574         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
23575         truncate_int_for_mode the result for the destination's mode.
23577         PR c/84909
23578         * hsa-gen.c (mem_type_for_type): Fix comment typo.
23579         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
23580         Likewise.
23581         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
23582         Likewise.
23584 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
23586         PR target/84876
23587         * lra-assigns.c (lra_split_hard_reg_for): Don't use
23588         regno_allocno_class_array and sorted_pseudos.
23589         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
23590         insns where regno is used.
23592 2018-03-16  Martin Liska  <mliska@suse.cz>
23594         PR ipa/84833
23595         * multiple_target.c (create_dispatcher_calls): Redirect
23596         reference in the symbol table.
23598 2018-03-16  Martin Liska  <mliska@suse.cz>
23600         PR ipa/84722
23601         * multiple_target.c (create_dispatcher_calls): Redirect also
23602         an alias.
23604 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23606         PR c++/79937
23607         PR c++/82410
23608         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
23609         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
23610         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
23612 2018-03-16  Julia Koval  <julia.koval@intel.com>
23614         * doc/invoke.texi (Skylake Server): Add CLWB.
23615         Cannonlake): Remove CLWB.
23617 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
23619         PR tree-optimization/84841
23620         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
23621         1 << 3.
23622         (FLOAT_ONE_CONST_TYPE): Define.
23623         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
23624         (sort_by_operand_rank): Put entries with higher constant_type last
23625         rather than first to match comments.
23627 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
23629         * config/nios2/nios2.md (movsi_internal): Fix thinko in
23630         split predicate.
23632 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23634         PR c++/79085
23635         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
23636         check and use address of target always.
23638 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
23640         PR target/84574
23641         * config/i386/i386.c (indirect_thunk_needed): Update comments.
23642         (indirect_thunk_bnd_needed): Likewise.
23643         (indirect_thunks_used): Likewise.
23644         (indirect_thunks_bnd_used): Likewise.
23645         (indirect_return_needed): New.
23646         (indirect_return_bnd_needed): Likewise.
23647         (output_indirect_thunk_function): Add a bool argument for
23648         function return.
23649         (output_indirect_thunk_function): Don't generate alias for
23650         function return thunk.
23651         (ix86_code_end): Call output_indirect_thunk_function to generate
23652         function return thunks.
23653         (ix86_output_function_return): Set indirect_return_bnd_needed
23654         and indirect_return_needed instead of indirect_thunk_bnd_needed
23655         and indirect_thunk_needed.
23657 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
23659         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
23660         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
23661         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
23663 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
23664             Paul Hua <paul.hua.gm@gmail.com>
23666         PR c/84852
23667         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
23669 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
23671         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
23672         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
23673         resp. SFmode cases.
23675 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
23677         PR target/84711
23678         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
23679         instead of GET_MODE_SIZE when comparing Units.
23681 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
23683         PR target/68256
23684         * varasm.c (hash_section): Return an unchangeble hash value
23685         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
23686         Return !aarch64_can_use_per_function_literal_pools_p ().
23688 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
23690         PR target/84860
23691         * optabs.c (emit_conditional_move): Pass address of cmode's copy
23692         rather than address of cmode as last argument to prepare_cmp_insn.
23694 2018-03-15  Julia Koval  <julia.koval@intel.com>
23696         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
23697         F_AVX512VNNI, F_AVX512BITALG): New.
23699 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
23701         PR target/83451
23702         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
23703         insn for floating-point loads and stores.
23705 2018-03-14  Carl Love  <cel@us.ibm.com>
23707         * config/rs6000/rs6000-c.c: Add macro definitions for
23708         ALTIVEC_BUILTIN_VEC_PERMXOR.
23709         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
23710         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
23711         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
23712         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
23713         UNSPEC_VPERMXOR.
23714         * config/doc/extend.texi: Add prototypes for vec_permxor.
23716 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
23718         PR c/84852
23719         * diagnostic-show-locus.c (class layout_point): Convert m_line
23720         from int to linenum_type.
23721         (line_span::comparator): Use linenum "compare" function when
23722         comparing line numbers.
23723         (test_line_span): New function.
23724         (layout_range::contains_point): Convert param "row" from int to
23725         linenum_type.
23726         (layout_range::intersects_line_p): Likewise.
23727         (layout::will_show_line_p): Likewise.
23728         (layout::print_source_line): Likewise.
23729         (layout::should_print_annotation_line_p): Likewise.
23730         (layout::print_annotation_line): Likewise.
23731         (layout::print_leading_fixits): Likewise.
23732         (layout::annotation_line_showed_range_p): Likewise.
23733         (struct line_corrections): Likewise for field m_row.
23734         (line_corrections::line_corrections): Likewise for param "row".
23735         (layout::print_trailing_fixits): Likewise.
23736         (layout::get_state_at_point): Likewise.
23737         (layout::get_x_bound_for_row): Likewise.
23738         (layout::print_line): Likewise.
23739         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
23740         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
23741         * input.c (selftest::test_linenum_comparisons): New function.
23742         (selftest::input_c_tests): Call it.
23743         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
23744         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
23745         * selftest.h (ASSERT_GT): New macro.
23746         (ASSERT_GT_AT): New macro.
23747         (ASSERT_LT): New macro.
23748         (ASSERT_LT_AT): New macro.
23750 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
23752         PR rtl-optimization/84780
23753         * combine.c (distribute_links): Don't make a link based on pc_rtx.
23755 2018-03-14  Martin Liska  <mliska@suse.cz>
23757         * tree.c (record_node_allocation_statistics): Use
23758         get_stats_node_kind.
23759         (get_stats_node_kind): New function extracted from
23760         record_node_allocation_statistics.
23761         (free_node): Use get_stats_node_kind.
23763 2018-03-14  Richard Biener  <rguenther@suse.de>
23765         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
23766         that the value-set of ANTIC_IN doesn't grow.
23768         Revert
23769         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
23770         member.
23771         (BB_VISITED_WITH_VISITED_SUCCS): New define.
23772         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
23774 2018-03-14  Julia Koval  <julia.koval@intel.com>
23776         * config.gcc (icelake-client, icelake-server): New.
23777         (icelake): Remove.
23778         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
23779         (initial_ix86_arch_features): Ditto.
23780         (PTA_SKYLAKE): Add SGX.
23781         (PTA_ICELAKE): Remove.
23782         (PTA_ICELAKE_CLIENT): New.
23783         (PTA_ICELAKE_SERVER): New.
23784         (ix86_option_override_internal): Split up icelake on icelake client and
23785         icelake server.
23786         (get_builtin_code_for_version): Ditto.
23787         (fold_builtin_cpu): Ditto.
23788         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
23789         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
23790         * config/i386/i386.h (processor_type): Ditto.
23791         * doc/invoke.texi: Ditto.
23793 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
23795         PR sanitizer/83392
23796         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
23797         INTEGER_CST offset, add it together with bitpos / 8 and
23798         sign extend based on POINTER_SIZE.
23800         PR target/84844
23801         Revert
23802         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
23804         PR target/78090
23805         * config/i386/constraints.md (Yc): New register constraint.
23806         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
23807         Use Yc constraint for alternative 2 of operand 0.  Remove
23808         preferred_for_speed attribute.
23810 2018-03-14  Richard Biener  <rguenther@suse.de>
23812         PR tree-optimization/84830
23813         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
23814         with the old one to avoid oscillations.
23816 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
23818         PR target/83712
23819         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
23820         pseudos.
23821         (assign_by_spills): Return a flag of reload assignment failure.
23822         Do not process the reload assignment failures.  Do not spill other
23823         reload pseudos if they has the same reg class.  Update n if
23824         necessary.
23825         (lra_assign): Add a return arg.  Set up from the result of
23826         assign_by_spills call.
23827         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
23828         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
23829         usage_insns if it is not NULL.
23830         (spill_hard_reg_in_range): New function.
23831         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
23832         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
23833         function prototypes.
23834         (lra_assign): Change prototype.
23835         * lra.c (lra): Add code to deal with fails by splitting hard reg
23836         live ranges.
23838 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
23840         * config/riscv/riscv.opt (mrelax): New option.
23841         * config/riscv/riscv.c (riscv_file_start): Emit ".option
23842         "norelax" when riscv_mrelax is disabled.
23843         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
23845 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
23847         PR target/84743
23848         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
23849         reassociation for int modes.
23851 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23853         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
23854         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
23855         for big-endian.
23856         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
23857         * config/aarch64/aarch64-sve.md
23858         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
23859         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
23860         (*extend<mode><Vwide>2): Rename to...
23861         (aarch64_sve_extend<mode><Vwide>2): ...this.
23862         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
23863         renaming the old pattern to...
23864         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
23865         unsigned packs.
23866         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
23867         define_expand, renaming the old pattern to...
23868         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
23869         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
23870         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
23871         account when deciding which SVE instruction the optab should use.
23872         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
23874 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23876         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
23877         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
23878         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
23879         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
23880         (tlsdesc_small_<mode>): Turn a define_expand and use
23881         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
23882         (tlsdesc_small_advsimd_<mode>): ...this.
23883         (tlsdesc_small_sve_<mode>): New pattern.
23885 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
23887         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
23888         (UNSPEC_UMUL_HIGHPART): New constants.
23889         (MUL_HIGHPART): New int iteraor.
23890         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
23891         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
23892         define_expand.
23893         (*<su>mul<mode>3_highpart): New define_insn.
23895 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
23897         PR lto/84805
23898         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
23899         incomplete types.
23901 2018-03-13  Martin Liska  <mliska@suse.cz>
23903         PR ipa/84658.
23904         * (sem_item_optimizer::sem_item_optimizer): Initialize new
23905         vector.
23906         (sem_item_optimizer::~sem_item_optimizer): Release it.
23907         (sem_item_optimizer::merge_classes): Register variable aliases.
23908         (sem_item_optimizer::fixup_pt_set): New function.
23909         (sem_item_optimizer::fixup_points_to_sets): Likewise.
23910         * ipa-icf.h: Declare new variables and functions.
23912 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
23914         PR middle-end/84834
23915         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
23916         integer_pow2p@2 and test integer_pow2p in condition.
23917         (A < 0 ? C : 0): Similarly for @1.
23919         PR middle-end/84831
23920         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
23921         characters starting at p contain '\0' character, don't look beyond
23922         that.
23924         PR target/84827
23925         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
23926         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
23928         PR target/84828
23929         * reg-stack.c (change_stack): Change update_end var from int to
23930         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
23931         also call set_block_for_insn on the newly added insns and rescan.
23933         PR target/84786
23934         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
23935         on the last operand.
23937         PR c++/84704
23938         * tree.c (stabilize_reference_1): Return save_expr (e) for
23939         STATEMENT_LIST even if it doesn't have side-effects.
23941 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
23943         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
23945 2018-03-12  Renlin Li  <renlin.li@arm.com>
23947         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
23948         aarch64_output_scalar_simd_mov_immediate.
23950 2018-03-12  Martin Sebor  <msebor@redhat.com>
23952         PR tree-optimization/83456
23953         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
23954         for perfectly overlapping calls to memcpy.
23955         (gimple_fold_builtin_memory_chk): Same.
23956         (gimple_fold_builtin_strcpy): Handle no-warning.
23957         (gimple_fold_builtin_stxcpy_chk): Same.
23958         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
23960 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23962         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
23963         parameter.  Use it for SFmode.
23964         (rs6000_function_arg_advance_1): Adjust.
23965         (rs6000_function_arg): Adjust.
23966         (rs6000_gimplify_va_arg): Pass false for that new parameter.
23968 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
23970         PR rtl-optimization/84169
23971         PR rtl-optimization/84780
23972         * combine.c (can_combine_p): Check for a 2-insn combination whether
23973         the destination register is used between the two insns, too.
23975 2018-03-12  Richard Biener  <rguenther@suse.de>
23977         PR tree-optimization/84803
23978         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
23979         for refs DR analysis didn't process.
23981 2018-03-12  Richard Biener  <rguenther@suse.de>
23983         PR tree-optimization/84777
23984         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
23985         force-vectorize loops ignore whether we are optimizing for size.
23987 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
23989         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
23990         (TARGET_MD_ASM_ADJUST): Define.
23992 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
23993             Kito Cheng  <kito.cheng@gmail.com>
23994             Chung-Ju Wu  <jasonwucj@gmail.com>
23996         * config/nds32/nds32.c (nds32_compute_stack_frame,
23997         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
23998         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
23999         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
24000         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
24001         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
24002         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
24003         * config/nds32/nds32.md (prologue, epilogue): Use macro
24004         NDS32_V3PUSH_AVAILABLE_P to do checking.
24006 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
24008         PR debug/58150
24009         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
24010         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
24011         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
24012         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
24013         addition of most attributes on !orig_type_die or the attribute not
24014         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
24016 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
24017             Chung-Ju Wu  <jasonwucj@gmail.com>
24019         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
24020         __NDS32_VH__ macro.
24021         * config/nds32/nds32.opt (mvh): New option.
24023 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
24024             Chung-Ju Wu  <jasonwucj@gmail.com>
24026         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
24027         function.
24028         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
24029         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
24030         definition.
24032 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
24033             Chung-Ju Wu  <jasonwucj@gmail.com>
24035         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
24036         function.
24037         * config/nds32/nds32-multiple.md (strlensi): New pattern.
24038         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
24040 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
24041             Kito Cheng  <kito.cheng@gmail.com>
24042             Chung-Ju Wu  <jasonwucj@gmail.com>
24044         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
24045         UNSPEC_FFMISM and UNSPEC_FLMISM.
24046         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
24047         for ffb, ffmism and flmism.
24048         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
24049         (unspec_ffmism): Ditto.
24050         (unspec_flmism): Ditto.
24051         (nds32_expand_builtin_impl): Check if string extension is available.
24052         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
24053         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
24055 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
24057         Reverting patch:
24058         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
24060         PR target/83712
24061         * lra-assigns.c (assign_by_spills): Return a flag of reload
24062         assignment failure.  Do not process the reload assignment
24063         failures.  Do not spill other reload pseudos if they has the same
24064         reg class.
24065         (lra_assign): Add a return arg.  Set up from the result of
24066         assign_by_spills call.
24067         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24068         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
24069         usage_insns if it is not NULL.
24070         (spill_hard_reg_in_range): New function.
24071         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24072         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24073         function prototypes.
24074         (lra_assign): Change prototype.
24075         * lra.c (lra): Add code to deal with fails by splitting hard reg
24076         live ranges.
24078 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
24080         PR target/84807
24081         * config/i386/i386.opt: Replace Enforcment with Enforcement.
24083 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
24085         PR debug/84620
24086         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
24087         (dw_val_node): Add val_symbolic_view.
24088         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
24089         (symview_upper_bound): New.
24090         (new_line_info_table): Initialize symviews_since_reset.
24091         (dwarf2out_source_line): Count symviews_since_reset and set
24092         symview_upper_bound.
24093         (dw_val_equal_p): Handle symview.
24094         (add_AT_symview): New.
24095         (print_dw_val): Handle symview.
24096         (attr_checksum, attr_checksum_ordered): Likewise.
24097         (same_dw_val_p, size_of_die): Likewise.
24098         (value_format, output_die): Likewise.
24099         (add_high_low_attributes): Use add_AT_symview for entry_view.
24100         (dwarf2out_finish): Reset symview_upper_bound, clear
24101         zero_view_p.
24103 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
24105         PR target/83969
24106         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
24107         Add strict argument and use it.
24108         (rs6000_split_multireg_move): Update for new strict argument.
24109         (mem_operand_gpr): Disallow all non-offsettable addresses.
24110         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
24112 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
24114         PR target/84772
24115         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
24116         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
24117         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
24119         PR c++/84767
24120         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
24121         decl, use remap_type if we want to use the type.
24123 2018-03-09  Martin Sebor  <msebor@redhat.com>
24125         PR tree-optimization/84526
24126         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
24127         Remove dead code.
24128         (builtin_access::generic_overlap): Be prepared to handle non-array
24129         base objects.
24131 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
24133         PR rtl-optimization/84682
24134         * lra-constraints.c (process_address_1): Check is_address flag
24135         for address constraints.
24136         (process_alt_operands): Likewise.
24137         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
24138         preprocess_constraints.
24139         * recog.h (preprocess_constraints): Add oploc parameter.
24140         Adjust callers.
24141         * recog.c (preprocess_constraints): Test address_operand for
24142         CT_ADDRESS constraints.
24144 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
24146         PR target/83712
24147         * lra-assigns.c (assign_by_spills): Return a flag of reload
24148         assignment failure.  Do not process the reload assignment
24149         failures.  Do not spill other reload pseudos if they has the same
24150         reg class.
24151         (lra_assign): Add a return arg.  Set up from the result of
24152         assign_by_spills call.
24153         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
24154         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
24155         usage_insns if it is not NULL.
24156         (spill_hard_reg_in_range): New function.
24157         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
24158         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
24159         function prototypes.
24160         (lra_assign): Change prototype.
24161         * lra.c (lra): Add code to deal with fails by splitting hard reg
24162         live ranges.
24164 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24166         PR target/83193
24167         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
24168         Accept complain bool parameter.  Only emit errors if it is true.
24169         (arm_parse_cpu_option_name): Likewise.
24170         (arm_target_thumb_only): Adjust callers of the above.
24171         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
24172         prototype to take a default true bool parameter.
24173         (arm_parse_arch_option_name): Likewise.
24175 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
24176             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
24178         PR jit/64089
24179         PR jit/84288
24180         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
24181         * configure: Regenerate.
24182         * configure.ac ("linker --version-script option"): New.
24183         ("linker soname option"): New.
24185 2018-03-09  Richard Biener  <rguenther@suse.de>
24187         PR tree-optimization/84775
24188         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
24189         immediate uses of predicate stmts and mark them modified.
24191         Revert
24192         PR tree-optimization/84178
24193         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24194         to caller.
24195         (version_loop_for_if_conversion): Delay update_ssa call.
24196         (tree_if_conversion): Delay update_ssa until after predicate
24197         insertion.
24199 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
24201         PR target/84763
24202         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
24203         when the function accesses prior frames.
24205 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
24207         PR debug/84456
24208         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
24209         gen_llsym, otherwise call maybe_gen_llsym.
24211         PR inline-asm/84742
24212         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
24213         has ',' character inside of it.
24215 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24217         PR target/84748
24218         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
24219         as clobbering CC_REGNUM.
24221 2018-03-08  Richard Biener  <rguenther@suse.de>
24223         PR middle-end/84552
24224         * tree-scalar-evolution.c: Include tree-into-ssa.h.
24225         (follow_copies_to_constant): Do not follow SSA names registered
24226         for update.
24228 2018-03-08  Richard Biener  <rguenther@suse.de>
24230         PR tree-optimization/84178
24231         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
24232         to caller.
24233         (version_loop_for_if_conversion): Delay update_ssa call.
24234         (tree_if_conversion): Delay update_ssa until after predicate
24235         insertion.
24237 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
24239         PR tree-optimization/84178
24240         * tree-if-conv.c (release_bb_predicate): Remove the
24241         the assertion that the stmts have NULL use_ops.
24242         Discard the statements, asserting that they haven't
24243         yet been added to a BB.
24245 2018-03-08  Richard Biener  <rguenther@suse.de>
24247         PR tree-optimization/84746
24248         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
24249         (phi_translate): Pass in destination ANTIC_OUT set.
24250         (phi_translate_1): Likewise.  For a simplified result lookup
24251         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
24252         (phi_translate_set): Adjust.
24253         (do_pre_regular_insertion): Likewise.
24254         (do_pre_partial_partial_insertion): Likewise.
24256 2018-03-08  Martin Liska  <mliska@suse.cz>
24258         PR gcov-profile/84735
24259         * doc/gcov.texi: Document usage of profile files.
24260         * gcov-io.h: Document changes in the format.
24262 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
24264         PR debug/84404
24265         PR debug/84408
24266         * dwarf2out.c (struct dw_line_info_table): Update comments for
24267         view == -1.
24268         (FORCE_RESET_NEXT_VIEW): New.
24269         (FORCE_RESETTING_VIEW_P): New.
24270         (RESETTING_VIEW_P): Check for -1 too.
24271         (ZERO_VIEW_P): Likewise.
24272         (new_line_info_table): Force-reset next view.
24273         (dwarf2out_begin_function): Likewise.
24274         (dwarf2out_source_line): Simplify zero_view_p initialization.
24275         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
24276         view directly.  Omit view when omitting .loc at line 0.
24278 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
24280         PR tree-optimization/84740
24281         * tree-switch-conversion.c (process_switch): Call build_constructors
24282         only if info.phi_count is non-zero.
24284         PR tree-optimization/84739
24285         * tree-tailcall.c (find_tail_calls): Check call arguments against
24286         DECL_ARGUMENTS (current_function_decl) rather than
24287         DECL_ARGUMENTS (func) when checking for tail recursion.
24289 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24291         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
24292         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
24293         Volker Reichelt's entry and add entries for people that perform
24294         GCC fuzzy testing and report numerous bugs.
24296 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
24298         PR target/82411
24299         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
24300         readonly data in sdata, if that is disabled.
24301         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
24302         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
24303         -mreadonly-in-sdata option.
24305 2018-03-07  Martin Sebor  <msebor@redhat.com>
24307         PR tree-optimization/84468
24308         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
24309         basic block when looking for nul assignment.
24311 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
24313         PR target/84277
24314         * except.h (output_function_exception_table): Adjust prototype.
24315         * except.c (output_function_exception_table): Remove FNNAME parameter
24316         and add SECTION parameter.  Ouput one part of the table at a time.
24317         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
24318         the first part of the exception table and emit unwind directives.
24319         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
24320         (i386_pe_seh_cold_init): Likewise.
24321         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
24322         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
24323         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
24324         (ix86_output_call_insn): Emit a nop in one more case for SEH.
24325         * config/i386/winnt.c: Include except.h.
24326         (struct seh_frame_state): Add reg_offset, after_prologue and
24327         in_cold_section fields.
24328         (i386_pe_seh_end_prologue): Set seh->after_prologue.
24329         (i386_pe_seh_cold_init): New function.
24330         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
24331         to seh->in_cold_section.
24332         (seh_emit_push): Record the offset of the push.
24333         (seh_emit_save): Record the offet of the save.
24334         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
24335         Test seh->after_prologue to disregard the epilogue.
24336         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
24337         (i386_pe_end_cold_function): New function.
24339 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
24341         PR fortran/84565
24342         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
24343         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
24345         PR c++/84704
24346         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
24347         on tmp_var.
24348         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
24349         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
24351         PR middle-end/84723
24352         * multiple_target.c: Include tree-inline.h and intl.h.
24353         (expand_target_clones): Diagnose and fail if node->definition and
24354         !tree_versionable_function_p (node->decl).
24356 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
24358         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
24359         sprint_ul.
24360         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
24361         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
24362         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
24364 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
24366         PR target/84710
24367         * combine.c (try_combine): Use reg_or_subregno instead of handling
24368         just paradoxical SUBREGs and REGs.
24370 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
24372         * config/arc/arc.c (arc_finalize_pic): Remove function.
24373         (arc_must_save_register): We use single base PIC register, remove
24374         checks to save/restore the PIC register.
24375         (arc_expand_prologue): Likewise.
24376         * config/arc/arc-protos.h (arc_set_default_type_attributes):
24377         Remove.
24378         (arc_verify_short): Likewise.
24379         (arc_attr_type): Likewise.
24380         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
24381         (walk_stores): Likewise.
24382         (arc_address_cost): Make it static.
24383         (arc_verify_short): Likewise.
24384         (branch_dest): Likewise.
24385         (arc_attr_type): Likewise.
24386         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
24387         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
24388         (arc_final_prescan_insn): Remove inserting the nops due to
24389         hardware hazards.  It is done in reorg step.
24390         (insn_length_variant_t): Remove.
24391         (insn_length_parameters_t): Likewise.
24392         (arc_insn_length_parameters): Likewise.
24393         (arc_get_insn_variants): Likewise.
24394         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
24396 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
24398         PR inline-asm/84683
24399         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
24400         assertion failure.
24402         PR tree-optimization/84687
24403         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
24404         on new_node->decl.
24405         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
24407 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24409         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
24410         Rename to ppc_speculation_barrier.
24411         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
24412         __builtin_ppc_speculation_barrier.
24414 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
24416         PR target/84700
24417         * combine.c (combine_simplify_rtx): Don't try to simplify if
24418         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
24419         are equal to x.
24421 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
24423         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
24424         to 32 bytes when compiling for POWER9.
24426 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
24428         PR target/84564
24429         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
24430         regparm >= 3 with no arg reg available also for calls with
24431         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
24433         PR target/84524
24434         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
24435         orig,vex.
24436         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
24438 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
24440         PR target/84264
24441         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
24443 2018-03-05  Richard Biener  <rguenther@suse.de>
24445         PR tree-optimization/84486
24446         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
24447         When inserting a __builtin_assume_aligned call set the LHS
24448         SSA name alignment info accordingly.
24450 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
24452         PR tree-optimization/84114
24453         * config/aarch64/aarch64.c (aarch64_reassociation_width)
24454         Avoid reassociation of FLOAT_MODE addition.
24456 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
24458         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
24459         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
24460         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
24461         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
24462         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
24463         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
24464         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
24465         and -mwbnoinvd.
24466         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
24467         __builtin_ia32_wbinvd): New builtins.
24468         (SPECIAL_ARGS2): New.
24469         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
24470         (SPECIAL_ARGS2): New.
24471         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
24472         (ix86_valid_target_attribute_inner_p): Ditto.
24473         (ix86_init_mmx_sse_builtins): Add special_args2.
24474         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
24475         TARGET_WBNOINVD_P): New.
24476         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
24477         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
24478         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
24479         * config/i386/immintrin.h (_wbinvd): New intrinsic.
24480         * config/i386/pconfigintrin.h: New file.
24481         * config/i386/wbnoinvdintrin.h: Ditto.
24482         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
24483         wbnoinvdintrin.h.
24484         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
24486 2018-03-05  Richard Biener  <rguenther@suse.de>
24488         PR tree-optimization/84670
24489         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
24490         member.
24491         (BB_VISITED_WITH_VISITED_SUCCS): New define.
24492         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
24493         (compute_antic_aux): Only assert the number of values in ANTIC_IN
24494         doesn't grow if all successors (recursively) were visited at least
24495         once.
24497 2018-03-05  Richard Biener  <rguenther@suse.de>
24499         PR tree-optimization/84650
24500         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
24501         if executed in the loop pipeline.
24503 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
24505         * doc/configfiles.texi (Configuration Files): Move info about
24506         conditionalizing $target-protos.h to...
24507         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
24508         differs from $target-protos.h.
24510 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
24511             Chung-Ju Wu  <jasonwucj@gmail.com>
24513         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
24514         * config/nds32/nds32-multiple.md (setmemsi): Define.
24515         * config/nds32/nds32-memory-manipulation.c
24516         (nds32_gen_dup_4_byte_to_word_value): New.
24517         (emit_setmem_word_loop): New.
24518         (emit_setmem_byte_loop): New.
24519         (nds32_expand_setmem_loop): New.
24520         (nds32_expand_setmem_loop_v3m): New.
24521         (nds32_expand_setmem_unroll): New.
24522         (nds32_expand_setmem): New.
24524 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24525             Chung-Ju Wu  <jasonwucj@gmail.com>
24527         * config/nds32/nds32-memory-manipulation.c
24528         (nds32_emit_load_store): New.
24529         (nds32_emit_post_inc_load_store): New.
24530         (nds32_emit_mem_move): New.
24531         (nds32_emit_mem_move_block): New.
24532         (nds32_expand_movmemsi_loop_unknown_size): New.
24533         (nds32_expand_movmemsi_loop_known_size): New.
24534         (nds32_expand_movmemsi_loop): New.
24535         (nds32_expand_movmemsi_unroll): New.
24536         (nds32_expand_movmemqi): Rename ...
24537         (nds32_expand_movmemsi): ... to this.
24538         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
24539         (movmemsi): ... to this.
24540         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
24541         (nds32_expand_movmemsi): ... to this.
24543 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24544             Monk Chiang  <sh.chiang04@gmail.com>
24545             Chung-Ju Wu  <jasonwucj@gmail.com>
24547         * config/nds32/nds32-protos.h
24548         (nds32_expand_load_multiple): New arguments.
24549         (nds32_expand_store_multiple): Ditto.
24550         (nds32_valid_multiple_load_store): Rename ...
24551         (nds32_valid_multiple_load_store_p): ... to this.
24552         * config/nds32/nds32-memory-manipulation.c
24553         (nds32_expand_load_multiple): Refine implementation.
24554         (nds32_expand_store_multiple): Ditto.
24555         * config/nds32/nds32-multiple.md
24556         (load_multiple): Update nds32_expand_load_multiple interface.
24557         (store_multiple): Update nds32_expand_store_multiple interface.
24558         * config/nds32/nds32-predicates.c
24559         (nds32_valid_multiple_load_store): Rename ...
24560         (nds32_valid_multiple_load_store_p): ... to this and refine
24561         implementation.
24562         * config/nds32/predicates.md
24563         (nds32_load_multiple_and_update_address_operation): New predicate.
24564         (nds32_store_multiple_and_update_address_operation): New predicate.
24566 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
24567             Chung-Ju Wu  <jasonwucj@gmail.com>
24569         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
24570         (combo): New attribute.
24571         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
24573 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
24575         * config/nds32/nds32.opt: Change -mcmodel= default value.
24577 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
24578             Monk Chiang  <sh.chiang04@gmail.com>
24579             Chung-Ju Wu  <jasonwucj@gmail.com>
24581         * config/nds32/constants.md (unspec_element): New enum.
24582         * config/nds32/constraints.md (Umw): New constraint.
24583         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
24584         * config/nds32/nds32-intrinsic.md: Likewise.
24585         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
24586         (nds32_valid_smw_lwm_base_p): New.
24587         (nds32_output_smw_single_word): New.
24588         (nds32_output_lmw_single_word): New.
24589         (nds32_expand_unaligned_load): New.
24590         (nds32_expand_unaligned_store): New.
24591         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
24592         (nds32_output_smw_single_word): Declare.
24593         (nds32_output_lmw_single_word): Declare.
24594         (nds32_expand_unaligned_load): Declare.
24595         (nds32_expand_unaligned_store): Declare.
24596         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
24597         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
24598         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
24599         NDS32_BUILTIN_UASTORE_DW.
24600         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
24601         predicate.
24603 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
24604             Kito Cheng  <kito.cheng@gmail.com>
24605             Chung-Ju Wu  <jasonwucj@gmail.com>
24607         * config/nds32/nds32-intrinsic.c
24608         (nds32_expand_builtin_null_ftype_reg): Delete.
24609         (nds32_expand_builtin_reg_ftype_imm): Ditto.
24610         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
24611         (nds32_read_argument): New.
24612         (nds32_legitimize_target): Ditto.
24613         (nds32_legitimize_argument): Ditto.
24614         (nds32_check_constant_argument): Ditto.
24615         (nds32_expand_unop_builtin): Ditto.
24616         (nds32_expand_unopimm_builtin): Ditto.
24617         (nds32_expand_binop_builtin): Ditto.
24618         (nds32_builtin_decl_impl): Ditto.
24619         (builtin_description): Ditto.
24620         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
24621         (nds32_init_builtins_impl): Ditto.
24622         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
24623         (nds32_builtin_decl): New.
24624         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
24625         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
24627 2018-03-02  Jeff Law  <law@redhat.com>
24629         * reorg.c (stop_search_p): Handle DEBUG_INSN.
24630         (redundant_insn, fill_simple_delay_slots): Likewise.
24631         (fill_slots_from_thread): Likewise.
24632         * resource.c (mark_referenced_resources): Likewise.
24633         (mark_set_resources, find_dead_or_set_registers): Likewise.
24635 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24637         * substring-locations.h (format_warning_va): Formatting fix for
24638         ATTRIBUTE_GCC_DIAG.
24639         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
24640         argument.
24641         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
24642         * substring-locations.c: Include intl.h.
24643         (format_warning_va): Turned into small wrapper around
24644         format_warning_n_va, renamed to ...
24645         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
24646         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
24647         use ngettext.
24648         (format_warning_at_substring_n): New function.
24649         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
24650         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
24651         format_warning_at_substring with just a shorter name instead of
24652         const function pointer.
24653         (fmtwarn_n): New function.
24654         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
24655         appropriate, get rid of all the fmtstr temporaries, move conditionals
24656         with G_() wrapped string literals directly into fmtwarn arguments,
24657         cast dir.len to (int), formatting fixes.
24659 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
24661         * doc/invoke.texi: Remove "Cilk Plus" references.
24663 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24664             Richard Biener  <rguenther@suse.de>
24666         PR ipa/84628
24667         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
24668         for error or warning attributes if CALL_FROM_THUNK_P is set.
24669         Formatting fixes.
24671 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24673         PR target/56540
24674         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
24675         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
24677         PR target/56540
24678         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
24679         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
24681         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
24682         instead of -1U in last predictors element's probability member.
24684 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
24686         PR ipa/83983
24687         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
24688         arguments if they are comparable.
24690 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24692         PR tree-optimization/84634
24693         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
24694         masks and masked_loop_p with a single loop_masks, making sure it's
24695         null for bb vectorization.
24697 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
24699         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
24700         (vect_analyze_data_ref_access): Use loop->safe_len rather than
24701         loop->force_vectorize to check whether there is no alias.
24703 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
24705         PR target/84614
24706         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
24707         prototypes.
24708         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
24709         comments.
24710         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
24711         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
24712         instead of a loop around prev_real_insn.
24713         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
24714         prev_real_insn.
24716         PR inline-asm/84625
24717         * config/i386/i386.c (ix86_print_operand): Use conditional
24718         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
24719         zero vector.
24721 2018-03-02  Richard Biener  <rguenther@suse.de>
24723         PR tree-optimization/84427
24724         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
24725         (bitmap_set_subtract_values): Rewrite to handle multiple
24726         exprs per value.
24727         (clean): Likewise.
24728         (prune_clobbered_mems): Likewise.
24729         (phi_translate): Take edge instead of pred/phiblock.
24730         (phi_translate_1): Likewise.
24731         (phi_translate_set): Likewise.  Insert all translated
24732         exprs for a value into the set, keeping possibly multiple
24733         expressions per value.
24734         (compute_antic_aux): Adjust for phi_translate changes.
24735         When intersecting union the expressions and prune those
24736         not in the final value set, keeping possibly multiple
24737         expressions per value.  Do not use value-insertion
24738         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
24739         all expressions.  Add verification that the value-sets
24740         only shrink during iteration.
24741         (compute_partial_antic_aux): Adjust for the phi_translate changes.
24742         (do_pre_regular_insertion): Likewise.
24743         (do_pre_partial_partial_insertion): Likewise.
24745 2018-03-02  Richard Biener  <rguenther@suse.de>
24747         PR target/82005
24748         * config/darwin.c (saved_debug_info_level): New static global.
24749         (darwin_asm_lto_start): Disable debug info generation for LTO out.
24750         (darwin_asm_lto_end): Restore debug info generation settings.
24752 2018-03-01  Martin Liska  <mliska@suse.cz>
24754         PR sanitizer/82484
24755         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
24756         volatile arguments.
24758 2018-03-01  Richard Biener  <rguenther@suse.de>
24760         PR debug/84645
24761         * dwarf2out.c (gen_variable_die): Properly handle late VLA
24762         type annotation with LTO when debug was disabled at compile-time.
24764 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
24766         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
24767         XINT with INTVAL.
24768         (mips_final_postscan_insn): Likewise.
24770 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
24772         PR rtl-optimization/84528
24773         * alias.c (init_alias_target): Add commentary.
24774         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
24775         a unique base value if the frame pointer is not eliminated
24776         to the stack pointer.
24778 2018-03-01  Tom de Vries  <tom@codesourcery.com>
24780         PR rtl-optimization/83327
24781         * lra-int.h (hard_regs_spilled_into): Declare.
24782         * lra.c (hard_regs_spilled_into): Define.
24783         (init_reg_info): Init hard_regs_spilled_into.
24784         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
24785         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
24786         (process_bb_lives): Handle hard_regs_spilled_into.
24787         (lra_create_live_ranges_1): Before doing liveness propagation, clear
24788         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
24790 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
24792         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
24793         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
24794         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
24795         * config/rs6000/aix72.h: New file.
24797 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
24799         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
24800         instead of warning_at with conditional singular and plural messages
24801         where possible.
24803         PR target/52991
24804         * stor-layout.c (update_alignment_for_field): For
24805         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
24806         && !DECL_PACKED (field), do the alignment update, just use
24807         only desired_align instead of MAX (type_align, desired_align)
24808         as the alignment.
24809         (place_field): Don't do known_align < desired_align handling
24810         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
24811         is non-NULL, instead do it after rli->prev_field handling and
24812         only if not within a bitfield word.  For DECL_PACKED (field)
24813         use type_align of BITS_PER_UNIT.
24815 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
24817         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
24818         superfluous parentheses and trailing spaces.
24820 2018-02-28  Richard Biener  <rguenther@suse.de>
24822         PR tree-optimization/84584
24823         * graphite-scop-detection.c (scop_detection::add_scop): Discard
24824         SCoPs with fake exit edge.
24826 2018-02-28  Martin Liska  <mliska@suse.cz>
24828         PR testsuite/84597
24829         * timevar.c (timer::print): Fix format to properly print 100%
24830         values.
24832 2018-02-28  Richard Biener  <rguenther@suse.de>
24834         PR middle-end/84607
24835         * genmatch.c (capture_info::walk_match): Do not mark
24836         captured expressions without operands as expr_p given
24837         they act more like predicates and should be subject to
24838         "lost tail" side-effect preserving.
24840 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
24842         PR rtl-optimization/81611
24843         * auto-inc-dec.c (attempt_change): Move dead note from
24844         mem_insn if it's the next use of regno
24845         (find_address): Take address use of reg holding
24846         non-incremented value.  Add parm to limit search to the named
24847         reg only.
24848         (merge_in_block): Attempt to use a mem insn that is the next
24849         use of the original regno.
24851 2018-02-27  Martin Sebor  <msebor@redhat.com>
24853         PR c++/83871
24854         * doc/invoke.texi (-Wmissing-attributes): New option.
24855         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
24857 2018-02-27  Martin Sebor  <msebor@redhat.com>
24859         PR translation/84207
24860         * diagnostic-core.h (warning_n, error_n, inform_n): Change
24861         n argument to unsigned HOST_WIDE_INT.
24862         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
24863         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
24864         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
24865         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
24867 2018-02-27  Richard Biener  <rguenther@suse.de>
24869         PR tree-optimization/84512
24870         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
24871         Do not use the estimate returned from record_stmt_cost for
24872         the scalar iteration cost but sum properly using add_stmt_cost.
24874 2018-02-27  Richard Biener  <rguenther@suse.de>
24876         PR tree-optimization/84466
24877         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
24878         Adjust last change to less strictly validate use operands.
24880 2018-02-27  Martin Liska  <mliska@suse.cz>
24882         PR gcov-profile/84548
24883         * gcov.c (process_file): Allow partial overlap and consider it
24884         also as group functions.
24885         (output_lines): Properly calculate range of lines for a group.
24887 2018-02-27  Martin Liska  <mliska@suse.cz>
24889         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
24890         'ggc' suffixes.  Change first column width.
24891         (timer::print): Fix formatting of the column.
24893 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
24895         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
24896         preserve inline entry blocks for the sake of debug inline
24897         entry point markers alone.
24898         (remove_unused_locals): Suggest in comments a better place to
24899         force the preservation of inline entry blocks that are
24900         otherwise unused, but do not preserve them.
24902 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24904         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
24906 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24908         PR target/84039
24909         * config/i386/constraints.md (Bs): Replace
24910         ix86_indirect_branch_register with
24911         TARGET_INDIRECT_BRANCH_REGISTER.
24912         (Bw): Likewise.
24913         * config/i386/i386.md (indirect_jump): Likewise.
24914         (tablejump): Likewise.
24915         (*sibcall_memory): Likewise.
24916         (*sibcall_value_memory): Likewise.
24917         Peepholes of indirect call and jump via memory: Likewise.
24918         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
24919         (*sibcall_value_GOT_32): Likewise.
24920         * config/i386/predicates.md (indirect_branch_operand): Likewise.
24921         (GOT_memory_operand): Likewise.
24922         (call_insn_operand): Likewise.
24923         (sibcall_insn_operand): Likewise.
24924         (GOT32_symbol_operand): Likewise.
24925         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
24927 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24929         PR rtl-optimization/83496
24930         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
24931         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
24932         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
24933         redundant insn, if any.
24934         (relax_delay_slots): Likewise.
24935         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
24937 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
24939         PR tree-optimization/83965
24940         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
24941         that grouped statements are part of a reduction chain.  Return
24942         true if the statement is not marked as a reduction itself but
24943         is part of a group.
24944         (vect_recog_dot_prod_pattern): Don't check whether the statement
24945         is part of a group here.
24946         (vect_recog_sad_pattern): Likewise.
24947         (vect_recog_widen_sum_pattern): Likewise.
24949 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
24951         PR debug/84545
24952         * final.c (rest_of_clean_state): Also look for calls inside sequences.
24954 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
24956         PR target/84530
24957         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
24958         the bool argument.
24959         (ix86_output_indirect_function_return): New prototype.
24960         (ix86_split_simple_return_pop_internal): Likewise.
24961         * config/i386/i386.c (indirect_return_via_cx): New.
24962         (indirect_return_via_cx_bnd): Likewise.
24963         (indirect_thunk_name): Handle return va CX_REG.
24964         (output_indirect_thunk_function): Create alias for
24965         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
24966         (ix86_output_indirect_jmp): Remove the bool argument.
24967         (ix86_output_indirect_function_return): New function.
24968         (ix86_split_simple_return_pop_internal): Likewise.
24969         * config/i386/i386.md (*indirect_jump): Don't pass false
24970         to ix86_output_indirect_jmp.
24971         (*tablejump_1): Likewise.
24972         (simple_return_pop_internal): Change it to define_insn_and_split.
24973         Call ix86_split_simple_return_pop_internal to split it for
24974         -mfunction-return=.
24975         (simple_return_indirect_internal): Call
24976         ix86_output_indirect_function_return instead of
24977         ix86_output_indirect_jmp.
24979 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
24981         PR bootstrap/84405
24982         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
24983         memset and value initialization afterwards.
24985 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
24987         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
24989 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
24991         PR target/84521
24992         * common/config/aarch64/aarch64-common.c
24993         (aarch_option_optimization_table[]): Switch
24994         off fomit-frame-pointer
24996 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
24997             Chung-Ju Wu  <jasonwucj@gmail.com>
24999         * config/nds32/nds32-multiple.md (load_multiple): Disallow
25000         volatile memory.
25001         (store_multiple): Ditto.
25003 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
25005         * config.gcc: Add --with-cpu support for nds32 target.
25006         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
25007         * config/nds32/nds32.opt: Add -mcpu= option.
25009 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
25011         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
25012         isel=yes): Warn for these deprecated options.
25014 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
25016         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
25017         ISA_2_5_MASKS_EMBEDDED.
25019 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
25021         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
25022         p->max as pointers rather than using iterative_hash_expr.
25024 2018-02-23  Carl Love  <cel@us.ibm.com>
25026         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
25027         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
25028         BU_P8V_OVERLOAD_2.
25029         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
25030         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
25031         P8V_BUILTIN_VEC_VUNSIGNED2.
25033 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
25035         PR target/81572
25036         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
25037         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
25038         LRA_UNKNOWN_ALT.
25039         * lra-constraints.c (curr_insn_transform): Set up
25040         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
25041         LRA_UNKNOWN_ALT.
25042         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
25043         * lra-eliminations.c (spill_pseudos): Ditto.
25044         (process_insn_for_elimination): Ditto.
25045         * lra-lives.c (reg_early_clobber_p): Use the new macros.
25046         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
25047         LRA_NON_CLOBBERED_ALT.
25049 2018-02-22  Martin Sebor  <msebor@redhat.com>
25051         PR tree-optimization/84480
25052         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
25053         to maybe_diag_stxncpy_trunc.  Call it.
25054         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
25055         from gimple_fold_builtin_strcpy.  Print inlining stack.
25056         (handle_builtin_stxncpy): Print inlining stack.
25057         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
25059 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
25061         PR target/84176
25062         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
25063         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
25064         and -fcheck-pointer-bounds are used together.
25065         (indirect_thunk_prefix): New enum.
25066         (indirect_thunk_need_prefix): New function.
25067         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
25068         "_nt" instead of "_bnd" for NOTRACK prefix.
25069         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
25070         (output_indirect_thunk_function): Likewise.
25071         (): Likewise.
25072         (ix86_code_end): Update output_indirect_thunk_function calls.
25073         (ix86_output_indirect_branch_via_reg): Replace
25074         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
25075         (ix86_output_indirect_branch_via_push): Likewise.
25076         (ix86_output_function_return): Likewise.
25077         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
25078         incompatible with -fcf-protection=branch and
25079         -fcheck-pointer-bounds.
25081 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
25083         PR target/83335
25084         * config/aarch64/aarch64.c (aarch64_print_address_internal):
25085         Change gcc_assert call to output_operand_lossage.
25087 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
25089         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
25091 2018-02-22  DJ Delorie  <dj@redhat.com>
25092             Sebastian Perta  <sebastian.perta@renesas.com>
25093             Oleg Endo  <olegendo@gcc.gnu.org>
25095         * config/rx/rx.c (rx_rtx_costs): New function.
25096         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
25098 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25100         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
25102 2018-02-22  Martin Liska  <mliska@suse.cz>
25104         PR driver/83193
25105         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
25106         Add "native" as a possible value.
25108 2018-02-22  Martin Liska  <mliska@suse.cz>
25110         PR driver/83193
25111         * config/i386/i386.c (ix86_option_override_internal):
25112         Add "native" as a possible value for -march and -mtune.
25114 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
25116         PR target/84502
25117         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
25118         to all type variants.
25120         PR tree-optimization/84503
25121         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
25122         width as info->bitpos + info->bitsize - start.
25123         (merged_store_group::merge_overlapping): Simplify width computation.
25124         (check_no_overlap): New function.
25125         (imm_store_chain_info::try_coalesce_bswap): Compute expected
25126         start + width and last_order of the group, fail if check_no_overlap
25127         fails.
25128         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
25129         to group if check_no_overlap fails.
25131 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
25133         * config/rs6000/altivec.md: Delete contraint arguments to
25134         define_expand, define_split, and define_peephole2, and in
25135         define_insn_and_split if always unused.
25136         * config/rs6000/darwin.md: Ditto.
25137         * config/rs6000/dfp.md: Ditto.
25138         * config/rs6000/rs6000.md: Ditto.
25139         * config/rs6000/sync.md: Ditto.
25140         * config/rs6000/vector.md: Ditto.
25141         * config/rs6000/vsx.md: Ditto.
25143 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
25145         * config/rs6000/altivec.md: Write output control strings as braced
25146         blocks instead of double-quoted strings.
25147         * config/rs6000/darwin.md: Ditto.
25148         * config/rs6000/rs6000.md: Ditto.
25149         * config/rs6000/vector.md: Ditto.
25150         * config/rs6000/vsx.md: Ditto.
25152 2018-02-21  Jason Merrill  <jason@redhat.com>
25154         PR c++/84314 - ICE with templates and fastcall attribute.
25155         * attribs.c (build_type_attribute_qual_variant): Remove assert.
25157 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
25159         * ipa-cp.c (determine_versionability): Fix comment typos.
25161 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
25163         PR c/84229
25164         * ipa-cp.c (determine_versionability): Do not version functions caling
25165         va_arg_pack.
25167 2018-02-21  Martin Liska  <mliska@suse.cz>
25169         PR driver/83193
25170         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
25171         Add "native" as a possible value.
25172         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
25173         the macro when native cpu detection is available.
25175 2018-02-21  Martin Liska  <mliska@suse.cz>
25177         PR driver/83193
25178         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
25179         Add "native" as a possible value.
25180         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
25181         when native cpu detection is available.
25183 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
25184             Martin Sebor  <msebor@redhat.com>
25186         PR tree-optimization/84478
25187         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
25188         false.
25189         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
25190         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
25191         support which is conservatively correct, for 2 only stay conservative
25192         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
25193         argument to the 2 argument get_range_strlen, adjust 6 arg
25194         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
25195         false.
25196         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
25197         (gimple_fold_builtin_strlen): Pass true as last argument to
25198         get_range_strlen.
25200 2018-02-20  Martin Sebor  <msebor@redhat.com>
25202         PR middle-end/84095
25203         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
25204         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
25205         (builtin_memref::builtin_memref): Factor out parts into
25206         set_base_and_offset and call it.
25208 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
25210         PR middle-end/84406
25211         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
25212         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
25213         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
25214         search at the associated MODE_INT.
25216 2018-02-20  Jeff Law  <law@redhat.com>
25218         PR middle-end/82123
25219         PR tree-optimization/81592
25220         PR middle-end/79257
25221         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
25222         for range data rather than using global data.
25223         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
25224         range data rather than using global data.
25225         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
25226         pass it to children as needed.
25227         (struct directive::fmtresult): Similarly.
25228         (struct directive::set_width): Similarly.
25229         (struct directive::set_precision): Similarly.
25230         (format_integer, format_directive, parse_directive): Similarly.
25231         (format_none): Accept unnamed vr_values parameter.
25232         (format_percent, format_floating, format_character): Similarly.
25233         (format_string, format_plain): Similarly.
25234         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
25235         the EVRP range analyzer for range data rather than using global data.
25236         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
25237         gimple-ssa-evrp-analyze.h
25238         (class sprintf_dom_walker): Add after_dom_children member function.
25239         Add evrp_range_analyzer member.
25240         (sprintf_dom_walker::before_dom_children): Call into the EVRP
25241         range analyzer as needed.
25242         (sprintf_dom_walker::after_dom_children): New member function.
25243         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
25244         if not optimizing.
25245         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
25246         (evrp_range_analyzer::pop_to_marker): Likewise.
25248 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
25250         PR tree-optimization/84419
25251         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
25252         with the required type if its current type is compatible but
25253         different.
25255 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
25257         PR middle-end/82004
25258         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
25259         after vectorization.
25261 2018-02-20  Martin Liska  <mliska@suse.cz>
25263         PR driver/83193
25264         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
25265         possible values if we don't have a hint.
25267 2018-02-20  Martin Liska  <mliska@suse.cz>
25269         PR c/84310
25270         PR target/79747
25271         * final.c (shorten_branches): Build align_tab array with one
25272         more element.
25273         * opts.c (finish_options): Add alignment option limit check.
25274         (MAX_CODE_ALIGN): Likewise.
25275         (MAX_CODE_ALIGN_VALUE): Likewise.
25276         * doc/invoke.texi: Document maximum allowed option value for
25277         all -falign-* options.
25279 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
25281         PR target/84146
25282         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
25283         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
25284         * var-tracking.c (emit_note_insn_var_location): Remove all references
25285         to NOTE_INSN_CALL_ARG_LOCATION.
25286         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
25287         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
25288         Use copy_rtx_if_shared.
25289         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
25290         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
25291         (dwarf2out_var_location): Remove handling of
25292         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
25293         on call_insn.
25294         * final.c (final_scan_insn): Remove all references to
25295         NOTE_INSN_CALL_ARG_LOCATION.
25296         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
25297         before dumping final insns.
25298         * except.c (emit_note_eh_region_end): Remove all references to
25299         NOTE_INSN_CALL_ARG_LOCATION.
25300         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
25301         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
25302         * config/arc/arc.c (hwloop_optimize): Likewise.
25303         * config/arm/arm.c (create_fix_barrier): Likewise.
25304         * config/s390/s390.c (s390_chunkify_start): Likewise.
25305         * config/sh/sh.c (find_barrier): Likewise.
25306         * config/i386/i386.c (rest_of_insert_endbranch,
25307         ix86_seh_fixup_eh_fallthru): Likewise.
25308         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
25309         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
25310         * config/frv/frv.c (frv_function_prologue): Likewise.
25311         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
25312         reg note.
25313         (note_outside_basic_block_p): Remove all references to
25314         NOTE_INSN_CALL_ARG_LOCATION.
25315         * gengtype.c (adjust_field_rtx_def): Likewise.
25316         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
25317         Likewise.
25318         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
25319         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
25321         PR c++/84444
25322         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
25323         is ADDR_EXPR.
25325         PR tree-optimization/84452
25326         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
25327         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
25328         is NULL.
25330 2018-02-19  Martin Liska  <mliska@suse.cz>
25332         PR sanitizer/82183
25333         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
25335 2018-02-19  Martin Liska  <mliska@suse.cz>
25336             Richard Sandiford  <richard.sandiford@linaro.org>
25338         PR tree-optimization/82491
25339         * gimple-fold.c (get_base_constructor): Make earlier bail out
25340         to prevent ubsan.
25342 2018-02-19  Carl Love  <cel@us.ibm.com>
25344         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
25345         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
25346         BU_P8V_OVERLOAD_1.
25347         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
25348         P8V_BUILTIN_VEC_NEG.
25350 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
25352         * config/rl78/rl78.md (movdf): New define expand.
25354 2018-02-19  Martin Liska  <mliska@suse.cz>
25356         PR other/80589
25357         * doc/invoke.texi: Fix typo.
25358         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
25360 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
25362         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
25363         handle rs6000_single_float and rs6000_double_float specially for
25364         e500 family CPUs.
25366 2018-02-16  Jeff Law  <law@redhat.com>
25368         * config/rx/rx.c (add_pop_cfi_notes): New function.;
25369         (pop_regs): Use it.
25371 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
25373         PR ipa/84425
25374         * ipa-inline.c (inline_small_functions): Fix a typo.
25376 2018-02-16  Nathan Sidwell  <nathan@acm.org>
25378         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
25380 2018-02-16  Carl Love  <cel@us.ibm.com>
25382         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
25383         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
25384         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
25385         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
25386         expansion to P8V_BUILTIN_VEC_FLOAT2.
25388 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
25390         PR rtl-optimization/70023
25391         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
25392         src_regno into account.
25394 2018-02-16  Carl Love  <cel@us.ibm.com>
25396         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
25397         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
25398         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
25399         * config/rs6000/rs6000.c: Remove case statements for
25400         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
25401         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
25402         and P9V_BUILTIN_VEC_VINSERT4B.
25403         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
25404         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
25405         * config/rs6000/vsx.md:
25406         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
25407         vec_insert4b.
25409 2018-02-16  Carl Love  <cel@us.ibm.com>
25411         * config/rs6000/altivec.h: Add builtin names vec_extract4b
25412         vec_insert4b.
25413         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
25414         definitions.
25415         * config/rs6000/rs6000-c.c: Add the definitions for
25416         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
25417         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
25418         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
25419         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
25420         definition for insert4b and define insn *insert3b_internal.
25421         * doc/extend.texi: Add documentation for vec_extract4b.
25423 2018-02-16  Nathan Sidwell  <nathan@acm.org>
25425         * doc/extend.texi (Backwards Compatibility): Mention friend
25426         injection.  Note for-scope is deprecated.
25427         * doc/invoke.texi (-ffriend-injection): Deprecate.
25429 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
25431         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
25432         that moved to I2, also allow destinations that are a paradoxical
25433         subreg (instead of a normal reg).
25435 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
25437         PR target/83831
25438         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
25439         to QImode.
25441 2018-02-16  Richard Biener  <rguenther@suse.de>
25443         PR tree-optimization/84037
25444         PR tree-optimization/84016
25445         PR target/82862
25446         * config/i386/i386.c (ix86_builtin_vectorization_cost):
25447         Adjust vec_construct for the fact we need additional higher latency
25448         128bit inserts for AVX256 and AVX512 vector builds.
25449         (ix86_add_stmt_cost): Scale vector construction cost for
25450         elementwise loads.
25452 2018-02-16  Richard Biener  <rguenther@suse.de>
25454         PR tree-optimization/84417
25455         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
25456         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
25457         (non_rewritable_lvalue_p): Likewise, use poly-ints.
25459 2018-02-16  Martin Liska  <mliska@suse.cz>
25461         PR sanitizer/84307
25462         * internal-fn.def (ASAN_CHECK): Set proper flags.
25463         (ASAN_MARK): Likewise.
25465 2018-02-16  Julia Koval  <julia.koval@intel.com>
25467         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
25468         from PTA_CANNONLAKE.
25470 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
25472         PR target/84272
25473         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
25474         Use ++iter rather than iter++ for std::list iterators.
25475         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
25476         defer deleting them until all nodes in the forest are processed.  Do
25477         free even leaf nodes.  Change to_process into auto_vec.
25479         PR bootstrap/84405
25480         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
25481         * vec.h (vec_default_construct): Use memset instead of placement new
25482         if BROKEN_VALUE_INITIALIZATION is defined.
25483         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
25484         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
25485         is defined.
25487         PR rtl-optimization/83723
25488         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
25489         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
25490         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
25491         recursive calls.
25492         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
25493         callers.
25494         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
25496 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
25498         PR rtl-optimization/81443
25499         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
25500         from inner REGs to paradoxical SUBREGs.
25502 2018-02-16  Richard Biener  <rguenther@suse.de>
25504         PR tree-optimization/84399
25505         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
25506         For operands we can analyze at their definition make sure we can
25507         analyze them at each use as well.
25509 2018-02-16  Richard Biener  <rguenther@suse.de>
25511         PR tree-optimization/84190
25512         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
25513         volatile accesses if the decl isn't volatile.
25515 2018-02-15  Jason Merrill  <jason@redhat.com>
25517         PR c++/84314 - ICE with templates and fastcall attribute.
25518         * attribs.c (build_type_attribute_qual_variant): Don't clobber
25519         TYPE_CANONICAL on an existing type.
25521 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
25523         PR tree-optimization/84383
25524         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
25525         dstoff nor call operand_equal_p if dstbase is NULL.
25527         PR tree-optimization/84334
25528         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
25529         also a CONSTANT_CLASS_P, punt.
25531 2018-02-14  Jim Wilson  <jimw@sifive.com>
25533         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
25534         first SMALL_OPERAND check.  New local min_second_step.  Move assert
25535         to where locals are set.  Add TARGET_RVC support.
25536         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
25538 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
25540         * doc/invoke.texi: Correct -Wformat-overflow code sample.
25542 2018-02-14  Martin Sebor  <msebor@redhat.com>
25544         PR tree-optimization/83698
25545         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
25546         arrays constrain the offset range to their bounds.
25547         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
25548         (builtin_access::overlap): Avoid setting the size of overlap if it's
25549         already been set.
25550         (maybe_diag_overlap): Also consider arrays when deciding what values
25551         of offsets to include in diagnostics.
25553 2018-02-14  Martin Sebor  <msebor@redhat.com>
25555         PR c/84108
25556         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
25557         that correspond to the kind of a declaration.
25559 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
25561         PR target/83984
25562         * config/pa/pa.md: Load address of PIC label using the linkage table
25563         if the label is nonlocal.
25565 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
25567         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
25568         warning message if user requests -maltivec=be.
25569         * doc/invoke.texi: Document deprecation of -maltivec=be.
25571 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
25573         PR target/84220
25574         * config/rs6000/rs6000-c.c: Update definitions for
25575         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
25576         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
25578 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
25580         PR target/84239
25581         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
25582         add _get_ssp intrinsics. Remove argument from
25583         __builtin_ia32_rdssp[d|q].
25584         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
25585         * config/i386/i386-builtin.def: Remove argument from
25586         __builtin_ia32_rdssp[d|q].
25587         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
25588         ix86_expand_special_args_builtin for _rdssp[d|q].
25589         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
25590         Clear register before usage.
25591         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
25592         Add documentation for new _get_ssp and _inc_ssp intrinsics.
25594 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
25596         PR tree-optimization/84357
25597         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
25598         operand 1 of an ARRAY_REF too.
25600 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
25602         PR target/83831
25603         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
25604         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25605         declarations.
25606         (set_of_reg): New struct.
25607         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
25608         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
25609         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
25610         functions.
25611         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
25612         Split into bitclr, bitset, bitinvert patterns if appropriate.
25613         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
25614         use rx_fuse_in_memory_bitop.
25615         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
25616         to named insn, correct maximum insn length.
25618 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
25620         PR target/79242
25621         * machmode.def: Define a complex mode for PARTIAL_INT.
25622         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
25623         MODE_PARTIAL_INT.
25624         * doc/rtl.texi: Document CSPImode.
25625         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
25626         handling.
25627         (msp430_hard_regno_nregs_with_padding): Likewise.
25629 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
25631         PR target/84279
25632         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
25634 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
25636         PR rtl-optimization/84169
25637         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
25638         we generated a parallel as new i3 and we split that to new i2 and i3
25639         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
25640         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
25641         those to i2, not i1.  Partially rewrite this scan code.
25643 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25645         PR c/82210
25646         * stor-layout.c (place_field): For variable length fields, adjust
25647         offset_align afterwards not just based on the field's alignment,
25648         but also on the size.
25650         PR middle-end/84309
25651         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
25652         of exps and logs in the use_exp2 case.
25654 2018-02-13  Jeff Law  <law@redhat.com>
25656         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
25657         entry for "vector".
25659         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
25660         ARGS as unused.
25662 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
25664         PR debug/84342
25665         PR debug/84319
25666         * common.opt (gas-loc-support, gas-locview-support): New.
25667         (ginline-points, ginternal-reset-location-views): New.
25668         * doc/invoke.texi: Document them.  Use @itemx where intended.
25669         (gvariable-location-views): Adjust.
25670         * target.def (reset_location_view): New.
25671         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
25672         (TARGET_RESET_LOCATION_VIEW): New.
25673         * doc/tm.texi: Rebuilt.
25674         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
25675         (dwarf2out_default_as_locview_support): New.
25676         (output_asm_line_debug_info): Use option variables.
25677         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
25678         (output_loc_list): Likewise.
25679         (add_high_low_attributes): Check option variables.
25680         Don't output entry view attribute in strict mode.
25681         (gen_inlined_subroutine_die): Check option variables.
25682         (dwarf2out_inline_entry): Likewise.
25683         (init_sections_and_labels): Likewise.
25684         (dwarf2out_early_finish): Likewise.
25685         (maybe_reset_location_view): New, from...
25686         (dwarf2out_var_location): ... here.  Call it.
25687         * debug.h (dwarf2out_default_as_loc_support): Declare.
25688         (dwarf2out_default_as_locview_support): Declare.
25689         * hooks.c (hook_int_rtx_insn_0): New.
25690         * hooks.h (hook_int_rtx_insn_0): Declare.
25691         * toplev.c (process_options): Take -gas-loc-support and
25692         -gas-locview-support from dwarf2out.  Enable
25693         -gvariable-location-views by default only with locview
25694         assembler support.  Enable -ginternal-reset-location-views by
25695         default only if the target defines the corresponding hook.
25696         Enable -ginline-points by default if location views are
25697         enabled; force it disabled if statement frontiers are
25698         disabled.
25699         * tree-inline.c (expand_call_inline): Check option variables.
25700         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
25702 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
25704         PR tree-optimization/84321
25705         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
25706         handling.  Also check whether the anti-range contains any values
25707         that satisfy the mask; switch to a VR_RANGE if not.
25709 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
25711         PR sanitizer/84340
25712         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
25714 2018-02-13  Martin Jambor  <mjambor@suse.cz>
25716         PR c++/83990
25717         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
25718         of call statements, also set location of a load to a temporary.
25720 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
25722         * config/rl78/rl78.c (add_vector_labels): New function.
25723         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
25724         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
25725         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
25726         which checks that no arguments are passed.
25727         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
25728         * doc/extend.texi: Documentation for the new attribute.
25730 2018-02-13  Andreas Schwab  <schwab@suse.de>
25732         * config/riscv/linux.h (CPP_SPEC): Define.
25734 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
25736         PR target/84335
25737         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
25738         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
25739         OPTION_MASK_ISA_AES as first argument to def_builtin_const
25740         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
25741         instead of OPTION_MASK_ISA_PCLMUL as first argument to
25742         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
25743         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
25744         temporarily for AES and PCLMUL builtins.
25746         PR tree-optimization/84339
25747         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
25748         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
25749         Formatting fixes.
25751         PR middle-end/84309
25752         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
25753         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
25754         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
25755         inline function.
25756         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
25757         inline function.
25758         * omp-simd-clone.h: New file.
25759         * omp-simd-clone.c: Include omp-simd-clone.h.
25760         (expand_simd_clones): No longer static.
25761         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
25762         cgraph.h and omp-simd-clone.h.
25763         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
25764         (vect_recog_widen_shift_pattern): Formatting fix.
25765         (vect_pattern_recog_1): Don't check optab for calls.
25767         PR target/84336
25768         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
25769         operands[2] into a REG before using gen_lowpart on it.
25771 2018-02-12  Jeff Law  <law@redhat.com>
25773         PR target/83760
25774         * config/sh/sh.c (find_barrier): Consider a sibling call
25775         a barrier as well.
25777         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
25778         successfully back substituting a reg.
25780 2018-02-12  Richard Biener  <rguenther@suse.de>
25782         PR tree-optimization/84037
25783         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
25784         parameter, move visited init to caller.
25785         (vect_slp_analyze_operations): Separate cost from validity
25786         check, initialize visited once for all instances.
25787         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
25788         for all instances.
25789         * tree-vect-stmts.c (vect_model_simple_cost): Make early
25790         out an assert.
25791         (vect_model_promotion_demotion_cost): Likewise.
25792         (vectorizable_bswap): Guard cost modeling with !slp_node
25793         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
25794         SLP stmts.
25795         (vectorizable_call): Likewise.
25796         (vectorizable_conversion): Likewise.
25797         (vectorizable_assignment): Likewise.
25798         (vectorizable_shift): Likewise.
25799         (vectorizable_operation): Likewise.
25800         (vectorizable_store): Likewise.
25801         (vectorizable_load): Likewise.
25802         (vectorizable_condition): Likewise.
25803         (vectorizable_comparison): Likewise.
25805 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
25807         PR sanitizer/84307
25808         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
25809         (ASAN_MARK): Fix fnspec to account for return value, change pointer
25810         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
25812 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
25814         PR middle-end/83665
25815         * params.def (inline-min-speedup): Increase from 8 to 15.
25816         (max-inline-insns-auto): Decrease from 40 to 30.
25817         * ipa-split.c (consider_split): Add some buffer for function to
25818         be considered inlining candidate.
25819         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
25820         default values.
25822 2018-02-12  Richard Biener  <rguenther@suse.de>
25824         PR tree-optimization/84037
25825         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
25826         matched stmts if we cannot swap the non-matched ones.
25828 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25830         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
25831         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
25832         _mm_maskz_scalef_round_ss): New intrinsics.
25833         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
25834         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
25835         __builtin_ia32_scalefss_round): Remove.
25836         (__builtin_ia32_scalefsd_mask_round,
25837         __builtin_ia32_scalefss_mask_round): New intrinsics.
25838         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
25839         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
25840         ((match_operand:VF_128 2 "<round_nimm_predicate>"
25841         "<round_constraint>")): Changed to ...
25842         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
25843         "<round_scalar_constraint>")): ... this.
25844         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
25845         %0, %1, %2<round_op3>}"): Changed to ...
25846         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
25847         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
25848         %2<round_scalar_mask_op3>}"): ... this.
25849         * config/i386/subst.md (round_scalar_nimm_predicate): New.
25851 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
25853         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
25854         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
25855         (_mm_maskz_sqrt_round_ss): New intrinsics.
25856         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
25857         (__builtin_ia32_sqrtsd_mask_round)
25858         (__builtin_ia32_sqrtss_mask_round): New builtins.
25859         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
25860         (__builtin_ia32_sqrtss_round): Remove.
25861         (__builtin_ia32_sqrtsd_mask_round)
25862         (__builtin_ia32_sqrtss_mask_round): New builtins.
25863         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
25864         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
25865         ((match_operand:VF_128 1 "vector_operand"
25866         "xBm,<round_constraint>")): Changed to ...
25867         ((match_operand:VF_128 1 "vector_operand"
25868         "xBm,<round_scalar_constraint>")): ... this.
25869         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
25870         %0, %2, %<iptr>1<round_op3>}): Changed to ...
25871         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
25872         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
25873         %<iptr>1<round_scalar_mask_op3>}): ... this.
25874         ((set_attr "prefix" "<round_prefix>")): Changed to ...
25875         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
25877 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
25879         PR target/84266
25880         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
25881         Cast vec_cmpeq result to correct type.
25882         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
25883         Cast vec_cmpgt result to correct type.
25885 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
25887         * final.c (final_scan_insn_1): Renamed from...
25888         (final_scan_insn): ... this.  New wrapper, to recover
25889         seen from the outermost call in recursive ones.
25890         * config/sparc/sparc.c (output_return): Drop seen from call.
25891         (output_sibcall): Likewise.
25892         * config/visium/visium.c (output_branch): Likewise.
25894 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
25896         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
25897         function label.
25899 2018-02-10  Alan Modra  <amodra@gmail.com>
25901         PR target/84300
25902         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
25903         Specify LR as an input.
25905 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
25907         PR sanitizer/83987
25908         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
25909         remove_member_access_dummy_vars): New functions.
25910         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
25911         lower_omp_1, execute_lower_omp): Use them.
25913         PR rtl-optimization/84308
25914         * shrink-wrap.c (spread_components): Release todo vector.
25916 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
25918         PR rtl-optimization/57193
25919         * ira-color.c (struct allocno_color_data): Add member
25920         conflict_allocno_hard_prefs.
25921         (update_conflict_allocno_hard_prefs): New.
25922         (bucket_allocno_compare_func): Add a preference based on
25923         conflict_allocno_hard_prefs.
25924         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
25925         (color_allocnos): Remove a dead code.  Initiate
25926         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
25928 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25930         PR target/84226
25931         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
25932         constraint from =wa to wa.  Avoid a subreg on the output operand,
25933         instead use a pseudo and subreg it in a move.
25934         (p9_xxbrd_<mode>): Changed to ...
25935         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
25936         (p9_xxbrd_v2df): New expander.
25937         (p9_xxbrw_<mode>): Changed to ...
25938         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
25939         (p9_xxbrw_v4sf): New expander.
25941 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25943         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
25945 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
25947         PR target/83926
25948         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
25949         multiply in 32-bit mode.
25950         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
25951         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
25952         mode.
25954 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
25956         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
25957         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
25958         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
25959         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
25961 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
25963         PR lto/84213
25964         * dwarf2out.c (is_trivial_indirect_ref): New function.
25965         (dwarf2out_late_global_decl): Do not generate a location
25966         attribute for variables that have a non-trivial DECL_VALUE_EXPR
25967         and that are not defined in the current unit.
25969 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25971         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
25972         instead of a libcall for UNORDERED.
25974 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
25976         PR target/82641
25977         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
25978         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
25980 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
25982         PR target/PR84295
25983         * config/s390/s390.c (s390_set_current_function): Invoke
25984         s390_indirect_branch_settings also if fndecl didn't change.
25986 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
25988         * config/rs6000/rs6000.md (blockage): Set length to zero.
25990 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
25992         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
25994 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
25996         PR sanitizer/84285
25997         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
25998         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
25999         -static-lib*san.
26001         PR debug/84252
26002         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
26003         PARALLEL incoming that failed vt_get_decl_and_offset check.
26005         PR middle-end/84237
26006         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
26007         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
26008         TREE_READONLY bit.
26009         (get_variable_section): For decls in named .bss* sections pass true as
26010         second argument to bss_initializer_p.
26012 2018-02-09  Marek Polacek  <polacek@redhat.com>
26013             Jakub Jelinek  <jakub@redhat.com>
26015         PR c++/83659
26016         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
26017         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
26018         Sync some changes from cxx_fold_indirect_ref.
26020 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
26022         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
26023         markers.
26024         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
26025         (BLOCK_INLINE_ENTRY_LABEL): New.
26026         (dwarf2out_var_location): Disregard inline entry markers.
26027         (inline_entry_data): New struct.
26028         (inline_entry_data_hasher): New hashtable type.
26029         (inline_entry_data_hasher::hash): New.
26030         (inline_entry_data_hasher::equal): New.
26031         (inline_entry_data_table): New variable.
26032         (add_high_low_attributes): Add DW_AT_entry_pc and
26033         DW_AT_GNU_entry_view attributes if a pending entry is found
26034         in inline_entry_data_table.  Add old entry_pc attribute only
26035         if debug nonbinding markers are disabled.
26036         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
26037         markers are enabled.
26038         (block_within_block_p, dwarf2out_inline_entry): New.
26039         (dwarf2out_finish): Check that no entries remained in
26040         inline_entry_data_table.
26041         * final.c (reemit_insn_block_notes): Handle inline entry notes.
26042         (final_scan_insn, notice_source_line): Likewise.
26043         (rest_of_clean_state): Skip inline entry markers.
26044         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
26045         markers.
26046         * gimple.c (gimple_build_debug_inline_entry): New.
26047         * gimple.h (enum gimple_debug_subcode): Add
26048         GIMPLE_DEBUG_INLINE_ENTRY.
26049         (gimple_build_debug_inline_entry): Declare.
26050         (gimple_debug_inline_entry_p): New.
26051         (gimple_debug_nonbind_marker_p): Adjust.
26052         * insn-notes.def (INLINE_ENTRY): New.
26053         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
26054         inline entry marker notes.
26055         (print_insn): Likewise.
26056         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
26057         (INSN_DEBUG_MARKER_KIND): Likewise.
26058         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
26059         * tree-inline.c (expand_call_inline): Build and insert
26060         debug_inline_entry stmt.
26061         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
26062         inline entry blocks early, if nonbind markers are enabled.
26063         (dump_scope_block): Dump fragment info.
26064         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
26065         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
26066         (gimple_build_debug_inline_entry): New.
26067         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
26068         Enable/disable inline entry points too.
26069         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
26070         (DEBUG_INSN): Describe inline entry markers.
26072         * common.opt (gvariable-location-views): New.
26073         (gvariable-location-views=incompat5): New.
26074         * config.in: Rebuilt.
26075         * configure: Rebuilt.
26076         * configure.ac: Test assembler for view support.
26077         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
26078         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
26079         * dwarf2out.c (var_loc_view): New typedef.
26080         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
26081         (dwarf2out_locviews_in_attribute): New.
26082         (dwarf2out_locviews_in_loclist): New.
26083         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
26084         (enum dw_line_info_opcode): Add LI_adv_address.
26085         (struct dw_line_info_table): Add view.
26086         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
26087         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
26088         (zero_view_p): New variable.
26089         (ZERO_VIEW_P): New macro.
26090         (output_asm_line_debug_info): New.
26091         (struct var_loc_node): Add view.
26092         (add_AT_view_list, AT_loc_list): New.
26093         (add_var_loc_to_decl): Add view param.  Test it against last.
26094         (new_loc_list): Add view params.  Record them.
26095         (AT_loc_list_ptr): Handle loc and view lists.
26096         (view_list_to_loc_list_val_node): New.
26097         (print_dw_val): Handle dw_val_class_view_list.
26098         (size_of_die): Likewise.
26099         (value_format): Likewise.
26100         (loc_list_has_views): New.
26101         (gen_llsym): Set vl_symbol too.
26102         (maybe_gen_llsym, skip_loc_list_entry): New.
26103         (dwarf2out_maybe_output_loclist_view_pair): New.
26104         (output_loc_list): Output view list or entries too.
26105         (output_view_list_offset): New.
26106         (output_die): Handle dw_val_class_view_list.
26107         (output_dwarf_version): New.
26108         (output_compilation_unit_header): Use it.
26109         (output_skeleton_debug_sections): Likewise.
26110         (output_rnglists, output_line_info): Likewise.
26111         (output_pubnames, output_aranges): Update version comments.
26112         (output_one_line_info_table): Output view numbers in asm comments.
26113         (dw_loc_list): Determine current endview, pass it to new_loc_list.
26114         Call maybe_gen_llsym.
26115         (loc_list_from_tree_1): Adjust.
26116         (add_AT_location_description): Create view list attribute if
26117         needed, check it's absent otherwise.
26118         (convert_cfa_to_fb_loc_list): Adjust.
26119         (maybe_emit_file): Call output_asm_line_debug_info for test.
26120         (dwarf2out_var_location): Reset views as needed.  Precompute
26121         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
26122         attribute.  Set view.
26123         (new_line_info_table): Reset next view.
26124         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
26125         (dwarf2out_source_line): Likewise.  Output view resets and labels to
26126         the assembler, or select appropriate line info opcodes.
26127         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
26128         (optimize_string_length): Catch it.  Adjust.
26129         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
26130         dw_val_class_view_list, and remove it if no longer needed.
26131         (hash_loc_list): Hash view numbers.
26132         (loc_list_hasher::equal): Compare them.
26133         (optimize_location_lists): Check whether a view list symbol is
26134         needed, and whether the locview attribute is present, and
26135         whether they match.  Remove the locview attribute if no longer
26136         needed.
26137         (index_location_lists): Call skip_loc_list_entry for test.
26138         (dwarf2out_finish): Call output_asm_line_debug_info for test.
26139         Use output_dwarf_version.
26140         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
26141         (struct dw_val_node): Add val_view_list.
26142         * final.c (SEEN_NEXT_VIEW): New.
26143         (set_next_view_needed): New.
26144         (clear_next_view_needed): New.
26145         (maybe_output_next_view): New.
26146         (final_start_function): Rename to...
26147         (final_start_function_1): ... this.  Take pointer to FIRST,
26148         add SEEN parameter.  Emit param bindings in the initial view.
26149         (final_start_function): Reintroduce SEEN-less interface.
26150         (final): Rename to...
26151         (final_1): ... this.  Take SEEN parameter.  Output final pending
26152         next view at the end.
26153         (final): Reintroduce seen-less interface.
26154         (final_scan_insn): Output pending next view before switching
26155         sections or ending a block.  Mark the next view as needed when
26156         outputting variable locations.  Notify debug backend of section
26157         changes, and of location view changes.
26158         (rest_of_handle_final): Adjust.
26159         * toplev.c (process_options): Autodetect value for debug variable
26160         location views option.  Warn on incompat5 without -gdwarf-5.
26161         * doc/invoke.texi (gvariable-location-views): New.
26162         (gvariable-location-views=incompat5): New.
26163         (gno-variable-location-views): New.
26165 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
26167         PR tree-optimization/84136
26168         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
26169         that the result of find_edge is non-NULL.
26171 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
26173         PR target/83008
26174         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
26175         storing integer register in SImode.  Fix cost of 256 and 512
26176         byte aligned SSE register store.
26178 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
26180         * config/i386/i386.c (ix86_multiplication_cost): Fix
26181         multiplication cost for TARGET_AVX512DQ.
26183 2018-02-08  Marek Polacek  <polacek@redhat.com>
26185         PR tree-optimization/84238
26186         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
26187         get_range_strlen.
26189 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26191         PR tree-optimization/84265
26192         * tree-vect-stmts.c (vectorizable_store): Don't treat
26193         VMAT_CONTIGUOUS accesses as grouped.
26194         (vectorizable_load): Likewise.
26196 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26198         PR tree-optimization/81635
26199         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
26200         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
26201         (test_round_for_mask): New functions.
26202         (wide_int_cc_tests): Call test_round_for_mask.
26203         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
26204         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
26205         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
26206         range returned by get_range_info.
26208 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
26210         PR ipa/81360
26211         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
26212         * symtab.c: Include builtins.h
26213         (symtab_node::output_to_lto_symbol_table_p): Move here
26214         from lto-streamer-out.c:output_symbol_p.
26215         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
26216         (output_symbol_p): Move all logic to symtab.c
26217         (produce_symtab): Update.
26219 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26221         * config/s390/s390-opts.h (enum indirect_branch): Define.
26222         * config/s390/s390-protos.h (s390_return_addr_from_memory)
26223         (s390_indirect_branch_via_thunk)
26224         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
26225         (enum s390_indirect_branch_type): Define.
26226         * config/s390/s390.c (struct s390_frame_layout, struct
26227         machine_function): Remove.
26228         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
26229         (indirect_branch_table_label_no, indirect_branch_table_name):
26230         Define variables.
26231         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
26232         (enum s390_indirect_branch_option): Define.
26233         (s390_return_addr_from_memory): New function.
26234         (s390_handle_string_attribute): New function.
26235         (s390_attribute_table): Add new attribute handler.
26236         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
26237         (s390_indirect_branch_via_thunk): New function.
26238         (s390_indirect_branch_via_inline_thunk): New function.
26239         (s390_function_ok_for_sibcall): When jumping via thunk disallow
26240         sibling call optimization for non z10 compiles.
26241         (s390_emit_call): Force indirect branch target to be a single
26242         register.  Add r1 clobber for non-z10 compiles.
26243         (s390_emit_epilogue): Emit return jump via return_use expander.
26244         (s390_reorg): Handle JUMP_INSNs as execute targets.
26245         (s390_option_override_internal): Perform validity checks for the
26246         new command line options.
26247         (s390_indirect_branch_attrvalue): New function.
26248         (s390_indirect_branch_settings): New function.
26249         (s390_set_current_function): Invoke s390_indirect_branch_settings.
26250         (s390_output_indirect_thunk_function):  New function.
26251         (s390_code_end): Implement target hook.
26252         (s390_case_values_threshold): Implement target hook.
26253         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
26254         macros.
26255         * config/s390/s390.h (struct s390_frame_layout)
26256         (struct machine_function): Move here from s390.c.
26257         (TARGET_INDIRECT_BRANCH_NOBP_RET)
26258         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
26259         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
26260         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
26261         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
26262         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
26263         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
26264         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
26265         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
26266         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
26267         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
26268         (mnemonic attribute): Add values which aren't recognized
26269         automatically.
26270         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
26271         pattern for branch conversion.  Fix mnemonic attribute.
26272         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
26273         indirect branch via thunk if requested.
26274         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
26275         ("*indirect_jump"): Disable for branch conversion using out of
26276         line thunks.
26277         ("indirect_jump_via_thunk<mode>_z10")
26278         ("indirect_jump_via_thunk<mode>")
26279         ("indirect_jump_via_inlinethunk<mode>_z10")
26280         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
26281         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
26282         ("casesi_jump_via_inlinethunk<mode>_z10")
26283         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
26284         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
26285         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
26286         ("*indirect2_jump"): Disable for branch conversion.
26287         ("casesi_jump"): Turn into expander and expand patterns for branch
26288         conversion.
26289         ("return_use"): New expander.
26290         ("*return"): Emit return via thunk and rename it to ...
26291         ("*return<mode>"): ... this one.
26292         * config/s390/s390.opt: Add new options and and enum for the
26293         option values.
26295 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
26297         * lra-constraints.c (match_reload): Unconditionally use
26298         gen_lowpart_SUBREG, rather than selecting between that
26299         and equivalent gen_rtx_SUBREG code.
26301 2018-02-08  Richard Biener  <rguenther@suse.de>
26303         PR tree-optimization/84233
26304         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
26305         changed flag instead of boguously re-using phi_inserted.
26307 2018-02-08  Martin Jambor  <mjambor@suse.cz>
26309         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
26310         static local variables.
26312 2018-02-08  Richard Biener  <rguenther@suse.de>
26314         PR tree-optimization/84278
26315         * tree-vect-stmts.c (vectorizable_store): When looking for
26316         smaller vector types to perform grouped strided loads/stores
26317         make sure the mode is supported by the target.
26318         (vectorizable_load): Likewise.
26320 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26322         * config/aarch64/aarch64.c (aarch64_components_for_bb):
26323         Increase LDP/STP opportunities by adding adjacent callee-saves.
26325 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
26327         PR rtl-optimization/84068
26328         PR rtl-optimization/83459
26329         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
26331 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
26333         PR tree-optimization/84224
26334         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
26335         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
26336         non-zero arguments.
26338 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
26340         PR target/84113
26341         * config/rs6000/altivec.md (*restore_world): Remove LR use.
26342         * config/rs6000/predicates.md (restore_world_operation): Adjust op
26343         count, remove one USE.
26345 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
26347         * doc/install.texi (Configuration): Document the
26348         --with-long-double-format={ibm,ieee} PowerPC configuration
26349         options.
26351         PR target/84154
26352         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
26353         Convert from define_expand to be define_insn_and_split.  Rework
26354         float/double/_Float128 conversions to QI/HI/SImode to work with
26355         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
26356         conversions to QI/HImode types did a store and then a load to
26357         truncate the value.  For conversions to VSX registers, don't split
26358         the insn, instead emit the code directly.  Use the code iterator
26359         any_fix to combine signed and unsigned conversions.
26360         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
26361         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
26362         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26363         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
26364         (fix_<mode>di2_hw): Likewise.
26365         (fixuns_<mode>di2_hw): Likewise.
26366         (fix_<mode>si2_hw): Likewise.
26367         (fixuns_<mode>si2_hw): Likewise.
26368         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
26369         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
26370         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
26371         fix<uns>_trunc<SFDF:mode>si2_p8.
26372         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
26373         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
26374         (fix<uns>_<mode>_mem): Likewise.
26375         (fctiw<u>z_<mode>_mem): Likewise.
26376         (fix<uns>_<mode>_mem): Likewise.
26377         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
26378         the register allocator from doing a direct move to the GPRs to do
26379         a store, and instead use the ISA 3.0 store byte/half-word from
26380         vector register instruction.  For IEEE 128-bit floating point,
26381         also optimize stores of 32-bit ints.
26382         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
26384 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
26386         * genextract.c (push_pathstr_operand): New function to support
26387         [a-zA-Z].
26388         (walk_rtx): Call push_pathstr_operand.
26389         (print_path): Support [a-zA-Z].
26391 2018-02-07  Richard Biener  <rguenther@suse.de>
26393         PR tree-optimization/84037
26394         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
26395         (cse_and_gimplify_to_preheader): Declare.
26396         (vect_get_place_in_interleaving_chain): Likewise.
26397         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
26398         ivexpr_map.
26399         (_loop_vec_info::~_loop_vec_info): Delete it.
26400         (cse_and_gimplify_to_preheader): New function.
26401         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
26402         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
26403         (vectorizable_load): Likewise.  For grouped stores always base
26404         the IV on the first element.
26405         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
26406         condition before gimplifying.
26408 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
26410         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
26411         *DIV_EXPR and *MOD_EXPR.
26413 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
26415         PR target/84248
26416         * config/i386/i386.c (ix86_option_override_internal): Mask out
26417         the CF_SET bit when checking -fcf-protection.
26419 2018-02-07  Tom de Vries  <tom@codesourcery.com>
26421         PR libgomp/84217
26422         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
26423         enough.
26425 2018-02-07  Richard Biener  <rguenther@suse.de>
26427         PR tree-optimization/84204
26428         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
26429         this place.
26431         PR tree-optimization/84205
26432         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
26433         special-case isl_ast_op_zdiv_r.
26435         PR tree-optimization/84223
26436         * graphite-scop-detection.c (gather_bbs::before_dom_children):
26437         Only add conditions from within the region.
26438         (gather_bbs::after_dom_children): Adjust.
26440 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
26442         PR target/84209
26443         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
26444         * config/avr/avr.md: Only post-reload split REG-REG moves if
26445         either register is GENERAL_REG_P.
26447 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
26449         PR tree-optimization/84235
26450         * tree-ssa-scopedtables.c
26451         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
26452         if the subtraction is performed in floating point type where NaNs are
26453         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
26454         build 1.  Formatting fix.
26456 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
26458         PR target/84146
26459         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
26460         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
26461         and skip it regardless of bb boundaries.  Use CALL_P macro,
26462         don't test INSN_P (insn) together with CALL_P or JUMP_P check
26463         unnecessarily, formatting fix.
26465 2018-02-06  Michael Collison  <michael.collison@arm.com>
26467         * config/arm/thumb2.md:
26468         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
26469         (*thumb_mov_notscc): Ditto.
26471 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
26473         PR target/84154
26474         * config/rs6000/rs6000.md (su code attribute): Use "u" for
26475         unsigned_fix, not "s".
26477 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26479         * configure.ac (gcc_fn_eh_frame_ro): New function.
26480         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
26481         correct .eh_frame permissions.
26482         * configure: Regenerate.
26484 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
26486         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
26487         irrelevant options.
26489 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26491         * config/rs6000/rs6000.c (rs6000_option_override_internal):
26492         Display warning message for -mno-speculate-indirect-jumps.
26494 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
26496         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
26497         Undocumented.
26498         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
26500 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
26502         PR tree-optimization/84225
26503         * tree-eh.c (find_trapping_overflow): Only call
26504         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
26506 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26508         PR target/84145
26509         * config/i386/i386.c: Reimplement the check of possible options
26510         -mibt/-mshstk conbination. Change error messages.
26511         * doc/invoke.texi: Fix a typo: remove extra '='.
26513 2018-02-06  Marek Polacek  <polacek@redhat.com>
26515         PR tree-optimization/84228
26516         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
26518 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
26520         PR target/82641
26521         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
26522         emitted arch directives.
26523         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
26524         __ARM_FEATURE_COPROC before changing architectures.
26526 2018-02-06  Richard Biener  <rguenther@suse.de>
26528         * config/i386/i386.c (print_reg): Fix typo.
26529         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
26531 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
26533         * configure: Regenerate.
26535 2018-02-05  Martin Sebor  <msebor@redhat.com>
26537         PR tree-optimization/83369
26538         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
26539         inlining context.
26541 2018-02-05  Martin Liska  <mliska@suse.cz>
26543         * doc/invoke.texi: Cherry-pick upstream r323995.
26545 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
26547         * ira.c (ira_init_register_move_cost): Adjust comment.
26549 2018-02-05  Martin Liska  <mliska@suse.cz>
26551         PR gcov-profile/84137
26552         * doc/gcov.texi: Fix typo in documentation.
26554 2018-02-05  Martin Liska  <mliska@suse.cz>
26556         PR gcov-profile/83879
26557         * doc/gcov.texi: Document necessity of --dynamic-list-data when
26558         using dlopen functionality.
26560 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
26562         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
26563         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
26564         _mm_maskz_range_ss, _mm_mask_range_round_ss,
26565         _mm_maskz_range_round_ss): New intrinsics.
26566         (__builtin_ia32_rangesd128_round)
26567         (__builtin_ia32_rangess128_round): Remove.
26568         (__builtin_ia32_rangesd128_mask_round,
26569         __builtin_ia32_rangess128_mask_round): New builtins.
26570         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
26571         __builtin_ia32_rangess128_round): Remove.
26572         (__builtin_ia32_rangesd128_mask_round,
26573         __builtin_ia32_rangess128_mask_round): New builtins.
26574         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
26575         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
26576         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
26577         "<round_saeonly_constraint>")): Changed to ...
26578         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
26579         "<round_saeonly_scalar_constraint>")): ... this.
26580         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
26581         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
26582         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
26583         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
26584         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
26586 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
26588         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
26589         options.
26590         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
26591         Remove all values except native, 8540 and 8548.
26593 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
26595         * config/i386/i386.c (ix86_output_function_return): Pass
26596         INVALID_REGNUM, instead of -1, as invalid register number to
26597         indirect_thunk_name and output_indirect_thunk.
26599 2018-02-02  Julia Koval  <julia.koval@intel.com>
26601         * config.gcc: Add -march=icelake.
26602         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
26603         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
26604         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
26605         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
26606         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
26607         (processor_target_table): Add icelake.
26608         (ix86_option_override_internal): Handle new PTAs.
26609         (get_builtin_code_for_version): Handle icelake.
26610         (M_INTEL_COREI7_ICELAKE): New.
26611         (fold_builtin_cpu): Handle icelake.
26612         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
26613         * doc/invoke.texi: Add -march=icelake.
26615 2018-02-02  Julia Koval  <julia.koval@intel.com>
26617         * config/i386/i386.c (ix86_option_override_internal): Change flags type
26618         to wide_int_bitmask.
26619         * wide-int-bitmask.h: New.
26621 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
26623         PR target/84066
26624         * config/i386/i386.md: Replace Pmode with word_mode in
26625         builtin_setjmp_setup and builtin_longjmp to support x32.
26627 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
26629         PR target/56010
26630         PR target/83743
26631         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
26632         #include "opts.h".
26633         (rs6000_supported_cpu_names): New static variable.
26634         (linux_cpu_translation_table): Likewise.
26635         (elf_platform) <cpu>: Define new static variable and use it.
26636         Translate kernel AT_PLATFORM name to canonical name if needed.
26637         Error if platform name is unknown.
26639 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
26641         PR target/84089
26642         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
26644 2018-02-01  Jeff Law  <law@redhat.com>
26646         PR target/84128
26647         * config/i386/i386.c (release_scratch_register_on_entry): Add new
26648         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
26649         the scratch if RELEASE_VIA_POP is false.
26650         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
26651         If we have to save a temporary register, decrement SIZE appropriately.
26652         Pass new arguments to release_scratch_register_on_entry.
26653         (ix86_adjust_stack_and_probe): Likewise.
26654         (ix86_emit_probe_stack_range): Pass new arguments to
26655         release_scratch_register_on_entry.
26657 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
26659         PR rtl-optimization/84157
26660         * combine.c (change_zero_ext): Use REG_P predicate in
26661         front of HARD_REGISTER_P predicate.
26663 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
26665         * config/avr/avr.c (avr_option_override): Move disabling of
26666         -fdelete-null-pointer-checks to...
26667         * common/config/avr/avr-common.c (avr_option_optimization_table):
26668         ...here.
26670 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26672         PR tree-optimization/81635
26673         * tree-data-ref.c (split_constant_offset_1): For types that
26674         wrap on overflow, try to use range info to prove that wrapping
26675         cannot occur.
26677 2018-02-01  Renlin Li  <renlin.li@arm.com>
26679         PR target/83370
26680         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
26681         TAILCALL_ADDR_REGS.
26682         (aarch64_register_move_cost): Likewise.
26683         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
26684         TAILCALL_ADDR_REGS.
26685         (REG_CLASS_NAMES): Likewise.
26686         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
26687         TAILCALL_ADDR_REGS. Remove IP registers.
26688         * config/aarch64/aarch64.md (Ucs): Update register constraint.
26690 2018-02-01  Richard Biener  <rguenther@suse.de>
26692         * domwalk.h (dom_walker::dom_walker): Add additional constructor
26693         for specifying RPO order and allow NULL for that.
26694         * domwalk.c (dom_walker::dom_walker): Likewise.
26695         (dom_walker::walk): Handle NULL RPO order.
26696         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
26697         in RPO order.
26698         (rewrite_update_dom_walker): Likewise.
26699         (mark_def_dom_walker): Likewise.
26701 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26703         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
26704         (aarch64_maybe_expand_sve_subreg_move): Declare.
26705         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
26706         * config/aarch64/predicates.md (aarch64_any_register_operand): New
26707         predicate.
26708         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
26709         that are semantically a reverse operation.
26710         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
26711         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
26712         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
26713         functions.
26714         (aarch64_can_change_mode_class): For big-endian, forbid changes
26715         between two SVE modes if they have different element sizes.
26717 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26719         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
26720         the TImode handling for big-endian targets.
26722 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26724         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
26725         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
26726         not just bytes.
26727         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
26728         Remove BSWAP handing for big-endian targets and use the form of
26729         LD1RQ appropariate for the mode.
26731 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26733         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
26734         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
26735         duplicated element.
26737 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
26739         PR tearget/83845
26740         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
26741         check for operands that need to go through aarch64_sve_reload_be.
26743 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
26745         PR tree-optimization/81661
26746         PR tree-optimization/84117
26747         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
26748         * tree-eh.c: Include gimplify.h.
26749         (find_trapping_overflow, replace_trapping_overflow,
26750         rewrite_to_non_trapping_overflow): New functions.
26751         * tree-vect-loop.c: Include tree-eh.h.
26752         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
26753         * tree-data-ref.c: Include tree-eh.h.
26754         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
26756 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
26758         PR rtl-optimization/84123
26759         * combine.c (change_zero_ext): Check if hard register satisfies
26760         can_change_dest_mode before calling gen_lowpart_SUBREG.
26762 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
26764         PR target/82444
26765         * ira.c (ira_init_register_move_cost): Remove assert.
26767 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26769         PR rtl-optimization/84071
26770         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
26771         * doc/tm.texi: Regenerate.
26773 2018-01-31  Richard Biener  <rguenther@suse.de>
26775         PR tree-optimization/84132
26776         * tree-data-ref.c (analyze_miv_subscript): Properly
26777         check whether evolution_function_is_affine_multivariate_p
26778         before calling gcd_of_steps_may_divide_p.
26780 2018-01-31  Julia Koval  <julia.koval@intel.com>
26782         PR target/83618
26783         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
26784         * config/i386/i386.md (rdpid_rex64) New.
26785         (rdpid): Make 32bit only.
26787 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
26789         PR lto/84105
26790         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
26791         an IDENTIFIER_NODE for FUNCTION_TYPE's.
26793 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26795         Revert
26796         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
26798         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
26800 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
26802         PR rtl-optimization/84071
26803         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
26804         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
26806 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26808         * config/arc/arc.c (arc_handle_aux_attribute): New function.
26809         (arc_attribute_table): Add 'aux' attribute.
26810         (arc_in_small_data_p): Consider aux like variables.
26811         (arc_is_aux_reg_p): New function.
26812         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
26813         (arc_get_aux_arg): New function.
26814         (prepare_move_operands): Handle aux-register access.
26815         (arc_handle_aux_attribute): New function.
26816         * doc/extend.texi (ARC Variable attributes): Add subsection.
26818 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
26820         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
26821         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
26822         (arc_attribute_table): Add 'uncached' attribute.
26823         (arc_print_operand): Print '.di' flag for uncached memory
26824         accesses.
26825         (arc_in_small_data_p): Do not consider for small data the uncached
26826         types.
26827         (arc_is_uncached_mem_p): New function.
26828         * config/arc/predicates.md (compact_store_memory_operand): Check
26829         for uncached memory accesses.
26830         (nonvol_nonimm_operand): Likewise.
26831         * doc/extend.texi (ARC Type Attribute): New subsection.
26833 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
26835         PR c/84100
26836         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
26837         falign-loops=): Add Optimization flag.
26839 2018-01-30  Jeff Law  <law@redhat.com>
26841         PR target/84064
26842         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
26843         INT_REGISTERS_SAVED.  Check it prior to calling
26844         get_scratch_register_on_entry.
26845         (ix86_adjust_stack_and_probe): Similarly.
26846         (ix86_emit_probe_stack_range): Similarly.
26847         (ix86_expand_prologue): Corresponding changes.
26849 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26851         PR target/40411
26852         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
26853         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
26855 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
26857         PR target/84112
26858         * lra-constraints.c (curr_insn_transform): Process AND in the
26859         address.
26861 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
26863         PR rtl-optimization/83986
26864         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
26865         dependence against last_pending_memory_flush in addition to
26866         pending_jump_insns.
26868 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
26870         PR tree-optimization/81611
26871         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
26872         copies.
26874 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26876         PR target/83758
26877         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
26878         a reg rtx.
26880 2018-01-30  Richard Biener  <rguenther@suse.de>
26881             Jakub Jelinek  <jakub@redhat.com>
26883         PR tree-optimization/84111
26884         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
26885         inner loops added during recursion, as they don't have up-to-date
26886         SSA form.
26888 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26890         PR ipa/81360
26891         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
26892         (can_inline_edge_by_limits_p): ... here.
26893         (can_early_inline_edge_p, check_callers,
26894         update_caller_keys, update_callee_keys, recursive_inlining,
26895         add_new_edges_to_heap, speculation_useful_p,
26896         inline_small_functions,
26897         inline_small_functions, flatten_function,
26898         inline_to_all_callers_1): Update.
26900 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
26902         * profile-count.c (profile_count::combine_with_ipa_count): Handle
26903         zeros correctly.
26905 2018-01-30  Richard Biener  <rguenther@suse.de>
26907         PR tree-optimization/83008
26908         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
26909         invariant and constant vector uses in stmts when they need
26910         more than one stmt.
26912 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
26914         PR bootstrap/84017
26915         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
26916         * configure: Regenerate.
26918 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26920         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
26921         pattern.
26922         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
26923         Use gen_rtx_REG rather than gen_lowpart.
26925 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26927         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
26928         rather than 0 when creating partial subregs.
26930 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
26932         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
26933         of usage.
26935 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
26937         PR target/81550
26938         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
26939         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
26940         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
26941         flags.  This restores the settings used before the 2017-07-24.
26942         Turning off pre increment/decrement/modify allows IVOPTS to
26943         optimize DF/SF loops where the index is an int.
26945 2018-01-29  Richard Biener  <rguenther@suse.de>
26946             Kelvin Nilsen  <kelvin@gcc.gnu.org>
26948         PR bootstrap/80867
26949         * tree-vect-stmts.c (vectorizable_call): Don't call
26950         targetm.vectorize_builtin_md_vectorized_function if callee is
26951         NULL.
26953 2018-01-22  Carl Love  <cel@us.ibm.com>
26955         * doc/extend.tex: Fix typo in second arg in
26956         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
26958 2018-01-29  Richard Biener  <rguenther@suse.de>
26960         PR tree-optimization/84086
26961         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
26962         (flush_ssaname_freelist): When SSA names were released reset
26963         the SCEV hash table.
26965 2018-01-29  Richard Biener  <rguenther@suse.de>
26967         PR tree-optimization/84057
26968         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
26969         removed paths when removing edges.
26971 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
26973         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
26974         -mfunction-return=@var{choice}.
26976 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
26978         PR diagnostic/84034
26979         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
26980         Handle CR like TAB.
26981         (layout::print_source_line): Likewise.
26982         (test_get_line_width_without_trailing_whitespace): Add test cases.
26984 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
26986         PR middle-end/84040
26987         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
26988         debug insns.
26990 2018-01-26  Jim Wilson  <jimw@sifive.com>
26992         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
26994         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
26995         specified.
26997 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26999         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
27000         and CMP + SUB-immediate -> SUBS.
27002 2018-01-26  Martin Sebor  <msebor@redhat.com>
27004         PR tree-optimization/83896
27005         * tree-ssa-strlen.c (get_string_len): Rename...
27006         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
27007         Avoid assuming length is constant.
27008         (handle_char_store): Use HOST_WIDE_INT for string length.
27010 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
27012         PR target/81763
27013         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
27014         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
27016 2018-01-26  Richard Biener  <rguenther@suse.de>
27018         PR rtl-optimization/84003
27019         * dse.c (record_store): Only record redundant stores when
27020         the earlier store aliases at least all accesses the later one does.
27022 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
27024         PR rtl-optimization/83985
27025         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
27026         REG_CFA_RESTORE insns.
27027         (delete_unmarked_insns): Don't ignore separate shrink wrapping
27028         REG_CFA_RESTORE insns here.
27030         PR c/83989
27031         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
27032         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
27034 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27036         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
27037         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
27038         (arc_init): Likewise.
27039         (arc_override_options): Likewise.
27040         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
27041         value.
27042         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
27043         support.
27044         * config/arc/arc.h (TARGET_DBNZ): Define.
27045         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
27046         properly set the tune attribute.
27047         (dbnz): Use TARGET_DBNZ guard.
27048         * config/arc/arc.opt (mtune): Add core3 option.
27050 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27052         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
27053         recognize new pic like addresses.
27054         (arc_delegitimize_address): Clean up.
27056 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27058         * config/arc/arc-arches.def: Option mrf16 valid for all
27059         architectures.
27060         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
27061         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
27062         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
27063         * config/arc/arc-tables.opt: Regenerate.
27064         * config/arc/arc.c (arc_conditional_register_usage): Handle
27065         reduced register file case.
27066         (arc_file_start): Set must have build attributes.
27067         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
27068         mrf16 option value.
27069         * config/arc/arc.opt (mrf16): Add new option.
27070         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
27071         * config/arc/genmultilib.awk: Handle new mrf16 option.
27072         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
27073         * config/arc/t-multilib: Regenerate.
27074         * doc/invoke.texi (ARC Options): Document mrf16 option.
27076 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27078         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
27079         * config/arc/arc.c (arc_handle_secure_attribute): New function.
27080         (arc_attribute_table): Add 'secure_call' attribute.
27081         (arc_print_operand): Print secure call operand.
27082         (arc_function_ok_for_sibcall): Don't optimize tail calls when
27083         secure.
27084         (arc_is_secure_call_p): New function.  * config/arc/arc.md
27085         (call_i): Add support for sjli instruction.
27086         (call_value_i): Likewise.
27087         * config/arc/constraints.md (Csc): New constraint.
27089 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
27090             John Eric Martin  <John.Martin@emmicro-us.com>
27092         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
27093         * config/arc/arc.c (_arc_jli_section): New struct.
27094         (arc_jli_section): New type.
27095         (rc_jli_sections): New static variable.
27096         (arc_handle_jli_attribute): New function.
27097         (arc_attribute_table): Add jli_always and jli_fixed attribute.
27098         (arc_file_end): New function.
27099         (TARGET_ASM_FILE_END): Define.
27100         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
27101         (arc_add_jli_section): New function.
27102         (jli_call_scan): Likewise.
27103         (arc_reorg): Call jli_call_scan.
27104         (arc_output_addsi): Remove 'S' from printing asm operand.
27105         (arc_is_jli_call_p): New function.
27106         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
27107         operand.
27108         (movhi_insn): Likewise.
27109         (movsi_insn): Likewise.
27110         (movsi_set_cc_insn): Likewise.
27111         (loadqi_update): Likewise.
27112         (load_zeroextendqisi_update): Likewise.
27113         (load_signextendqisi_update): Likewise.
27114         (loadhi_update): Likewise.
27115         (load_zeroextendhisi_update): Likewise.
27116         (load_signextendhisi_update): Likewise.
27117         (loadsi_update): Likewise.
27118         (loadsf_update): Likewise.
27119         (movsicc_insn): Likewise.
27120         (bset_insn): Likewise.
27121         (bxor_insn): Likewise.
27122         (bclr_insn): Likewise.
27123         (bmsk_insn): Likewise.
27124         (bicsi3_insn): Likewise.
27125         (cmpsi_cc_c_insn): Likewise.
27126         (movsi_ne): Likewise.
27127         (movsi_cond_exec): Likewise.
27128         (clrsbsi2): Likewise.
27129         (norm_f): Likewise.
27130         (normw): Likewise.
27131         (swap): Likewise.
27132         (divaw): Likewise.
27133         (flag): Likewise.
27134         (sr): Likewise.
27135         (kflag): Likewise.
27136         (ffs): Likewise.
27137         (ffs_f): Likewise.
27138         (fls): Likewise.
27139         (call_i): Remove 'S' asm letter, add jli instruction.
27140         (call_value_i): Likewise.
27141         * config/arc/arc.op (mjli-always): New option.
27142         * config/arc/constraints.md (Cji): New constraint.
27143         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
27144         operand.
27145         (subsf3_fpx): Likewise.
27146         (mulsf3_fpx): Likewise.
27147         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
27148         asm operand.
27149         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
27150         function attrbutes.
27151         * doc/invoke.texi (ARC): Document mjli-always option.
27153 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
27155         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
27156         avoid addition with 0 and use incw and decw where possible.
27158 2018-01-26  Richard Biener  <rguenther@suse.de>
27160         PR tree-optimization/81082
27161         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
27162         association if it requires casting to unsigned.
27163         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
27164         from fold_plusminus_mult_expr to catch important cases late when
27165         range info is available.
27167 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27169         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
27170         * configure.ac (hidden_linkonce): New test.
27171         * configure: Regenerate.
27172         * config.in: Regenerate.
27174 2018-01-26  Julia Koval  <julia.koval@intel.com>
27176         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
27177         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
27178         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
27179         _mm_mask_bitshuffle_epi64_mask): Fix type.
27180         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
27181         USI_FTYPE_V4DI_V4DI_USI): Remove.
27182         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
27183         __builtin_ia32_vpshufbitqmb256_mask,
27184         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
27185         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
27186         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
27188 2018-01-26  Alan Modra  <amodra@gmail.com>
27190         PR target/84033
27191         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
27192         UNSPEC_VBPERMQ.  Sort other unspecs.
27194 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
27196         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
27198 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
27200         PR middle-end/83055
27201         * predict.c (drop_profile): Do not push/pop cfun; update also
27202         node->count.
27203         (handle_missing_profiles): Fix logic looking for zero profiles.
27205 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
27207         PR middle-end/83977
27208         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
27209         on functions with #pragma omp declare simd or functions with simd
27210         attribute.
27211         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
27212         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
27213         Remove trailing \n from warning_at calls.
27215 2018-01-25  Tom de Vries  <tom@codesourcery.com>
27217         PR target/84028
27218         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27219         for neutered workers.
27221 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
27223         PR target/68467
27224         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
27225         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
27227 2018-01-24  Jeff Law  <law@redhat.com>
27229         PR target/83994
27230         * i386.c (get_probe_interval): Move to earlier point.
27231         (ix86_compute_frame_layout): If -fstack-clash-protection and
27232         the frame is larger than the probe interval, then use pushes
27233         to save registers rather than reg->mem moves.
27234         (ix86_expand_prologue): Remove conditional for int_registers_saved
27235         assertion.
27237 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
27239         PR target/84014
27240         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
27241         min/max for never referenced object.
27243 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
27245         PR middle-end/83977
27246         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
27247         here.
27248         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
27249         attributes from DECL_ATTRIBUTES (decl) without affecting
27250         DECL_ATTRIBUTES (current_function_decl).
27251         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
27252         functions with non-NULL DECL_ABSTRACT_ORIGIN.
27254 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
27256         PR tree-optimization/83979
27257         * fold-const.c (fold_comparison): Use constant_boolean_node
27258         instead of boolean_{true,false}_node.
27260 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
27262         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
27263         with zero counts.
27265 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27267         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27268         Simplify the clause that sets the length attribute.
27269         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27270         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
27271         clause that sets the length attribute.
27272         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27274 2018-01-24  Tom de Vries  <tom@codesourcery.com>
27276         PR target/83589
27277         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
27278         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
27279         Add strict parameter.
27280         (prevent_branch_around_nothing): Insert dummy insn between branch to
27281         label and label with no ptx insn inbetween.
27282         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
27284 2018-01-24  Tom de Vries  <tom@codesourcery.com>
27286         PR target/81352
27287         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
27288         for neutered threads in warp.
27289         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
27291 2018-01-24  Richard Biener  <rguenther@suse.de>
27293         PR tree-optimization/83176
27294         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
27295         operands.
27297 2018-01-24  Richard Biener  <rguenther@suse.de>
27299         PR tree-optimization/82819
27300         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
27301         code generating pluses that are no-ops in the target precision.
27303 2018-01-24  Richard Biener  <rguenther@suse.de>
27305         PR middle-end/84000
27306         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
27308 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27310         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
27311         to merge probabilities.
27312         * predict.c (probably_never_executed): Also mark as cold functions
27313         with global 0 profile and guessed local profile.
27314         * profile-count.c (profile_probability::combine_with_count): New
27315         member function.
27316         * profile-count.h (profile_probability::operator*,
27317         profile_probability::operator*=, profile_probability::operator/,
27318         profile_probability::operator/=): Reduce precision to adjusted
27319         and set value to guessed on contradictory divisions.
27320         (profile_probability::combine_with_freq): Remove.
27321         (profile_probability::combine_wiht_count): Declare.
27322         (profile_count::force_nonzero):: Set to adjusted.
27323         (profile_count::probability_in):: Set quality to adjusted.
27324         * tree-ssa-tail-merge.c (replace_block_by): Use
27325         combine_with_count.
27327 2018-01-23  Andrew Waterman  <andrew@sifive.com>
27328             Jim Wilson  <jimw@sifive.com>
27330         * config/riscv/riscv.c (riscv_stack_boundary): New.
27331         (riscv_option_override): Set riscv_stack_boundary.  Handle
27332         riscv_preferred_stack_boundary_arg.
27333         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
27334         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
27335         (STACK_BOUNDARY): Set to riscv_stack_boundary.
27336         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
27337         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
27338         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
27340 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
27342         PR target/83905
27343         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
27344         of struct ix86_frame.
27345         (ix86_expand_epilogue): Likewise.  Add a local variable for
27346         the reg_save_offset field in struct ix86_frame.
27348 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
27350         PR tree-optimization/82604
27351         * tree-loop-distribution.c (enum partition_kind): New enum item
27352         PKIND_PARTIAL_MEMSET.
27353         (partition_builtin_p): Support above new enum item.
27354         (generate_code_for_partition): Ditto.
27355         (compute_access_range): Differentiate cases that equality can be
27356         proven at all loops, the innermost loops or no loops.
27357         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
27358         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
27359         (finalize_partitions, distribute_loop): Don't fuse partition of
27360         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
27361         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
27362         parloop is enabled.
27364 2018-01-23  Martin Liska  <mliska@suse.cz>
27366         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
27367         order to ignore the predictor.
27368         (PRED_POLYMORPHIC_CALL): Likewise.
27369         (PRED_RECURSIVE_CALL): Likewise.
27371 2018-01-23  Martin Liska  <mliska@suse.cz>
27373         * tree-profile.c (tree_profiling): Print function header to
27374         aware reader which function we are working on.
27375         * value-prof.c (gimple_find_values_to_profile): Do not print
27376         not interesting value histograms.
27378 2018-01-23  Martin Liska  <mliska@suse.cz>
27380         * profile-count.h (enum profile_quality): Add
27381         profile_uninitialized as the first value. Do not number values
27382         as they are zero based.
27383         (profile_count::verify): Update sanity check.
27384         (profile_probability::verify): Likewise.
27386 2018-01-23  Nathan Sidwell  <nathan@acm.org>
27388         * doc/invoke.texi (ffor-scope): Deprecate.
27390 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27392         PR tree-optimization/83510
27393         * domwalk.c (set_all_edges_as_executable): New function.
27394         (dom_walker::dom_walker): Convert bool param
27395         "skip_unreachable_blocks" to enum reachability.  Move setup of
27396         edge flags to set_all_edges_as_executable and only do it when
27397         reachability is REACHABLE_BLOCKS.
27398         * domwalk.h (enum dom_walker::reachability): New enum.
27399         (dom_walker::dom_walker): Convert bool param
27400         "skip_unreachable_blocks" to enum reachability.
27401         (set_all_edges_as_executable): New decl.
27402         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
27403         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
27404         "reachability".
27405         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
27406         but converting true to REACHABLE_BLOCKS.
27407         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
27408         * tree-vrp.c
27409         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
27410         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
27411         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
27412         REACHABLE_BLOCKS.
27413         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
27414         if check_all_array_refs will be called.
27416 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27418         * tree.c (selftest::test_location_wrappers): Add more test
27419         coverage.
27421 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
27423         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
27424         (selftest::test_bit_in_range): Likewise.
27426 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27428         PR testsuite/83888
27429         * doc/sourcebuild.texi (vect_float): Say that the selector
27430         only describes the situation when -funsafe-math-optimizations is on.
27431         (vect_float_strict): Document.
27433 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
27435         PR tree-optimization/83965
27436         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
27437         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
27438         instead of checking only for a reduction.
27439         (vect_recog_widen_sum_pattern): Likewise.
27441 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27443         * predict.c (probably_never_executed): Only use precise profile info.
27444         (compute_function_frequency): Skip after inlining hack since we now
27445         have quality checking.
27447 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
27449         * profile-count.h (profile_probability::very_unlikely,
27450         profile_probability::unlikely, profile_probability::even): Set
27451         precision to guessed.
27453 2018-01-23  Richard Biener  <rguenther@suse.de>
27455         PR tree-optimization/83963
27456         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
27457         Properly terminate dominator walk when crossing the exit edge not
27458         when visiting its source block.
27460 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
27462         PR c++/83918
27463         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
27464         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
27466 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
27468         PR tree-optimization/83957
27469         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
27470         semicolon after for body surrounded by braces.
27472         PR tree-optimization/83081
27473         * profile-count.h (profile_probability::split): New method.
27474         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
27475         Use profile_probability::split.
27476         (do_compare_rtx_and_jump): Fix adjustment of probabilities
27477         when splitting a single conditional jump into 2.
27479 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
27481         PR tree-optimization/69452
27482         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
27483         decl.
27485 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27487         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
27488         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
27489         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
27491 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27493         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
27494         declaration.
27495         * config/rl78/rl78.md (movdi): New define_expand.
27496         * config/rl78/rl78.c (rl78_split_movdi): New function.
27498 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
27500         PR target/83862
27501         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
27502         no longer used.
27503         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
27504         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
27505         128-bit to produce an UNSPEC move to get the double word with the
27506         signbit and then a shift directly to do signbit.
27507         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
27508         implementation with a new version that just does either a direct
27509         move or a regular move.  Move memory interface to separate insns.
27510         Move insns so they are next to the expander.
27511         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
27512         with signbit move.  Split big and little endian case.
27513         (signbit<mode>2_dm_mem_le): Likewise.
27514         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
27515         (signbit<mode>2_dm2): Likewise.
27517 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27519         * config/rl78/rl78.md (anddi3): New define_expand.
27521 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27523         * config/rl78/rl78.md (umindi3): New define_expand.
27525 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27527         * config/rl78/rl78.md (smindi3): New define_expand.
27529 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27531         * config/rl78/rl78.md (smaxdi3): New define_expand.
27533 2018-01-22  Carl Love  <cel@us.ibm.com>
27535         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
27536         LVX_V1TI): Add macro expansion.
27537         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
27538         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
27539         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
27540         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
27541         Change check to determine if the instruction is a byte reversing
27542         entry.  Fix typo in comment.
27543         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
27544         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
27545         Add def_builtin calls for new builtins.
27546         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
27547         Add define_insn expansion.
27549 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27551         * config/rl78/rl78.md (umaxdi3): New define_expand.
27553 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
27555         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
27556         for non-QImode registers.
27558 2018-01-22  Richard Biener  <rguenther@suse.de>
27560         PR tree-optimization/83963
27561         * graphite-scop-detection.c (scop_detection::get_sese): Delay
27562         including the loop exit block.
27563         (scop_detection::merge_sese): Likewise.
27564         (scop_detection::add_scop): Do it here instead.
27566 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27568         * doc/sourcebuild.texi (arm_softfloat): Document.
27570 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
27572         PR gcc/77734
27573         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
27574         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
27575         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
27577 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27578             David Edelsohn  <dje.gcc@gmail.com>
27580         PR target/83946
27581         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27582         Change "crset eq" to "crset 2".
27583         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27584         (*call_indirect_aix<mode>_nospec): Likewise.
27585         (*call_value_indirect_aix<mode>_nospec): Likewise.
27586         (*call_indirect_elfv2<mode>_nospec): Likewise.
27587         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
27588         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
27589         change assembly output from . to $.
27590         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27591         (indirect_jump<mode>_nospec): Change assembly output from . to $.
27592         (*tablejump<mode>_internal1_nospec): Likewise.
27594 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
27596         PR target/80870
27597         * config/sh/sh_optimize_sett_clrt.cc:
27598         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
27600 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
27602         PR tree-optimization/83940
27603         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
27604         offset_dt to vect_constant_def rather than vect_unknown_def_type.
27605         (vect_check_load_store_mask): Add a mask_dt_out parameter and
27606         use it to pass back the definition type.
27607         (vect_check_store_rhs): Likewise rhs_dt_out.
27608         (vect_build_gather_load_calls): Add a mask_dt argument and use
27609         it instead of a call to vect_is_simple_use.
27610         (vectorizable_store): Update calls to vect_check_load_store_mask
27611         and vect_check_store_rhs.  Use the dt returned by the latter instead
27612         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
27613         instead of calls to vect_is_simple_use.  Pass the scalar rather
27614         than the vector operand to vect_is_simple_use when handling
27615         second and subsequent copies of an rhs value.
27616         (vectorizable_load): Update calls to vect_check_load_store_mask
27617         and vect_build_gather_load_calls.  Use the cached mask_dt and
27618         gs_info.offset_dt instead of calls to vect_is_simple_use.
27620 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
27622         PR middle-end/83945
27623         * tree-emutls.c: Include gimplify.h.
27624         (lower_emutls_2): New function.
27625         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
27626         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
27627         it before further processing.
27629         PR target/83930
27630         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
27631         UINTVAL (trueop1) instead of INTVAL (op1).
27633 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
27635         PR debug/81570
27636         PR debug/83728
27637         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
27638         INCOMING_FRAME_SP_OFFSET if not defined.
27639         (scan_trace): Add ENTRY argument.  If true and
27640         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
27641         emit a note to adjust the CFA offset.
27642         (create_cfi_notes): Adjust scan_trace callers.
27643         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
27644         INCOMING_FRAME_SP_OFFSET in the CIE.
27645         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
27646         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
27647         Likewise.
27648         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
27649         * doc/tm.texi: Regenerated.
27651 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27653         PR rtl-optimization/83147
27654         * lra-constraints.c (remove_inheritance_pseudos): Use
27655         lra_substitute_pseudo_within_insn.
27657 2018-01-19  Tom de Vries  <tom@codesourcery.com>
27658             Cesar Philippidis  <cesar@codesourcery.com>
27660         PR target/83920
27661         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
27663 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
27665         PR target/83790
27666         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
27667         spaces for function labels.
27669 2018-01-19  Martin Liska  <mliska@suse.cz>
27671         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
27672         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
27673         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
27674         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
27675         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
27676         (PRED_CONST_RETURN): Change from 69 to 65.
27677         (PRED_NULL_RETURN): Change from 91 to 71.
27678         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
27679         (PRED_LOOP_GUARD): Change from 66 to 73.
27681 2018-01-19  Martin Liska  <mliska@suse.cz>
27683         * predict.c (predict_insn_def): Add new assert.
27684         (struct branch_predictor): Change type to signed integer.
27685         (test_prediction_value_range): Amend test to cover
27686         PROB_UNINITIALIZED.
27687         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
27688         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
27689         (PRED_LOOP_ITERATIONS_MAX): Likewise.
27690         (PRED_LOOP_IV_COMPARE): Likewise.
27691         * predict.h (PROB_UNINITIALIZED): Define new constant.
27693 2018-01-19  Martin Liska  <mliska@suse.cz>
27695         * predict.c (dump_prediction): Add new format for
27696         analyze_brprob.py script which is enabled with -details
27697         suboption.
27698         * profile-count.h (precise_p): New function.
27700 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27702         PR tree-optimization/83922
27703         * tree-vect-loop.c (vect_verify_full_masking): Return false if
27704         there are no statements that need masking.
27705         (vect_active_double_reduction_p): New function.
27706         (vect_analyze_loop_operations): Use it when handling phis that
27707         are not in the loop header.
27709 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
27711         PR tree-optimization/83914
27712         * tree-vect-loop.c (vectorizable_induction): Don't convert
27713         init_expr or apply the peeling adjustment for inductions
27714         that are nested within the vectorized loop.
27716 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27718         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
27719         instead of NEG.
27721 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27723         PR sanitizer/81715
27724         PR testsuite/83882
27725         * function.h (gimplify_parameters): Add gimple_seq * argument.
27726         * function.c: Include gimple.h and options.h.
27727         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
27728         for the added local temporaries if needed.
27729         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
27730         if there are any parameter cleanups, wrap whole body into a
27731         try/finally with the cleanups.
27733 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
27735         PR target/82964
27736         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27737         Use GET_MODE_CLASS for scalar floating point.
27739 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27741         PR ipa/82256
27742         patch by PaX Team
27743         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
27744         Fix call of call_cgraph_insertion_hooks.
27746 2018-01-18  Martin Sebor  <msebor@redhat.com>
27748         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
27750 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
27752         PR ipa/83619
27753         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
27754         frequencies.
27756 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
27758         PR other/70268
27759         * common.opt: (-ffile-prefix-map): New option.
27760         * opts.c (common_handle_option): Defer it.
27761         * opts-global.c (handle_common_deferred_options): Handle it.
27762         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
27763         * file-prefix-map.h: New file.
27764         (remap_debug_filename, add_debug_prefix_map): ...here.
27765         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
27766         * final.c (debug_prefix_map, add_debug_prefix_map
27767         remap_debug_filename): Move to...
27768         * file-prefix-map.c: New file.
27769         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
27770         generalize, get rid of alloca(), use strrchr() instead of strchr().
27771         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
27772         Implement in terms of add_prefix_map().
27773         (remap_macro_filename, remap_debug_filename): Implement in term of
27774         remap_filename().
27775         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
27776         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
27777         * dbxout.c: Include file-prefix-map.h.
27778         * varasm.c: Likewise.
27779         * vmsdbgout.c: Likewise.
27780         * xcoffout.c: Likewise.
27781         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
27782         * doc/cppopts.texi (-fmacro-prefix-map): Document.
27783         * doc/invoke.texi (-ffile-prefix-map): Document.
27784         (-fdebug-prefix-map): Update description.
27786 2018-01-18  Martin Liska  <mliska@suse.cz>
27788         * config/i386/i386.c (indirect_thunk_name): Document that also
27789         lfence is emitted.
27790         (output_indirect_thunk): Document why both instructions
27791         (pause and lfence) are generated.
27793 2018-01-18  Richard Biener  <rguenther@suse.de>
27795         PR tree-optimization/83887
27796         * graphite-scop-detection.c
27797         (scop_detection::get_nearest_dom_with_single_entry): Remove.
27798         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
27799         (scop_detection::merge_sese): Re-implement with a flood-fill
27800         algorithm that properly finds a SESE region if it exists.
27802 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
27804         PR c/61240
27805         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
27806         pointer_diff optimizations use view_convert instead of convert.
27808 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27810         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
27811         Generate different code for -mno-speculate-indirect-jumps.
27812         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
27813         (*call_indirect_aix<mode>): Disable for
27814         -mno-speculate-indirect-jumps.
27815         (*call_indirect_aix<mode>_nospec): New define_insn.
27816         (*call_value_indirect_aix<mode>): Disable for
27817         -mno-speculate-indirect-jumps.
27818         (*call_value_indirect_aix<mode>_nospec): New define_insn.
27819         (*sibcall_nonlocal_sysv<mode>): Generate different code for
27820         -mno-speculate-indirect-jumps.
27821         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
27823 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
27825         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
27826         long double type, set the flags for noting the default long double
27827         type, even if we don't pass or return a long double type.
27829 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
27831         PR ipa/83051
27832         * ipa-inline.c (flatten_function): Do not overwrite final inlining
27833         failure.
27835 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
27837         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
27838         support for merge[hl].
27839         (fold_mergehl_helper): New helper function.
27840         (tree-vector-builder.h): New #include for tree_vector_builder usage.
27841         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
27842         (altivec_vmrglw_direct): Add xxmrglw insn.
27844 2018-01-17  Andrew Waterman  <andrew@sifive.com>
27846         * config/riscv/riscv.c (riscv_conditional_register_usage): If
27847         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
27849 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
27851         PR lto/83121
27852         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
27853         call the lto_location_cache before reading the
27854         DECL_SOURCE_LOCATION of the types.
27856 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
27857             Richard Sandiford  <richard.sandiford@linaro.org>
27859         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
27860         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
27861         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
27862         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
27863         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
27864         Add declaration.
27865         * config/aarch64/constraints.md (aarch64_movti_operand):
27866         Limit immediates.
27867         * config/aarch64/predicates.md (Uti): Add new constraint.
27869 2018-01-17  Carl Love  <cel@us.ibm.com>
27871         * config/rs6000/vsx.md (define_expand xl_len_r,
27872         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
27873         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
27874         lxvll.
27875         (define_expand, define_insn): Move the shift left from  the
27876         define_insn to the define_expand for lxvl and stxvl instructions.
27877         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
27878         and XL_LEN_R definitions to PURE.
27880 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
27882         * config/i386/i386.c (indirect_thunk_name): Declare regno
27883         as unsigned int.  Compare regno with INVALID_REGNUM.
27884         (output_indirect_thunk): Ditto.
27885         (output_indirect_thunk_function): Ditto.
27886         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
27887         in the call to output_indirect_thunk_function.
27889 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
27891         PR middle-end/83884
27892         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
27893         rather than the size of inner_type to determine the stack slot size
27894         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
27896 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
27898         PR target/83546
27899         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
27900         to PTA_SILVERMONT.
27902 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
27904         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
27905         endian Linux systems to optionally enable multilibs for selecting
27906         the long double type if the user configured an explicit type.
27907         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
27908         have no long double multilibs if not defined.
27909         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
27910         warn if the user used -mabi={ieee,ibm}longdouble and we built
27911         multilibs for long double.
27912         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
27913         appropriate multilib option.
27914         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
27915         multilib options.
27916         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
27917         for building long double multilibs.
27918         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
27920 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
27922         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
27923         copies.
27925         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
27926         64 bits.
27927         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
27928         128 bits.
27930         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
27931         variables.
27933         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
27934         return value.
27936 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27938         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
27939         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
27941 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
27943         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
27944         different rtl trees depending on TARGET_64BIT.
27945         (rs6000_gen_lvx): Likewise.
27947 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
27949         * config/visium/visium.md (nop): Tweak comment.
27950         (hazard_nop): Likewise.
27952 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27954         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
27955         -mspeculate-indirect-jumps.
27956         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
27957         for -mno-speculate-indirect-jumps.
27958         (*call_indirect_elfv2<mode>_nospec): New define_insn.
27959         (*call_value_indirect_elfv2<mode>): Disable for
27960         -mno-speculate-indirect-jumps.
27961         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
27962         (indirect_jump): Emit different RTL for
27963         -mno-speculate-indirect-jumps.
27964         (*indirect_jump<mode>): Disable for
27965         -mno-speculate-indirect-jumps.
27966         (*indirect_jump<mode>_nospec): New define_insn.
27967         (tablejump): Emit different RTL for
27968         -mno-speculate-indirect-jumps.
27969         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
27970         (tablejumpsi_nospec): New define_expand.
27971         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
27972         (tablejumpdi_nospec): New define_expand.
27973         (*tablejump<mode>_internal1): Disable for
27974         -mno-speculate-indirect-jumps.
27975         (*tablejump<mode>_internal1_nospec): New define_insn.
27976         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
27977         option.
27979 2018-01-16  Artyom Skrobov tyomitch@gmail.com
27981         * caller-save.c (insert_save): Drop unnecessary parameter.  All
27982         callers updated.
27984 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
27985             Richard Biener  <rguenth@suse.de>
27987         PR libgomp/83590
27988         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
27989         return early, inline manually is_gimple_sizepos.  Make sure if we
27990         call gimplify_expr we don't end up with a gimple constant.
27991         * tree.c (variably_modified_type_p): Don't return true for
27992         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
27993         * gimplify.h (is_gimple_sizepos): Remove.
27995 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
27997         PR tree-optimization/83857
27998         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
27999         vectorizable_live_operation for pure SLP statements.
28000         (vectorizable_live_operation): Handle PHIs.
28002 2018-01-16  Richard Biener  <rguenther@suse.de>
28004         PR tree-optimization/83867
28005         * tree-vect-stmts.c (vect_transform_stmt): Precompute
28006         nested_in_vect_loop_p since the scalar stmt may get invalidated.
28008 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
28010         PR c/83844
28011         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
28012         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
28013         If off is not INTEGER_CST, issue a may not be aligned warning
28014         rather than isn't aligned.  Use isn%'t rather than isn't.
28015         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
28016         into MULT_EXPR.
28017         <case MULT_EXPR>: Improve the case when bottom and one of the
28018         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
28019         operand, in that case check if the other operand is multiple of
28020         bottom divided by the INTEGER_CST operand.
28022 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28024         PR target/83858
28025         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
28026         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
28027         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
28028         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
28029         * config/pa/pa.c (pa_function_arg_advance): Likewise.
28030         (pa_function_arg, pa_arg_partial_bytes): Likewise.
28031         (pa_function_arg_size): New function.
28033 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28035         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
28036         in a separate statement.
28038 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
28040         PR tree-optimization/83847
28041         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
28042         group gathers and scatters.
28044 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
28046         PR rtl-optimization/86620
28047         * params.def (max-sched-ready-insns): Bump minimum value to 1.
28049         PR rtl-optimization/83213
28050         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
28051         to last if both are JUMP_INSNs.
28053         PR tree-optimization/83843
28054         * gimple-ssa-store-merging.c
28055         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
28056         store_immediate_info for bswap/nop orig_stores.
28058 2018-01-15  Andrew Waterman  <andrew@sifive.com>
28060         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
28061         !TARGET_MUL.
28062         <UDIV>: Increase cost if !TARGET_DIV.
28064 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
28066         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
28067         (define_attr "cr_logical_3op"): New.
28068         (cceq_ior_compare): Adjust.
28069         (cceq_ior_compare_complement): Adjust.
28070         (*cceq_rev_compare): Adjust.
28071         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
28072         (is_cracked_insn): Adjust.
28073         (insn_must_be_first_in_group): Adjust.
28074         * config/rs6000/40x.md: Adjust.
28075         * config/rs6000/440.md: Adjust.
28076         * config/rs6000/476.md: Adjust.
28077         * config/rs6000/601.md: Adjust.
28078         * config/rs6000/603.md: Adjust.
28079         * config/rs6000/6xx.md: Adjust.
28080         * config/rs6000/7450.md: Adjust.
28081         * config/rs6000/7xx.md: Adjust.
28082         * config/rs6000/8540.md: Adjust.
28083         * config/rs6000/cell.md: Adjust.
28084         * config/rs6000/e300c2c3.md: Adjust.
28085         * config/rs6000/e500mc.md: Adjust.
28086         * config/rs6000/e500mc64.md: Adjust.
28087         * config/rs6000/e5500.md: Adjust.
28088         * config/rs6000/e6500.md: Adjust.
28089         * config/rs6000/mpc.md: Adjust.
28090         * config/rs6000/power4.md: Adjust.
28091         * config/rs6000/power5.md: Adjust.
28092         * config/rs6000/power6.md: Adjust.
28093         * config/rs6000/power7.md: Adjust.
28094         * config/rs6000/power8.md: Adjust.
28095         * config/rs6000/power9.md: Adjust.
28096         * config/rs6000/rs64.md: Adjust.
28097         * config/rs6000/titan.md: Adjust.
28099 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28101         * config/i386/predicates.md (indirect_branch_operand): Rewrite
28102         ix86_indirect_branch_register logic.
28104 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28106         * config/i386/constraints.md (Bs): Update
28107         ix86_indirect_branch_register check.  Don't check
28108         ix86_indirect_branch_register with GOT_memory_operand.
28109         (Bw): Likewise.
28110         * config/i386/predicates.md (GOT_memory_operand): Don't check
28111         ix86_indirect_branch_register here.
28112         (GOT32_symbol_operand): Likewise.
28114 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28116         * config/i386/predicates.md (constant_call_address_operand):
28117         Rewrite ix86_indirect_branch_register logic.
28118         (sibcall_insn_operand): Likewise.
28120 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28122         * config/i386/constraints.md (Bs): Replace
28123         ix86_indirect_branch_thunk_register with
28124         ix86_indirect_branch_register.
28125         (Bw): Likewise.
28126         * config/i386/i386.md (indirect_jump): Likewise.
28127         (tablejump): Likewise.
28128         (*sibcall_memory): Likewise.
28129         (*sibcall_value_memory): Likewise.
28130         Peepholes of indirect call and jump via memory: Likewise.
28131         * config/i386/i386.opt: Likewise.
28132         * config/i386/predicates.md (indirect_branch_operand): Likewise.
28133         (GOT_memory_operand): Likewise.
28134         (call_insn_operand): Likewise.
28135         (sibcall_insn_operand): Likewise.
28136         (GOT32_symbol_operand): Likewise.
28138 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
28140         PR middle-end/83837
28141         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
28142         type rather than type addr's type points to.
28143         (expand_omp_atomic_mutex): Likewise.
28144         (expand_omp_atomic): Likewise.
28146 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
28148         PR target/83839
28149         * config/i386/i386.c (output_indirect_thunk_function): Use
28150         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
28151         for  __x86_return_thunk.
28153 2018-01-15  Richard Biener  <rguenther@suse.de>
28155         PR middle-end/83850
28156         * expmed.c (extract_bit_field_1): Fix typo.
28158 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
28160         PR target/83687
28161         * config/arm/iterators.md (VF): New mode iterator.
28162         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
28163         Remove integer-related logic from pattern.
28164         (neon_vabd<mode>_3): Likewise.
28166 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
28168         PR middle-end/82694
28169         * common.opt (fstrict-overflow): No longer an alias.
28170         (fwrapv-pointer): New option.
28171         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
28172         also for pointer types based on flag_wrapv_pointer.
28173         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
28174         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
28175         opts->x_flag_wrapv got set.
28176         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
28177         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
28178         POINTER_TYPE_OVERFLOW_UNDEFINED.
28179         * match.pd: Likewise in address comparison pattern.
28180         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
28182 2018-01-15  Richard Biener  <rguenther@suse.de>
28184         PR lto/83804
28185         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
28186         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
28187         Reset type names to their identifier if their TYPE_DECL doesn't
28188         have linkage (and thus is used for ODR and devirt).
28189         (save_debug_info_for_decl): Remove.
28190         (save_debug_info_for_type): Likewise.
28191         (add_tree_to_fld_list): Adjust.
28192         * tree-pretty-print.c (dump_generic_node): Make dumping of
28193         type names more robust.
28195 2018-01-15  Richard Biener  <rguenther@suse.de>
28197         * BASE-VER: Bump to 8.0.1.
28199 2018-01-14  Martin Sebor  <msebor@redhat.com>
28201         PR other/83508
28202         * builtins.c (check_access): Avoid warning when the no-warning bit
28203         is set.
28205 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
28207         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
28208         * ira-color (allocno_hard_regs_compare): Likewise.
28210 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
28212         PR target/83013
28213         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
28214         Use .pushsection/.popsection.
28216 2018-01-14  Martin Sebor  <msebor@redhat.com>
28218         PR c++/81327
28219         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
28221 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
28223         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
28224         entry from extra_headers.
28225         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
28226         extra_headers, make the list bitwise identical to the i?86-*-* one.
28228 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28230         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
28231         -mcmodel=large with -mindirect-branch=thunk,
28232         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
28233         -mfunction-return=thunk-extern.
28234         * doc/invoke.texi: Document -mcmodel=large is incompatible with
28235         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
28236         -mfunction-return=thunk and -mfunction-return=thunk-extern.
28238 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28240         * config/i386/i386.c (print_reg): Print the name of the full
28241         integer register without '%'.
28242         (ix86_print_operand): Handle 'V'.
28243         * doc/extend.texi: Document 'V' modifier.
28245 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28247         * config/i386/constraints.md (Bs): Disallow memory operand for
28248         -mindirect-branch-register.
28249         (Bw): Likewise.
28250         * config/i386/predicates.md (indirect_branch_operand): Likewise.
28251         (GOT_memory_operand): Likewise.
28252         (call_insn_operand): Likewise.
28253         (sibcall_insn_operand): Likewise.
28254         (GOT32_symbol_operand): Likewise.
28255         * config/i386/i386.md (indirect_jump): Call convert_memory_address
28256         for -mindirect-branch-register.
28257         (tablejump): Likewise.
28258         (*sibcall_memory): Likewise.
28259         (*sibcall_value_memory): Likewise.
28260         Disallow peepholes of indirect call and jump via memory for
28261         -mindirect-branch-register.
28262         (*call_pop): Replace m with Bw.
28263         (*call_value_pop): Likewise.
28264         (*sibcall_pop_memory): Replace m with Bs.
28265         * config/i386/i386.opt (mindirect-branch-register): New option.
28266         * doc/invoke.texi: Document -mindirect-branch-register option.
28268 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28270         * config/i386/i386-protos.h (ix86_output_function_return): New.
28271         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
28272         set function_return_type.
28273         (indirect_thunk_name): Add ret_p to indicate thunk for function
28274         return.
28275         (output_indirect_thunk_function): Pass false to
28276         indirect_thunk_name.
28277         (ix86_output_indirect_branch_via_reg): Likewise.
28278         (ix86_output_indirect_branch_via_push): Likewise.
28279         (output_indirect_thunk_function): Create alias for function
28280         return thunk if regno < 0.
28281         (ix86_output_function_return): New function.
28282         (ix86_handle_fndecl_attribute): Handle function_return.
28283         (ix86_attribute_table): Add function_return.
28284         * config/i386/i386.h (machine_function): Add
28285         function_return_type.
28286         * config/i386/i386.md (simple_return_internal): Use
28287         ix86_output_function_return.
28288         (simple_return_internal_long): Likewise.
28289         * config/i386/i386.opt (mfunction-return=): New option.
28290         (indirect_branch): Mention -mfunction-return=.
28291         * doc/extend.texi: Document function_return function attribute.
28292         * doc/invoke.texi: Document -mfunction-return= option.
28294 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
28296         * config/i386/i386-opts.h (indirect_branch): New.
28297         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
28298         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
28299         with local indirect jump when converting indirect call and jump.
28300         (ix86_set_indirect_branch_type): New.
28301         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
28302         (indirectlabelno): New.
28303         (indirect_thunk_needed): Likewise.
28304         (indirect_thunk_bnd_needed): Likewise.
28305         (indirect_thunks_used): Likewise.
28306         (indirect_thunks_bnd_used): Likewise.
28307         (INDIRECT_LABEL): Likewise.
28308         (indirect_thunk_name): Likewise.
28309         (output_indirect_thunk): Likewise.
28310         (output_indirect_thunk_function): Likewise.
28311         (ix86_output_indirect_branch_via_reg): Likewise.
28312         (ix86_output_indirect_branch_via_push): Likewise.
28313         (ix86_output_indirect_branch): Likewise.
28314         (ix86_output_indirect_jmp): Likewise.
28315         (ix86_code_end): Call output_indirect_thunk_function if needed.
28316         (ix86_output_call_insn): Call ix86_output_indirect_branch if
28317         needed.
28318         (ix86_handle_fndecl_attribute): Handle indirect_branch.
28319         (ix86_attribute_table): Add indirect_branch.
28320         * config/i386/i386.h (machine_function): Add indirect_branch_type
28321         and has_local_indirect_jump.
28322         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
28323         to true.
28324         (tablejump): Likewise.
28325         (*indirect_jump): Use ix86_output_indirect_jmp.
28326         (*tablejump_1): Likewise.
28327         (simple_return_indirect_internal): Likewise.
28328         * config/i386/i386.opt (mindirect-branch=): New option.
28329         (indirect_branch): New.
28330         (keep): Likewise.
28331         (thunk): Likewise.
28332         (thunk-inline): Likewise.
28333         (thunk-extern): Likewise.
28334         * doc/extend.texi: Document indirect_branch function attribute.
28335         * doc/invoke.texi: Document -mindirect-branch= option.
28337 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
28339         PR ipa/83051
28340         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
28342 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
28344         * ipa-inline.c (want_inline_small_function_p): Return false if
28345         inlining has already failed with CIF_FINAL_ERROR.
28346         (update_caller_keys): Call want_inline_small_function_p before
28347         can_inline_edge_p.
28348         (update_callee_keys): Likewise.
28350 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
28352         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
28353         New function.
28354         (rs6000_quadword_masked_address_p): Likewise.
28355         (quad_aligned_load_p): Likewise.
28356         (quad_aligned_store_p): Likewise.
28357         (const_load_sequence_p): Add comment to describe the outer-most loop.
28358         (mimic_memory_attributes_and_flags): New function.
28359         (rs6000_gen_stvx): Likewise.
28360         (replace_swapped_aligned_store): Likewise.
28361         (rs6000_gen_lvx): Likewise.
28362         (replace_swapped_aligned_load): Likewise.
28363         (replace_swapped_load_constant): Capitalize argument name in
28364         comment describing this function.
28365         (rs6000_analyze_swaps): Add a third pass to search for vector loads
28366         and stores that access quad-word aligned addresses and replace
28367         with stvx or lvx instructions when appropriate.
28368         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
28369         New function prototype.
28370         (rs6000_quadword_masked_address_p): Likewise.
28371         (rs6000_gen_lvx): Likewise.
28372         (rs6000_gen_stvx): Likewise.
28373         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
28374         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
28375         when memory address is aligned.
28376         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
28377         this split to select lvx instruction when memory address is aligned.
28378         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
28379         instruction when memory address is aligned.
28380         (*vsx_le_perm_load_v16qi): Likewise.
28381         (four unnamed splitters): Modify to select the stvx instruction
28382         when memory is aligned.
28384 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
28386         * predict.c (determine_unlikely_bbs): Handle correctly BBs
28387         which appears in the queue multiple times.
28389 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28390             Alan Hayward  <alan.hayward@arm.com>
28391             David Sherwood  <david.sherwood@arm.com>
28393         * tree-vectorizer.h (vec_lower_bound): New structure.
28394         (_loop_vec_info): Add check_nonzero and lower_bounds.
28395         (LOOP_VINFO_CHECK_NONZERO): New macro.
28396         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
28397         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
28398         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
28399         fields.  Make seg_len the distance travelled, not including the
28400         access size.
28401         (dr_direction_indicator): Declare.
28402         (dr_zero_step_indicator): Likewise.
28403         (dr_known_forward_stride_p): Likewise.
28404         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
28405         tree-ssanames.h.
28406         (runtime_alias_check_p): Allow runtime alias checks with
28407         variable strides.
28408         (operator ==): Compare access_size and align.
28409         (prune_runtime_alias_test_list): Rework for new distinction between
28410         the access_size and seg_len.
28411         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
28412         segment lengths.
28413         (get_segment_min_max): New function.
28414         (create_intersect_range_checks): Use it.
28415         (dr_step_indicator): New function.
28416         (dr_direction_indicator): Likewise.
28417         (dr_zero_step_indicator): Likewise.
28418         (dr_known_forward_stride_p): Likewise.
28419         * tree-loop-distribution.c (data_ref_segment_size): Return
28420         DR_STEP * (niters - 1).
28421         (compute_alias_check_pairs): Update call to the dr_with_seg_len
28422         constructor.
28423         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
28424         (vect_preserves_scalar_order_p): New function, split out from...
28425         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
28426         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
28427         (vect_vfa_access_size): New function.
28428         (vect_vfa_align): Likewise.
28429         (vect_compile_time_alias): Take access_size_a and access_b arguments.
28430         (dump_lower_bound): New function.
28431         (vect_check_lower_bound): Likewise.
28432         (vect_small_gap_p): Likewise.
28433         (vectorizable_with_step_bound_p): Likewise.
28434         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
28435         depencies if the vectorization factor is 1.  Convert the checks
28436         for nonzero steps into checks on the bounds of DR_STEP.  Try using
28437         a bunds check for variable steps if the minimum required step is
28438         relatively small. Update calls to the dr_with_seg_len
28439         constructor and to vect_compile_time_alias.
28440         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
28441         function.
28442         (vect_loop_versioning): Call it.
28443         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
28444         when retrying.
28445         (vect_estimate_min_profitable_iters): Account for any bounds checks.
28447 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28448             Alan Hayward  <alan.hayward@arm.com>
28449             David Sherwood  <david.sherwood@arm.com>
28451         * doc/sourcebuild.texi (vect_scatter_store): Document.
28452         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
28453         optabs.
28454         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
28455         Document.
28456         * genopinit.c (main): Add supports_vec_scatter_store and
28457         supports_vec_scatter_store_cached to target_optabs.
28458         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
28459         IFN_MASK_SCATTER_STORE.
28460         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
28461         functions.
28462         * internal-fn.h (internal_store_fn_p): Declare.
28463         (internal_fn_stored_value_index): Likewise.
28464         * internal-fn.c (scatter_store_direct): New macro.
28465         (expand_scatter_store_optab_fn): New function.
28466         (direct_scatter_store_optab_supported_p): New macro.
28467         (internal_store_fn_p): New function.
28468         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
28469         IFN_MASK_SCATTER_STORE.
28470         (internal_fn_mask_index): Likewise.
28471         (internal_fn_stored_value_index): New function.
28472         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
28473         for scatter stores.
28474         * optabs-query.h (supports_vec_scatter_store_p): Declare.
28475         * optabs-query.c (supports_vec_scatter_store_p): New function.
28476         * tree-vectorizer.h (vect_get_store_rhs): Declare.
28477         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
28478         true for scatter stores.
28479         (vect_gather_scatter_fn_p): Handle scatter stores too.
28480         (vect_check_gather_scatter): Consider using scatter stores if
28481         supports_vec_scatter_store_p.
28482         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
28483         scatter stores too.
28484         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28485         internal_fn_stored_value_index.
28486         (check_load_store_masking): Handle scatter stores too.
28487         (vect_get_store_rhs): Make public.
28488         (vectorizable_call): Use internal_store_fn_p.
28489         (vectorizable_store): Handle scatter store internal functions.
28490         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
28491         when deciding whether the end of the group has been reached.
28492         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
28493         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
28494         (mask_scatter_store<mode>): New insns.
28496 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28497             Alan Hayward  <alan.hayward@arm.com>
28498             David Sherwood  <david.sherwood@arm.com>
28500         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
28501         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
28502         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
28503         function.
28504         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
28505         Use vect_truncate_gather_scatter_offset if we can't treat the
28506         operation as a normal gather load or scatter store.
28507         (get_group_load_store_type): Take the gather_scatter_info
28508         as argument.  Try using a gather load or scatter store for
28509         single-element groups.
28510         (get_load_store_type): Update calls to get_group_load_store_type
28511         and vect_use_strided_gather_scatters_p.
28513 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28514             Alan Hayward  <alan.hayward@arm.com>
28515             David Sherwood  <david.sherwood@arm.com>
28517         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
28518         optional tree argument.
28519         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
28520         null target hooks.
28521         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
28522         but continue to use the current value as a fallback.
28523         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
28524         to compare the updates.
28525         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
28526         (get_load_store_type): Use it when handling a strided access.
28527         (vect_get_strided_load_store_ops): New function.
28528         (vect_get_data_ptr_increment): Likewise.
28529         (vectorizable_load): Handle strided gather loads.  Always pass
28530         a step to vect_create_data_ref_ptr and bump_vector_ptr.
28532 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28533             Alan Hayward  <alan.hayward@arm.com>
28534             David Sherwood  <david.sherwood@arm.com>
28536         * doc/md.texi (gather_load@var{m}): Document.
28537         (mask_gather_load@var{m}): Likewise.
28538         * genopinit.c (main): Add supports_vec_gather_load and
28539         supports_vec_gather_load_cached to target_optabs.
28540         * optabs-tree.c (init_tree_optimization_optabs): Use
28541         ggc_cleared_alloc to allocate target_optabs.
28542         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
28543         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
28544         functions.
28545         * internal-fn.h (internal_load_fn_p): Declare.
28546         (internal_gather_scatter_fn_p): Likewise.
28547         (internal_fn_mask_index): Likewise.
28548         (internal_gather_scatter_fn_supported_p): Likewise.
28549         * internal-fn.c (gather_load_direct): New macro.
28550         (expand_gather_load_optab_fn): New function.
28551         (direct_gather_load_optab_supported_p): New macro.
28552         (direct_internal_fn_optab): New function.
28553         (internal_load_fn_p): Likewise.
28554         (internal_gather_scatter_fn_p): Likewise.
28555         (internal_fn_mask_index): Likewise.
28556         (internal_gather_scatter_fn_supported_p): Likewise.
28557         * optabs-query.c (supports_at_least_one_mode_p): New function.
28558         (supports_vec_gather_load_p): Likewise.
28559         * optabs-query.h (supports_vec_gather_load_p): Declare.
28560         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
28561         and memory_type field.
28562         (NUM_PATTERNS): Bump to 15.
28563         * tree-vect-data-refs.c: Include internal-fn.h.
28564         (vect_gather_scatter_fn_p): New function.
28565         (vect_describe_gather_scatter_call): Likewise.
28566         (vect_check_gather_scatter): Try using internal functions for
28567         gather loads.  Recognize existing calls to a gather load function.
28568         (vect_analyze_data_refs): Consider using gather loads if
28569         supports_vec_gather_load_p.
28570         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
28571         (vect_get_gather_scatter_offset_type): Likewise.
28572         (vect_convert_mask_for_vectype): Likewise.
28573         (vect_add_conversion_to_patterm): Likewise.
28574         (vect_try_gather_scatter_pattern): Likewise.
28575         (vect_recog_gather_scatter_pattern): New pattern recognizer.
28576         (vect_vect_recog_func_ptrs): Add it.
28577         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
28578         internal_fn_mask_index and internal_gather_scatter_fn_p.
28579         (check_load_store_masking): Take the gather_scatter_info as an
28580         argument and handle gather loads.
28581         (vect_get_gather_scatter_ops): New function.
28582         (vectorizable_call): Check internal_load_fn_p.
28583         (vectorizable_load): Likewise.  Handle gather load internal
28584         functions.
28585         (vectorizable_store): Update call to check_load_store_masking.
28586         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
28587         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
28588         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
28589         (aarch64_gather_scale_operand_d): New predicates.
28590         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
28591         (mask_gather_load<mode>): New insns.
28593 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28594             Alan Hayward  <alan.hayward@arm.com>
28595             David Sherwood  <david.sherwood@arm.com>
28597         * optabs.def (fold_left_plus_optab): New optab.
28598         * doc/md.texi (fold_left_plus_@var{m}): Document.
28599         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
28600         * internal-fn.c (fold_left_direct): Define.
28601         (expand_fold_left_optab_fn): Likewise.
28602         (direct_fold_left_optab_supported_p): Likewise.
28603         * fold-const-call.c (fold_const_fold_left): New function.
28604         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
28605         * tree-parloops.c (valid_reduction_p): New function.
28606         (gather_scalar_reductions): Use it.
28607         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
28608         (vect_finish_replace_stmt): Declare.
28609         * tree-vect-loop.c (fold_left_reduction_fn): New function.
28610         (needs_fold_left_reduction_p): New function, split out from...
28611         (vect_is_simple_reduction): ...here.  Accept reductions that
28612         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
28613         (vect_force_simple_reduction): Also store the reduction type in
28614         the assignment's STMT_VINFO_REDUC_TYPE.
28615         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
28616         (merge_with_identity): New function.
28617         (vect_expand_fold_left): Likewise.
28618         (vectorize_fold_left_reduction): Likewise.
28619         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
28620         scalar phi in place for it.  Check for target support and reject
28621         cases that would reassociate the operation.  Defer the transform
28622         phase to vectorize_fold_left_reduction.
28623         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
28624         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
28625         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
28627 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28629         * tree-if-conv.c (predicate_mem_writes): Remove redundant
28630         call to ifc_temp_var.
28632 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28633             Alan Hayward  <alan.hayward@arm.com>
28634             David Sherwood  <david.sherwood@arm.com>
28636         * target.def (legitimize_address_displacement): Take the original
28637         offset as a poly_int.
28638         * targhooks.h (default_legitimize_address_displacement): Update
28639         accordingly.
28640         * targhooks.c (default_legitimize_address_displacement): Likewise.
28641         * doc/tm.texi: Regenerate.
28642         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
28643         as an argument, moving assert of ad->disp == ad->disp_term to...
28644         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
28645         Try calling targetm.legitimize_address_displacement before expanding
28646         the address rather than afterwards, and adjust for the new interface.
28647         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
28648         Match the new hook interface.  Handle SVE addresses.
28649         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
28650         new hook interface.
28652 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28654         * Makefile.in (OBJS): Add early-remat.o.
28655         * target.def (select_early_remat_modes): New hook.
28656         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
28657         * doc/tm.texi: Regenerate.
28658         * targhooks.h (default_select_early_remat_modes): Declare.
28659         * targhooks.c (default_select_early_remat_modes): New function.
28660         * timevar.def (TV_EARLY_REMAT): New timevar.
28661         * passes.def (pass_early_remat): New pass.
28662         * tree-pass.h (make_pass_early_remat): Declare.
28663         * early-remat.c: New file.
28664         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
28665         function.
28666         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
28668 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28669             Alan Hayward  <alan.hayward@arm.com>
28670             David Sherwood  <david.sherwood@arm.com>
28672         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
28673         vfm1 with a bound_epilog parameter.
28674         (vect_do_peeling): Update calls accordingly, and move the prologue
28675         call earlier in the function.  Treat the base bound_epilog as 0 for
28676         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
28677         this base when peeling for gaps.
28678         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
28679         with fully-masked loops.
28680         (vect_estimate_min_profitable_iters): Handle the single peeled
28681         iteration in that case.
28683 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28684             Alan Hayward  <alan.hayward@arm.com>
28685             David Sherwood  <david.sherwood@arm.com>
28687         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
28688         single-element interleaving even if the size is not a power of 2.
28689         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
28690         accesses for single-element interleaving if the group size is
28691         not a power of 2.
28693 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28694             Alan Hayward  <alan.hayward@arm.com>
28695             David Sherwood  <david.sherwood@arm.com>
28697         * doc/md.texi (fold_extract_last_@var{m}): Document.
28698         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
28699         * optabs.def (fold_extract_last_optab): New optab.
28700         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
28701         * internal-fn.c (fold_extract_direct): New macro.
28702         (expand_fold_extract_optab_fn): Likewise.
28703         (direct_fold_extract_optab_supported_p): Likewise.
28704         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
28705         * tree-vect-loop.c (vect_model_reduction_cost): Handle
28706         EXTRACT_LAST_REDUCTION.
28707         (get_initial_def_for_reduction): Do not create an initial vector
28708         for EXTRACT_LAST_REDUCTION reductions.
28709         (vectorizable_reduction): Leave the scalar phi in place for
28710         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
28711         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
28712         epilogue code for EXTRACT_LAST_REDUCTION and defer the
28713         transform phase to vectorizable_condition.
28714         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
28715         split out from...
28716         (vect_finish_stmt_generation): ...here.
28717         (vect_finish_replace_stmt): New function.
28718         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
28719         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
28720         pattern.
28721         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
28723 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28724             Alan Hayward  <alan.hayward@arm.com>
28725             David Sherwood  <david.sherwood@arm.com>
28727         * doc/md.texi (extract_last_@var{m}): Document.
28728         * optabs.def (extract_last_optab): New optab.
28729         * internal-fn.def (EXTRACT_LAST): New internal function.
28730         * internal-fn.c (cond_unary_direct): New macro.
28731         (expand_cond_unary_optab_fn): Likewise.
28732         (direct_cond_unary_optab_supported_p): Likewise.
28733         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
28734         loops using EXTRACT_LAST.
28735         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
28736         (extract_last_<mode>): ...this optab.
28737         (vec_extract<mode><Vel>): Update accordingly.
28739 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28740             Alan Hayward  <alan.hayward@arm.com>
28741             David Sherwood  <david.sherwood@arm.com>
28743         * target.def (empty_mask_is_expensive): New hook.
28744         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
28745         * doc/tm.texi: Regenerate.
28746         * targhooks.h (default_empty_mask_is_expensive): Declare.
28747         * targhooks.c (default_empty_mask_is_expensive): New function.
28748         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
28749         if the target says that empty masks are expensive.
28750         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
28751         New function.
28752         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
28754 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28755             Alan Hayward  <alan.hayward@arm.com>
28756             David Sherwood  <david.sherwood@arm.com>
28758         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
28759         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
28760         (vect_use_loop_mask_for_alignment_p): New function.
28761         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
28762         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
28763         niters_skip argument.  Make sure that the first niters_skip elements
28764         of the first iteration are inactive.
28765         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
28766         Update call to vect_set_loop_masks_directly.
28767         (get_misalign_in_elems): New function, split out from...
28768         (vect_gen_prolog_loop_niters): ...here.
28769         (vect_update_init_of_dr): Take a code argument that specifies whether
28770         the adjustment should be added or subtracted.
28771         (vect_update_init_of_drs): Likewise.
28772         (vect_prepare_for_masked_peels): New function.
28773         (vect_do_peeling): Skip prologue peeling if we're using a mask
28774         instead.  Update call to vect_update_inits_of_drs.
28775         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28776         mask_skip_niters.
28777         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
28778         alignment.  Do not include the number of peeled iterations in
28779         the minimum threshold in that case.
28780         (vectorizable_induction): Adjust the start value down by
28781         LOOP_VINFO_MASK_SKIP_NITERS iterations.
28782         (vect_transform_loop): Call vect_prepare_for_masked_peels.
28783         Take the number of skipped iterations into account when calculating
28784         the loop bounds.
28785         * tree-vect-stmts.c (vect_gen_while_not): New function.
28787 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28788             Alan Hayward  <alan.hayward@arm.com>
28789             David Sherwood  <david.sherwood@arm.com>
28791         * doc/sourcebuild.texi (vect_fully_masked): Document.
28792         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
28793         default value to 0.
28794         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
28795         split out from...
28796         (vect_analyze_loop_2): ...here. Don't check the vectorization
28797         factor against the number of loop iterations if the loop is
28798         fully-masked.
28800 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28801             Alan Hayward  <alan.hayward@arm.com>
28802             David Sherwood  <david.sherwood@arm.com>
28804         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
28805         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
28806         (dump_groups): Update accordingly.
28807         (iv_use::mem_type): New member variable.
28808         (address_p): New function.
28809         (record_use): Add a mem_type argument and initialize the new
28810         mem_type field.
28811         (record_group_use): Add a mem_type argument.  Use address_p.
28812         Remove obsolete null checks of base_object.  Update call to record_use.
28813         (find_interesting_uses_op): Update call to record_group_use.
28814         (find_interesting_uses_cond): Likewise.
28815         (find_interesting_uses_address): Likewise.
28816         (get_mem_type_for_internal_fn): New function.
28817         (find_address_like_use): Likewise.
28818         (find_interesting_uses_stmt): Try find_address_like_use before
28819         calling find_interesting_uses_op.
28820         (addr_offset_valid_p): Use the iv mem_type field as the type
28821         of the addressed memory.
28822         (add_autoinc_candidates): Likewise.
28823         (get_address_cost): Likewise.
28824         (split_small_address_groups_p): Use address_p.
28825         (split_address_groups): Likewise.
28826         (add_iv_candidate_for_use): Likewise.
28827         (autoinc_possible_for_pair): Likewise.
28828         (rewrite_groups): Likewise.
28829         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
28830         (determine_group_iv_cost): Update after split of USE_ADDRESS.
28831         (get_alias_ptr_type_for_ptr_address): New function.
28832         (rewrite_use_address): Rewrite address uses in calls that were
28833         identified by find_address_like_use.
28835 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28836             Alan Hayward  <alan.hayward@arm.com>
28837             David Sherwood  <david.sherwood@arm.com>
28839         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
28840         TARGET_MEM_REFs.
28841         * gimple-expr.h (is_gimple_addressable: Likewise.
28842         * gimple-expr.c (is_gimple_address): Likewise.
28843         * internal-fn.c (expand_call_mem_ref): New function.
28844         (expand_mask_load_optab_fn): Use it.
28845         (expand_mask_store_optab_fn): Likewise.
28847 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28848             Alan Hayward  <alan.hayward@arm.com>
28849             David Sherwood  <david.sherwood@arm.com>
28851         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
28852         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
28853         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
28854         (cond_umax@var{mode}): Document.
28855         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
28856         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
28857         (cond_umin_optab, cond_umax_optab): New optabs.
28858         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
28859         (COND_IOR, COND_XOR): New internal functions.
28860         * internal-fn.h (get_conditional_internal_fn): Declare.
28861         * internal-fn.c (cond_binary_direct): New macro.
28862         (expand_cond_binary_optab_fn): Likewise.
28863         (direct_cond_binary_optab_supported_p): Likewise.
28864         (get_conditional_internal_fn): New function.
28865         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
28866         Cope with reduction statements that are vectorized as calls rather
28867         than assignments.
28868         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
28869         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
28870         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
28871         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
28872         (UNSPEC_COND_EOR): New unspecs.
28873         (optab): Add mappings for them.
28874         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
28875         (sve_int_op, sve_fp_op): New int attributes.
28877 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28878             Alan Hayward  <alan.hayward@arm.com>
28879             David Sherwood  <david.sherwood@arm.com>
28881         * optabs.def (while_ult_optab): New optab.
28882         * doc/md.texi (while_ult@var{m}@var{n}): Document.
28883         * internal-fn.def (WHILE_ULT): New internal function.
28884         * internal-fn.h (direct_internal_fn_supported_p): New override
28885         that takes two types as argument.
28886         * internal-fn.c (while_direct): New macro.
28887         (expand_while_optab_fn): New function.
28888         (convert_optab_supported_p): Likewise.
28889         (direct_while_optab_supported_p): New macro.
28890         * wide-int.h (wi::udiv_ceil): New function.
28891         * tree-vectorizer.h (rgroup_masks): New structure.
28892         (vec_loop_masks): New typedef.
28893         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
28894         and fully_masked_p.
28895         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
28896         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
28897         (vect_max_vf): New function.
28898         (slpeel_make_loop_iterate_ntimes): Delete.
28899         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
28900         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
28901         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
28902         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
28903         internal-fn.h, stor-layout.h and optabs-query.h.
28904         (vect_set_loop_mask): New function.
28905         (add_preheader_seq): Likewise.
28906         (add_header_seq): Likewise.
28907         (interleave_supported_p): Likewise.
28908         (vect_maybe_permute_loop_masks): Likewise.
28909         (vect_set_loop_masks_directly): Likewise.
28910         (vect_set_loop_condition_masked): Likewise.
28911         (vect_set_loop_condition_unmasked): New function, split out from
28912         slpeel_make_loop_iterate_ntimes.
28913         (slpeel_make_loop_iterate_ntimes): Rename to..
28914         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
28915         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
28916         (vect_do_peeling): Update call accordingly.
28917         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
28918         loops.
28919         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
28920         mask_compare_type, can_fully_mask_p and fully_masked_p.
28921         (release_vec_loop_masks): New function.
28922         (_loop_vec_info): Use it to free the loop masks.
28923         (can_produce_all_loop_masks_p): New function.
28924         (vect_get_max_nscalars_per_iter): Likewise.
28925         (vect_verify_full_masking): Likewise.
28926         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
28927         retries, and free the mask rgroups before retrying.  Check loop-wide
28928         reasons for disallowing fully-masked loops.  Make the final decision
28929         about whether use a fully-masked loop or not.
28930         (vect_estimate_min_profitable_iters): Do not assume that peeling
28931         for the number of iterations will be needed for fully-masked loops.
28932         (vectorizable_reduction): Disable fully-masked loops.
28933         (vectorizable_live_operation): Likewise.
28934         (vect_halve_mask_nunits): New function.
28935         (vect_double_mask_nunits): Likewise.
28936         (vect_record_loop_mask): Likewise.
28937         (vect_get_loop_mask): Likewise.
28938         (vect_transform_loop): Handle the case in which the final loop
28939         iteration might handle a partial vector.  Call vect_set_loop_condition
28940         instead of slpeel_make_loop_iterate_ntimes.
28941         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
28942         (check_load_store_masking): New function.
28943         (prepare_load_store_mask): Likewise.
28944         (vectorizable_store): Handle fully-masked loops.
28945         (vectorizable_load): Likewise.
28946         (supportable_widening_operation): Use vect_halve_mask_nunits for
28947         booleans.
28948         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
28949         (vect_gen_while): New function.
28950         * config/aarch64/aarch64.md (umax<mode>3): New expander.
28951         (aarch64_uqdec<mode>): New insn.
28953 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28954             Alan Hayward  <alan.hayward@arm.com>
28955             David Sherwood  <david.sherwood@arm.com>
28957         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
28958         (reduc_xor_scal_optab): New optabs.
28959         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
28960         (reduc_xor_scal_@var{m}): Document.
28961         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
28962         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
28963         internal functions.
28964         * fold-const-call.c (fold_const_call): Handle them.
28965         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
28966         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
28967         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
28968         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
28969         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
28970         (UNSPEC_XORV): New unspecs.
28971         (optab): Add entries for them.
28972         (BITWISEV): New int iterator.
28973         (bit_reduc_op): New int attributes.
28975 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
28976             Alan Hayward  <alan.hayward@arm.com>
28977             David Sherwood  <david.sherwood@arm.com>
28979         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
28980         * internal-fn.def (VEC_SHL_INSERT): New internal function.
28981         * optabs.def (vec_shl_insert_optab): New optab.
28982         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
28983         (duplicate_and_interleave): Likewise.
28984         * tree-vect-loop.c: Include internal-fn.h.
28985         (neutral_op_for_slp_reduction): New function, split out from
28986         get_initial_defs_for_reduction.
28987         (get_initial_def_for_reduction): Handle option 2 for variable-length
28988         vectors by loading the neutral value into a vector and then shifting
28989         the initial value into element 0.
28990         (get_initial_defs_for_reduction): Replace the code argument with
28991         the neutral value calculated by neutral_op_for_slp_reduction.
28992         Use gimple_build_vector for constant-length vectors.
28993         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
28994         but the first group_size elements have a neutral value.
28995         Use duplicate_and_interleave otherwise.
28996         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
28997         Update call to get_initial_defs_for_reduction.  Handle SLP
28998         reductions for variable-length vectors by creating one vector
28999         result for each scalar result, with the elements associated
29000         with other scalar results stubbed out with the neutral value.
29001         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
29002         Require IFN_VEC_SHL_INSERT for double reductions on
29003         variable-length vectors, or SLP reductions that have
29004         a neutral value.  Require can_duplicate_and_interleave_p
29005         support for variable-length unchained SLP reductions if there
29006         is no neutral value, such as for MIN/MAX reductions.  Also require
29007         the number of vector elements to be a multiple of the number of
29008         SLP statements when doing variable-length unchained SLP reductions.
29009         Update call to vect_create_epilog_for_reduction.
29010         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
29011         and remove initial values.
29012         (duplicate_and_interleave): Make public.
29013         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
29014         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
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         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
29021         (can_duplicate_and_interleave_p): New function.
29022         (vect_get_and_check_slp_defs): Take the vector of statements
29023         rather than just the current one.  Remove excess parentheses.
29024         Restriction rejectinon of vect_constant_def and vect_external_def
29025         for variable-length vectors to boolean types, or types for which
29026         can_duplicate_and_interleave_p is false.
29027         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
29028         (duplicate_and_interleave): New function.
29029         (vect_get_constant_vectors): Use gimple_build_vector for
29030         constant-length vectors and suitable variable-length constant
29031         vectors.  Use duplicate_and_interleave for other variable-length
29032         vectors.  Don't defer the update when inserting new statements.
29034 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29035             Alan Hayward  <alan.hayward@arm.com>
29036             David Sherwood  <david.sherwood@arm.com>
29038         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
29039         min_profitable_iters doesn't go negative.
29041 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29042             Alan Hayward  <alan.hayward@arm.com>
29043             David Sherwood  <david.sherwood@arm.com>
29045         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
29046         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
29047         * optabs.def (vec_mask_load_lanes_optab): New optab.
29048         (vec_mask_store_lanes_optab): Likewise.
29049         * internal-fn.def (MASK_LOAD_LANES): New internal function.
29050         (MASK_STORE_LANES): Likewise.
29051         * internal-fn.c (mask_load_lanes_direct): New macro.
29052         (mask_store_lanes_direct): Likewise.
29053         (expand_mask_load_optab_fn): Handle masked operations.
29054         (expand_mask_load_lanes_optab_fn): New macro.
29055         (expand_mask_store_optab_fn): Handle masked operations.
29056         (expand_mask_store_lanes_optab_fn): New macro.
29057         (direct_mask_load_lanes_optab_supported_p): Likewise.
29058         (direct_mask_store_lanes_optab_supported_p): Likewise.
29059         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
29060         parameter.
29061         (vect_load_lanes_supported): Likewise.
29062         * tree-vect-data-refs.c (strip_conversion): New function.
29063         (can_group_stmts_p): Likewise.
29064         (vect_analyze_data_ref_accesses): Use it instead of checking
29065         for a pair of assignments.
29066         (vect_store_lanes_supported): Take a masked_p parameter.
29067         (vect_load_lanes_supported): Likewise.
29068         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
29069         vect_store_lanes_supported and vect_load_lanes_supported.
29070         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
29071         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
29072         parameter.  Don't allow gaps for masked accesses.
29073         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
29074         and vect_load_lanes_supported.
29075         (get_load_store_type): Take a masked_p parameter and update
29076         call to get_group_load_store_type.
29077         (vectorizable_store): Update call to get_load_store_type.
29078         Handle IFN_MASK_STORE_LANES.
29079         (vectorizable_load): Update call to get_load_store_type.
29080         Handle IFN_MASK_LOAD_LANES.
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         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
29087         modes for SVE.
29088         * config/aarch64/aarch64-protos.h
29089         (aarch64_sve_struct_memory_operand_p): Declare.
29090         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
29091         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
29092         (VPRED, vpred): Handle SVE structure modes.
29093         * config/aarch64/constraints.md (Utx): New constraint.
29094         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
29095         (aarch64_sve_struct_nonimmediate_operand): New predicates.
29096         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
29097         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
29098         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
29099         structure modes.  Split into pieces after RA.
29100         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
29101         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
29102         New patterns.
29103         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
29104         SVE structure modes.
29105         (aarch64_classify_address): Likewise.
29106         (sizetochar): Move earlier in file.
29107         (aarch64_print_operand): Handle SVE register lists.
29108         (aarch64_array_mode): New function.
29109         (aarch64_sve_struct_memory_operand_p): Likewise.
29110         (TARGET_ARRAY_MODE): Redefine.
29112 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29113             Alan Hayward  <alan.hayward@arm.com>
29114             David Sherwood  <david.sherwood@arm.com>
29116         * target.def (array_mode): New target hook.
29117         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
29118         * doc/tm.texi: Regenerate.
29119         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
29120         * hooks.c (hook_optmode_mode_uhwi_none): New function.
29121         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
29122         targetm.array_mode.
29123         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
29124         type sizes.
29126 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29127             Alan Hayward  <alan.hayward@arm.com>
29128             David Sherwood  <david.sherwood@arm.com>
29130         * fold-const.c (fold_binary_loc): Check the argument types
29131         rather than the result type when testing for a vector operation.
29133 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29135         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
29136         * doc/tm.texi: Regenerate.
29138 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29139             Alan Hayward  <alan.hayward@arm.com>
29140             David Sherwood  <david.sherwood@arm.com>
29142         * doc/invoke.texi (-msve-vector-bits=): Document new option.
29143         (sve): Document new AArch64 extension.
29144         * doc/md.texi (w): Extend the description of the AArch64
29145         constraint to include SVE vectors.
29146         (Upl, Upa): Document new AArch64 predicate constraints.
29147         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
29148         enum.
29149         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
29150         (msve-vector-bits=): New option.
29151         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
29152         SVE when these are disabled.
29153         (sve): New extension.
29154         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
29155         modes.  Adjust their number of units based on aarch64_sve_vg.
29156         (MAX_BITSIZE_MODE_ANY_MODE): Define.
29157         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
29158         aarch64_addr_query_type.
29159         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
29160         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
29161         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
29162         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
29163         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
29164         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
29165         (aarch64_simd_imm_zero_p): Delete.
29166         (aarch64_check_zero_based_sve_index_immediate): Declare.
29167         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29168         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29169         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29170         (aarch64_sve_float_mul_immediate_p): Likewise.
29171         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29172         rather than an rtx.
29173         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
29174         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
29175         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
29176         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
29177         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
29178         (aarch64_regmode_natural_size): Likewise.
29179         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
29180         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
29181         left one place.
29182         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
29183         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
29184         for VG and the SVE predicate registers.
29185         (V_ALIASES): Add a "z"-prefixed alias.
29186         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
29187         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
29188         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
29189         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
29190         (REG_CLASS_NAMES): Add entries for them.
29191         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
29192         and the predicate registers.
29193         (aarch64_sve_vg): Declare.
29194         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
29195         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
29196         (REGMODE_NATURAL_SIZE): Define.
29197         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
29198         SVE macros.
29199         * config/aarch64/aarch64.c: Include cfgrtl.h.
29200         (simd_immediate_info): Add a constructor for series vectors,
29201         and an associated step field.
29202         (aarch64_sve_vg): New variable.
29203         (aarch64_dbx_register_number): Handle VG and the predicate registers.
29204         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
29205         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
29206         (VEC_ANY_DATA, VEC_STRUCT): New constants.
29207         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
29208         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
29209         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
29210         (aarch64_get_mask_mode): New functions.
29211         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
29212         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
29213         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
29214         predicate modes and predicate registers.  Explicitly restrict
29215         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
29216         to store a vector mode if it is recognized by
29217         aarch64_classify_vector_mode.
29218         (aarch64_regmode_natural_size): New function.
29219         (aarch64_hard_regno_caller_save_mode): Return the original mode
29220         for predicates.
29221         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
29222         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
29223         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
29224         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
29225         functions.
29226         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
29227         does not overlap dest if the function is frame-related.  Handle
29228         SVE constants.
29229         (aarch64_split_add_offset): New function.
29230         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
29231         them aarch64_add_offset.
29232         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
29233         and update call to aarch64_sub_sp.
29234         (aarch64_add_cfa_expression): New function.
29235         (aarch64_expand_prologue): Pass extra temporary registers to the
29236         functions above.  Handle the case in which we need to emit new
29237         DW_CFA_expressions for registers that were originally saved
29238         relative to the stack pointer, but now have to be expressed
29239         relative to the frame pointer.
29240         (aarch64_output_mi_thunk): Pass extra temporary registers to the
29241         functions above.
29242         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
29243         IP0 and IP1 values for SVE frames.
29244         (aarch64_expand_vec_series): New function.
29245         (aarch64_expand_sve_widened_duplicate): Likewise.
29246         (aarch64_expand_sve_const_vector): Likewise.
29247         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
29248         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
29249         into the register, rather than emitting a SET directly.
29250         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
29251         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
29252         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
29253         (offset_9bit_signed_scaled_p): New functions.
29254         (aarch64_replicate_bitmask_imm): New function.
29255         (aarch64_bitmask_imm): Use it.
29256         (aarch64_cannot_force_const_mem): Reject expressions involving
29257         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
29258         (aarch64_classify_index): Handle SVE indices, by requiring
29259         a plain register index with a scale that matches the element size.
29260         (aarch64_classify_address): Handle SVE addresses.  Assert that
29261         the mode of the address is VOIDmode or an integer mode.
29262         Update call to aarch64_classify_symbol.
29263         (aarch64_classify_symbolic_expression): Update call to
29264         aarch64_classify_symbol.
29265         (aarch64_const_vec_all_in_range_p): New function.
29266         (aarch64_print_vector_float_operand): Likewise.
29267         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
29268         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
29269         and the FP immediates 1.0 and 0.5.
29270         (aarch64_print_address_internal): Handle SVE addresses.
29271         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
29272         (aarch64_regno_regclass): Handle predicate registers.
29273         (aarch64_secondary_reload): Handle big-endian reloads of SVE
29274         data modes.
29275         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
29276         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
29277         (aarch64_convert_sve_vector_bits): New function.
29278         (aarch64_override_options): Use it to handle -msve-vector-bits=.
29279         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
29280         rather than an rtx.
29281         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
29282         Handle SVE vector and predicate modes.  Accept VL-based constants
29283         that need only one temporary register, and VL offsets that require
29284         no temporary registers.
29285         (aarch64_conditional_register_usage): Mark the predicate registers
29286         as fixed if SVE isn't available.
29287         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
29288         Return true for SVE vector and predicate modes.
29289         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
29290         rather than an unsigned int.  Handle SVE modes.
29291         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
29292         SVE modes.
29293         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
29294         if SVE is enabled.
29295         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
29296         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
29297         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
29298         (aarch64_sve_float_mul_immediate_p): New functions.
29299         (aarch64_sve_valid_immediate): New function.
29300         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
29301         Explicitly reject structure modes.  Check for INDEX constants.
29302         Handle PTRUE and PFALSE constants.
29303         (aarch64_check_zero_based_sve_index_immediate): New function.
29304         (aarch64_simd_imm_zero_p): Delete.
29305         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
29306         vector modes.  Accept constants in the range of CNT[BHWD].
29307         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
29308         ask for an Advanced SIMD mode.
29309         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
29310         (aarch64_simd_vector_alignment): Handle SVE predicates.
29311         (aarch64_vectorize_preferred_vector_alignment): New function.
29312         (aarch64_simd_vector_alignment_reachable): Use it instead of
29313         the vector size.
29314         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
29315         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
29316         functions.
29317         (MAX_VECT_LEN): Delete.
29318         (expand_vec_perm_d): Add a vec_flags field.
29319         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
29320         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
29321         (aarch64_evpc_ext): Don't apply a big-endian lane correction
29322         for SVE modes.
29323         (aarch64_evpc_rev): Rename to...
29324         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
29325         (aarch64_evpc_rev_global): New function.
29326         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
29327         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
29328         MAX_VECT_LEN.
29329         (aarch64_evpc_sve_tbl): New function.
29330         (aarch64_expand_vec_perm_const_1): Update after rename of
29331         aarch64_evpc_rev.  Handle SVE permutes too, trying
29332         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
29333         than aarch64_evpc_tbl.
29334         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
29335         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
29336         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
29337         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
29338         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
29339         (aarch64_expand_sve_vcond): New functions.
29340         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
29341         of aarch64_vector_mode_p.
29342         (aarch64_dwarf_poly_indeterminate_value): New function.
29343         (aarch64_compute_pressure_classes): Likewise.
29344         (aarch64_can_change_mode_class): Likewise.
29345         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
29346         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
29347         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
29348         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
29349         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
29350         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
29351         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
29352         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
29353         constraints.
29354         (Dn, Dl, Dr): Accept const as well as const_vector.
29355         (Dz): Likewise.  Compare against CONST0_RTX.
29356         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
29357         of "vector" where appropriate.
29358         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
29359         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
29360         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
29361         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
29362         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
29363         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
29364         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
29365         (v_int_equiv): Extend to SVE modes.
29366         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
29367         mode attributes.
29368         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
29369         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
29370         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
29371         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
29372         (SVE_COND_FP_CMP): New int iterators.
29373         (perm_hilo): Handle the new unpack unspecs.
29374         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
29375         attributes.
29376         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
29377         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
29378         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
29379         (aarch64_equality_operator, aarch64_constant_vector_operand)
29380         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
29381         (aarch64_sve_nonimmediate_operand): Likewise.
29382         (aarch64_sve_general_operand): Likewise.
29383         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
29384         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
29385         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
29386         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
29387         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
29388         (aarch64_sve_float_arith_immediate): Likewise.
29389         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
29390         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
29391         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
29392         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
29393         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
29394         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
29395         (aarch64_sve_float_arith_operand): Likewise.
29396         (aarch64_sve_float_arith_with_sub_operand): Likewise.
29397         (aarch64_sve_float_mul_operand): Likewise.
29398         (aarch64_sve_vec_perm_operand): Likewise.
29399         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
29400         (aarch64_mov_operand): Accept const_poly_int and const_vector.
29401         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
29402         as well as const_vector.
29403         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
29404         in file.  Use CONST0_RTX and CONSTM1_RTX.
29405         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
29406         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
29407         Use aarch64_simd_imm_zero.
29408         * config/aarch64/aarch64-sve.md: New file.
29409         * config/aarch64/aarch64.md: Include it.
29410         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
29411         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
29412         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
29413         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
29414         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
29415         (sve): New attribute.
29416         (enabled): Disable instructions with the sve attribute unless
29417         TARGET_SVE.
29418         (movqi, movhi): Pass CONST_POLY_INT operaneds through
29419         aarch64_expand_mov_immediate.
29420         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
29421         CNT[BHSD] immediates.
29422         (movti): Split CONST_POLY_INT moves into two halves.
29423         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
29424         Split additions that need a temporary here if the destination
29425         is the stack pointer.
29426         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
29427         (*add<mode>3_poly_1): New instruction.
29428         (set_clobber_cc): New expander.
29430 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29432         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
29433         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
29434         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
29435         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
29436         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
29437         Change innermode from fixed_mode_size to machine_mode.
29438         (simplify_subreg): Update call accordingly.  Handle a constant-sized
29439         subreg of a variable-length CONST_VECTOR.
29441 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
29442             Alan Hayward  <alan.hayward@arm.com>
29443             David Sherwood  <david.sherwood@arm.com>
29445         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
29446         (add_offset_to_base): New function, split out from...
29447         (create_mem_ref): ...here.  When handling a scale other than 1,
29448         check first whether the address is valid without the offset.
29449         Add it into the base if so, leaving the index and scale as-is.
29451 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29453         PR c++/83778
29454         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
29455         fold_for_warn before checking if arg2 is INTEGER_CST.
29457 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
29459         * config/rs6000/predicates.md (load_multiple_operation): Delete.
29460         (store_multiple_operation): Delete.
29461         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
29462         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
29463         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
29464         guarded by TARGET_STRING.
29465         (rs6000_output_load_multiple): Delete.
29466         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29467         OPTION_MASK_STRING / TARGET_STRING handling.
29468         (print_operand) <'N', 'O'>: Add comment that these are unused now.
29469         (const rs6000_opt_masks) <"string">: Change mask to 0.
29470         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
29471         (MASK_STRING): Delete.
29472         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
29473         parts.  Simplify.
29474         (load_multiple): Delete.
29475         (*ldmsi8): Delete.
29476         (*ldmsi7): Delete.
29477         (*ldmsi6): Delete.
29478         (*ldmsi5): Delete.
29479         (*ldmsi4): Delete.
29480         (*ldmsi3): Delete.
29481         (store_multiple): Delete.
29482         (*stmsi8): Delete.
29483         (*stmsi7): Delete.
29484         (*stmsi6): Delete.
29485         (*stmsi5): Delete.
29486         (*stmsi4): Delete.
29487         (*stmsi3): Delete.
29488         (movmemsi_8reg): Delete.
29489         (corresponding unnamed define_insn): Delete.
29490         (movmemsi_6reg): Delete.
29491         (corresponding unnamed define_insn): Delete.
29492         (movmemsi_4reg): Delete.
29493         (corresponding unnamed define_insn): Delete.
29494         (movmemsi_2reg): Delete.
29495         (corresponding unnamed define_insn): Delete.
29496         (movmemsi_1reg): Delete.
29497         (corresponding unnamed define_insn): Delete.
29498         * config/rs6000/rs6000.opt (mno-string): New.
29499         (mstring): Replace by deprecation warning stub.
29500         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
29502 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
29504         * regrename.c (regrename_do_replace): If replacing the same
29505         reg multiple times, try to reuse last created gen_raw_REG.
29507         PR debug/81155
29508         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
29509         main to workaround a bug in GDB.
29511 2018-01-12  Tom de Vries  <tom@codesourcery.com>
29513         PR target/83737
29514         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
29516 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
29518         PR rtl-optimization/80481
29519         * ira-color.c (get_cap_member): New function.
29520         (allocnos_conflict_by_live_ranges_p): Use it.
29521         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
29522         (setup_slot_coalesced_allocno_live_ranges): Ditto.
29524 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
29526         PR target/83628
29527         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
29528         (*saddl_se_1): Ditto.
29529         (*ssubsi_1): Ditto.
29530         (*ssubl_se_1): Ditto.
29532 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29534         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
29535         rather than wi::to_widest for DR_INITs.
29536         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
29537         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
29538         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
29539         INTEGER_CSTs.
29540         (vect_analyze_group_access_1): Note that here.
29542 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29544         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
29545         polynomial type sizes.
29547 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
29549         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
29550         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
29551         (gimple_add_tmp_var): Likewise.
29553 2018-01-12  Martin Liska  <mliska@suse.cz>
29555         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
29556         (gimple_alloc_sizes): Likewise.
29557         (dump_gimple_statistics): Use PRIu64 in printf format.
29558         * gimple.h: Change uint64_t to int.
29560 2018-01-12  Martin Liska  <mliska@suse.cz>
29562         * tree-core.h: Use uint64_t instead of int.
29563         * tree.c (tree_node_counts): Likewise.
29564         (tree_node_sizes): Likewise.
29565         (dump_tree_statistics): Use PRIu64 in printf format.
29567 2018-01-12  Martin Liska  <mliska@suse.cz>
29569         * Makefile.in: As qsort_chk is implemented in vec.c, add
29570         vec.o to linkage of gencfn-macros.
29571         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
29572         passing the info to record_node_allocation_statistics.
29573         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
29574         and pass the info.
29575         * ggc-common.c (struct ggc_usage): Add operator== and use
29576         it in operator< and compare function.
29577         * mem-stats.h (struct mem_usage): Likewise.
29578         * vec.c (struct vec_usage): Remove operator< and compare
29579         function. Can be simply inherited.
29581 2018-01-12  Martin Jambor  <mjambor@suse.cz>
29583         PR target/81616
29584         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
29585         * tree-ssa-math-opts.c: Include domwalk.h.
29586         (convert_mult_to_fma_1): New function.
29587         (fma_transformation_info): New type.
29588         (fma_deferring_state): Likewise.
29589         (cancel_fma_deferring): New function.
29590         (result_of_phi): Likewise.
29591         (last_fma_candidate_feeds_initial_phi): Likewise.
29592         (convert_mult_to_fma): Added deferring logic, split actual
29593         transformation to convert_mult_to_fma_1.
29594         (math_opts_dom_walker): New type.
29595         (math_opts_dom_walker::after_dom_children): New method, body moved
29596         here from pass_optimize_widening_mul::execute, added deferring logic
29597         bits.
29598         (pass_optimize_widening_mul::execute): Moved most of code to
29599         math_opts_dom_walker::after_dom_children.
29600         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
29601         * config/i386/i386.c (ix86_option_override_internal): Added
29602         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
29604 2018-01-12  Richard Biener  <rguenther@suse.de>
29606         PR debug/83157
29607         * dwarf2out.c (gen_variable_die): Do not reset old_die for
29608         inline instance vars.
29610 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
29612         PR target/81819
29613         * config/rx/rx.c (rx_is_restricted_memory_address):
29614         Handle SUBREG case.
29616 2018-01-12  Richard Biener  <rguenther@suse.de>
29618         PR tree-optimization/80846
29619         * target.def (split_reduction): New target hook.
29620         * targhooks.c (default_split_reduction): New function.
29621         * targhooks.h (default_split_reduction): Declare.
29622         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
29623         target requests first reduce vectors by combining low and high
29624         parts.
29625         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
29626         (get_vectype_for_scalar_type_and_size): Export.
29627         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
29628         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
29629         * doc/tm.texi: Regenerate.
29630         * config/i386/i386.c (ix86_split_reduction): Implement
29631         TARGET_VECTORIZE_SPLIT_REDUCTION.
29633 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29635         PR target/83368
29636         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
29637         in PIC mode except for TARGET_VXWORKS_RTP.
29638         * config/sparc/sparc.c: Include cfgrtl.h.
29639         (TARGET_INIT_PIC_REG): Define.
29640         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
29641         (sparc_pic_register_p): New predicate.
29642         (sparc_legitimate_address_p): Use it.
29643         (sparc_legitimize_pic_address): Likewise.
29644         (sparc_delegitimize_address): Likewise.
29645         (sparc_mode_dependent_address_p): Likewise.
29646         (gen_load_pcrel_sym): Remove 4th parameter.
29647         (load_got_register): Adjust call to above.  Remove obsolete stuff.
29648         (sparc_expand_prologue): Do not call load_got_register here.
29649         (sparc_flat_expand_prologue): Likewise.
29650         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
29651         (sparc_use_pseudo_pic_reg): New function.
29652         (sparc_init_pic_reg): Likewise.
29653         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
29654         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
29656 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
29658         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
29659         Add item for branch_cost.
29661 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
29663         PR rtl-optimization/83565
29664         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
29665         not extend the result to a larger mode for rotate operations.
29666         (num_sign_bit_copies1): Likewise.
29668 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
29670         PR target/40411
29671         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
29672         -symbolic.
29673         Use values-Xc.o for -pedantic.
29674         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
29676 2018-01-12  Martin Liska  <mliska@suse.cz>
29678         PR ipa/83054
29679         * ipa-devirt.c (final_warning_record::grow_type_warnings):
29680         New function.
29681         (possible_polymorphic_call_targets): Use it.
29682         (ipa_devirt): Likewise.
29684 2018-01-12  Martin Liska  <mliska@suse.cz>
29686         * profile-count.h (enum profile_quality): Use 0 as invalid
29687         enum value of profile_quality.
29689 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
29691         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
29692         -mext-string options.
29694 2018-01-12  Richard Biener  <rguenther@suse.de>
29696         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
29697         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
29698         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
29699         Likewise.
29700         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
29702 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
29704         * configure.ac (--with-long-double-format): Add support for the
29705         configuration option to change the default long double format on
29706         PowerPC systems.
29707         * config.gcc (powerpc*-linux*-*): Likewise.
29708         * configure: Regenerate.
29709         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
29710         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
29711         used without modification.
29713 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
29715         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
29716         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
29717         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
29718         MISC_BUILTIN_SPEC_BARRIER.
29719         (rs6000_init_builtins): Likewise.
29720         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
29721         enum value.
29722         (speculation_barrier): New define_insn.
29723         * doc/extend.texi: Document __builtin_speculation_barrier.
29725 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29727         PR target/83203
29728         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
29729         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
29730         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
29731         iterators.
29732         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
29733         integral modes instead of "ss" and "sd".
29734         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
29735         vectors with 32-bit and 64-bit elements.
29736         (vecdupssescalarmodesuffix): New mode attribute.
29737         (vec_dup<mode>): Use it.
29739 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
29741         PR target/83330
29742         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
29743         frame if argument is passed on stack.
29745 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
29747         PR target/82682
29748         * ree.c (combine_reaching_defs): Optimize also
29749         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
29750         reg2=any_extend(exp); reg1=reg2;, formatting fix.
29752 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29754         PR middle-end/83189
29755         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
29757 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
29759         PR middle-end/83718
29760         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
29761         after they are computed.
29763 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
29765         PR tree-optimization/83695
29766         * gimple-loop-linterchange.cc
29767         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
29768         reset cached scev information after interchange.
29769         (pass_linterchange::execute): Remove call to scev_reset_htab.
29771 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29773         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
29774         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
29775         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
29776         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
29777         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
29778         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
29779         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
29780         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
29781         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
29782         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
29783         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
29784         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
29785         (V_lane_reg): Likewise.
29786         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
29787         New define_expand.
29788         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
29789         (vfmal_lane_low<mode>_intrinsic,
29790         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
29791         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
29792         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
29793         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
29794         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
29795         vfmsl_lane_high<mode>_intrinsic): New define_insns.
29797 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29799         * config/arm/arm-cpus.in (fp16fml): New feature.
29800         (ALL_SIMD): Add fp16fml.
29801         (armv8.2-a): Add fp16fml as an option.
29802         (armv8.3-a): Likewise.
29803         (armv8.4-a): Add fp16fml as part of fp16.
29804         * config/arm/arm.h (TARGET_FP16FML): Define.
29805         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
29806         when appropriate.
29807         * config/arm/arm-modes.def (V2HF): Define.
29808         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
29809         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
29810         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
29811         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
29812         vfmsl_low, vfmsl_high): New set of builtins.
29813         * config/arm/iterators.md (PLUSMINUS): New code iterator.
29814         (vfml_op): New code attribute.
29815         (VFMLHALVES): New int iterator.
29816         (VFML, VFMLSEL): New mode attributes.
29817         (V_reg): Define mapping for V2HF.
29818         (V_hi, V_lo): New mode attributes.
29819         (VF_constraint): Likewise.
29820         (vfml_half, vfml_half_selector): New int attributes.
29821         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
29822         define_expand.
29823         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
29824         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
29825         New define_insn.
29826         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
29827         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
29828         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
29829         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
29830         documentation.
29831         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
29832         Document new effective target and option set.
29834 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29836         * config/arm/arm-cpus.in (armv8_4): New feature.
29837         (ARMv8_4a): New fgroup.
29838         (armv8.4-a): New arch.
29839         * config/arm/arm-tables.opt: Regenerate.
29840         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
29841         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
29842         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
29843         Add matching rules for -march=armv8.4-a and extensions.
29844         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
29846 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
29848         PR target/81821
29849         * config/rx/rx.md (BW): New mode attribute.
29850         (sync_lock_test_and_setsi): Add mode suffix to insn output.
29852 2018-01-11  Richard Biener  <rguenther@suse.de>
29854         PR tree-optimization/83435
29855         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
29856         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
29857         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
29859 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29860             Alan Hayward  <alan.hayward@arm.com>
29861             David Sherwood  <david.sherwood@arm.com>
29863         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
29864         field.
29865         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
29866         (aarch64_print_address_internal): Use it to check for a zero offset.
29868 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29869             Alan Hayward  <alan.hayward@arm.com>
29870             David Sherwood  <david.sherwood@arm.com>
29872         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
29873         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
29874         Return a poly_int64 rather than a HOST_WIDE_INT.
29875         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
29876         rather than a HOST_WIDE_INT.
29877         * config/aarch64/aarch64.h (aarch64_frame): Protect with
29878         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
29879         hard_fp_offset, frame_size, initial_adjust, callee_offset and
29880         final_offset from HOST_WIDE_INT to poly_int64.
29881         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
29882         to_constant when getting the number of units in an Advanced SIMD
29883         mode.
29884         (aarch64_builtin_vectorized_function): Check for a constant number
29885         of units.
29886         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
29887         GET_MODE_SIZE.
29888         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
29889         attribute instead of GET_MODE_NUNITS.
29890         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
29891         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
29892         GET_MODE_SIZE for fixed-size registers.
29893         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
29894         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
29895         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
29896         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
29897         (aarch64_print_operand, aarch64_print_address_internal)
29898         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
29899         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
29900         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
29901         Handle polynomial GET_MODE_SIZE.
29902         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
29903         wider than SImode without modification.
29904         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
29905         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
29906         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
29907         passing and returning SVE modes.
29908         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
29909         rather than GEN_INT.
29910         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
29911         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
29912         (aarch64_allocate_and_probe_stack_space): Likewise.
29913         (aarch64_layout_frame): Cope with polynomial offsets.
29914         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
29915         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
29916         polynomial offsets.
29917         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
29918         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
29919         poly_int64 rather than a HOST_WIDE_INT.
29920         (aarch64_get_separate_components, aarch64_process_components)
29921         (aarch64_expand_prologue, aarch64_expand_epilogue)
29922         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
29923         (aarch64_anchor_offset): New function, split out from...
29924         (aarch64_legitimize_address): ...here.
29925         (aarch64_builtin_vectorization_cost): Handle polynomial
29926         TYPE_VECTOR_SUBPARTS.
29927         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
29928         GET_MODE_NUNITS.
29929         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
29930         number of elements from the PARALLEL rather than the mode.
29931         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
29932         rather than GET_MODE_BITSIZE.
29933         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
29934         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
29935         (aarch64_expand_vec_perm_const_1): Handle polynomial
29936         d->perm.length () and d->perm elements.
29937         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
29938         Apply to_constant to d->perm elements.
29939         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
29940         polynomial CONST_VECTOR_NUNITS.
29941         (aarch64_move_pointer): Take amount as a poly_int64 rather
29942         than an int.
29943         (aarch64_progress_pointer): Avoid temporary variable.
29944         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
29945         the mode attribute instead of GET_MODE.
29947 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29948             Alan Hayward  <alan.hayward@arm.com>
29949             David Sherwood  <david.sherwood@arm.com>
29951         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
29952         x exists before using it.
29953         (aarch64_add_constant_internal): Rename to...
29954         (aarch64_add_offset_1): ...this.  Replace regnum with separate
29955         src and dest rtxes.  Handle the case in which they're different,
29956         including when the offset is zero.  Replace scratchreg with an rtx.
29957         Use 2 additions if there is no spare register into which we can
29958         move a 16-bit constant.
29959         (aarch64_add_constant): Delete.
29960         (aarch64_add_offset): Replace reg with separate src and dest
29961         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
29962         Use aarch64_add_offset_1.
29963         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
29964         an rtx rather than an int.  Take the delta as a poly_int64
29965         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
29966         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
29967         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
29968         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
29969         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
29970         and aarch64_add_sp.
29971         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
29972         aarch64_add_constant.
29974 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29976         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
29977         Use scalar_float_mode.
29979 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
29981         * config/aarch64/aarch64-simd.md
29982         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
29983         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
29984         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
29985         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
29986         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
29987         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
29988         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
29989         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
29990         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
29991         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
29993 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
29995         PR target/83514
29996         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
29997         targ_options->x_arm_arch_string is non NULL.
29999 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
30001         * config/aarch64/aarch64.h
30002         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
30004 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
30006         PR target/82096
30007         * expmed.c (emit_store_flag_force): Swap if const op0
30008         and change VOIDmode to mode of op0.
30010 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
30012         PR rtl-optimization/83761
30013         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
30014         than bytes to mode_for_size.
30016 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30018         PR middle-end/83189
30019         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
30020         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
30021         profile.
30023 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30025         PR middle-end/83575
30026         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
30027         when in layout mode.
30028         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
30029         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
30030         partition fixup.
30032 2018-01-10  Michael Collison  <michael.collison@arm.com>
30034         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
30035         * config/aarch64/aarch64-option-extension.def: Add
30036         AARCH64_OPT_EXTENSION of 'fp16fml'.
30037         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30038         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
30039         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
30040         * config/aarch64/constraints.md (Ui7): New constraint.
30041         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
30042         (VFMLA_SEL_W): Ditto.
30043         (f16quad): Ditto.
30044         (f16mac1): Ditto.
30045         (VFMLA16_LOW): New int iterator.
30046         (VFMLA16_HIGH): Ditto.
30047         (UNSPEC_FMLAL): New unspec.
30048         (UNSPEC_FMLSL): Ditto.
30049         (UNSPEC_FMLAL2): Ditto.
30050         (UNSPEC_FMLSL2): Ditto.
30051         (f16mac): New code attribute.
30052         * config/aarch64/aarch64-simd-builtins.def
30053         (aarch64_fmlal_lowv2sf): Ditto.
30054         (aarch64_fmlsl_lowv2sf): Ditto.
30055         (aarch64_fmlalq_lowv4sf): Ditto.
30056         (aarch64_fmlslq_lowv4sf): Ditto.
30057         (aarch64_fmlal_highv2sf): Ditto.
30058         (aarch64_fmlsl_highv2sf): Ditto.
30059         (aarch64_fmlalq_highv4sf): Ditto.
30060         (aarch64_fmlslq_highv4sf): Ditto.
30061         (aarch64_fmlal_lane_lowv2sf): Ditto.
30062         (aarch64_fmlsl_lane_lowv2sf): Ditto.
30063         (aarch64_fmlal_laneq_lowv2sf): Ditto.
30064         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
30065         (aarch64_fmlalq_lane_lowv4sf): Ditto.
30066         (aarch64_fmlsl_lane_lowv4sf): Ditto.
30067         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
30068         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
30069         (aarch64_fmlal_lane_highv2sf): Ditto.
30070         (aarch64_fmlsl_lane_highv2sf): Ditto.
30071         (aarch64_fmlal_laneq_highv2sf): Ditto.
30072         (aarch64_fmlsl_laneq_highv2sf): Ditto.
30073         (aarch64_fmlalq_lane_highv4sf): Ditto.
30074         (aarch64_fmlsl_lane_highv4sf): Ditto.
30075         (aarch64_fmlalq_laneq_highv4sf): Ditto.
30076         (aarch64_fmlsl_laneq_highv4sf): Ditto.
30077         * config/aarch64/aarch64-simd.md:
30078         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
30079         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
30080         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
30081         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
30082         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
30083         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
30084         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
30085         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
30086         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
30087         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
30088         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
30089         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
30090         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
30091         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
30092         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
30093         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
30094         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
30095         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
30096         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
30097         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
30098         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
30099         (vfmlsl_low_u32): Ditto.
30100         (vfmlalq_low_u32): Ditto.
30101         (vfmlslq_low_u32): Ditto.
30102         (vfmlal_high_u32): Ditto.
30103         (vfmlsl_high_u32): Ditto.
30104         (vfmlalq_high_u32): Ditto.
30105         (vfmlslq_high_u32): Ditto.
30106         (vfmlal_lane_low_u32): Ditto.
30107         (vfmlsl_lane_low_u32): Ditto.
30108         (vfmlal_laneq_low_u32): Ditto.
30109         (vfmlsl_laneq_low_u32): Ditto.
30110         (vfmlalq_lane_low_u32): Ditto.
30111         (vfmlslq_lane_low_u32): Ditto.
30112         (vfmlalq_laneq_low_u32): Ditto.
30113         (vfmlslq_laneq_low_u32): Ditto.
30114         (vfmlal_lane_high_u32): Ditto.
30115         (vfmlsl_lane_high_u32): Ditto.
30116         (vfmlal_laneq_high_u32): Ditto.
30117         (vfmlsl_laneq_high_u32): Ditto.
30118         (vfmlalq_lane_high_u32): Ditto.
30119         (vfmlslq_lane_high_u32): Ditto.
30120         (vfmlalq_laneq_high_u32): Ditto.
30121         (vfmlslq_laneq_high_u32): Ditto.
30122         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
30123         (AARCH64_FL_FOR_ARCH8_4): New.
30124         (AARCH64_ISA_F16FML): New ISA flag.
30125         (TARGET_F16FML): New feature flag for fp16fml.
30126         (doc/invoke.texi): Document new fp16fml option.
30128 2018-01-10  Michael Collison  <michael.collison@arm.com>
30130         * config/aarch64/aarch64-builtins.c:
30131         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
30132         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30133         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
30134         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
30135         (AARCH64_ISA_SHA3): New ISA flag.
30136         (TARGET_SHA3): New feature flag for sha3.
30137         * config/aarch64/iterators.md (sha512_op): New int attribute.
30138         (CRYPTO_SHA512): New int iterator.
30139         (UNSPEC_SHA512H): New unspec.
30140         (UNSPEC_SHA512H2): Ditto.
30141         (UNSPEC_SHA512SU0): Ditto.
30142         (UNSPEC_SHA512SU1): Ditto.
30143         * config/aarch64/aarch64-simd-builtins.def
30144         (aarch64_crypto_sha512hqv2di): New builtin.
30145         (aarch64_crypto_sha512h2qv2di): Ditto.
30146         (aarch64_crypto_sha512su0qv2di): Ditto.
30147         (aarch64_crypto_sha512su1qv2di): Ditto.
30148         (aarch64_eor3qv8hi): Ditto.
30149         (aarch64_rax1qv2di): Ditto.
30150         (aarch64_xarqv2di): Ditto.
30151         (aarch64_bcaxqv8hi): Ditto.
30152         * config/aarch64/aarch64-simd.md:
30153         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
30154         (aarch64_crypto_sha512su0qv2di): Ditto.
30155         (aarch64_crypto_sha512su1qv2di): Ditto.
30156         (aarch64_eor3qv8hi): Ditto.
30157         (aarch64_rax1qv2di): Ditto.
30158         (aarch64_xarqv2di): Ditto.
30159         (aarch64_bcaxqv8hi): Ditto.
30160         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
30161         (vsha512h2q_u64): Ditto.
30162         (vsha512su0q_u64): Ditto.
30163         (vsha512su1q_u64): Ditto.
30164         (veor3q_u16): Ditto.
30165         (vrax1q_u64): Ditto.
30166         (vxarq_u64): Ditto.
30167         (vbcaxq_u16): Ditto.
30168         * config/arm/types.md (crypto_sha512): New type attribute.
30169         (crypto_sha3): Ditto.
30170         (doc/invoke.texi): Document new sha3 option.
30172 2018-01-10  Michael Collison  <michael.collison@arm.com>
30174         * config/aarch64/aarch64-builtins.c:
30175         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
30176         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30177         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
30178         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
30179         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
30180         (AARCH64_ISA_SM4): New ISA flag.
30181         (TARGET_SM4): New feature flag for sm4.
30182         * config/aarch64/aarch64-simd-builtins.def
30183         (aarch64_sm3ss1qv4si): Ditto.
30184         (aarch64_sm3tt1aq4si): Ditto.
30185         (aarch64_sm3tt1bq4si): Ditto.
30186         (aarch64_sm3tt2aq4si): Ditto.
30187         (aarch64_sm3tt2bq4si): Ditto.
30188         (aarch64_sm3partw1qv4si): Ditto.
30189         (aarch64_sm3partw2qv4si): Ditto.
30190         (aarch64_sm4eqv4si): Ditto.
30191         (aarch64_sm4ekeyqv4si): Ditto.
30192         * config/aarch64/aarch64-simd.md:
30193         (aarch64_sm3ss1qv4si): Ditto.
30194         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
30195         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
30196         (aarch64_sm4eqv4si): Ditto.
30197         (aarch64_sm4ekeyqv4si): Ditto.
30198         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
30199         (sm3part_op): Ditto.
30200         (CRYPTO_SM3TT): Ditto.
30201         (CRYPTO_SM3PART): Ditto.
30202         (UNSPEC_SM3SS1): New unspec.
30203         (UNSPEC_SM3TT1A): Ditto.
30204         (UNSPEC_SM3TT1B): Ditto.
30205         (UNSPEC_SM3TT2A): Ditto.
30206         (UNSPEC_SM3TT2B): Ditto.
30207         (UNSPEC_SM3PARTW1): Ditto.
30208         (UNSPEC_SM3PARTW2): Ditto.
30209         (UNSPEC_SM4E): Ditto.
30210         (UNSPEC_SM4EKEY): Ditto.
30211         * config/aarch64/constraints.md (Ui2): New constraint.
30212         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
30213         * config/arm/types.md (crypto_sm3): New type attribute.
30214         (crypto_sm4): Ditto.
30215         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
30216         (vsm3tt1aq_u32): Ditto.
30217         (vsm3tt1bq_u32): Ditto.
30218         (vsm3tt2aq_u32): Ditto.
30219         (vsm3tt2bq_u32): Ditto.
30220         (vsm3partw1q_u32): Ditto.
30221         (vsm3partw2q_u32): Ditto.
30222         (vsm4eq_u32): Ditto.
30223         (vsm4ekeyq_u32): Ditto.
30224         (doc/invoke.texi): Document new sm4 option.
30226 2018-01-10  Michael Collison  <michael.collison@arm.com>
30228         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
30229         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
30230         (AARCH64_FL_FOR_ARCH8_4): New.
30231         (AARCH64_FL_V8_4): New flag.
30232         (doc/invoke.texi): Document new armv8.4-a option.
30234 2018-01-10  Michael Collison  <michael.collison@arm.com>
30236         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
30237         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
30238         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
30239         * config/aarch64/aarch64-option-extension.def: Add
30240         AARCH64_OPT_EXTENSION of 'sha2'.
30241         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
30242         (crypto): Disable sha2 and aes if crypto disabled.
30243         (crypto): Enable aes and sha2 if enabled.
30244         (simd): Disable sha2 and aes if simd disabled.
30245         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
30246         New flags.
30247         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
30248         (TARGET_SHA2): New feature flag for sha2.
30249         (TARGET_AES): New feature flag for aes.
30250         * config/aarch64/aarch64-simd.md:
30251         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
30252         conditional on TARGET_AES.
30253         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
30254         (aarch64_crypto_sha1hsi): Make pattern conditional
30255         on TARGET_SHA2.
30256         (aarch64_crypto_sha1hv4si): Ditto.
30257         (aarch64_be_crypto_sha1hv4si): Ditto.
30258         (aarch64_crypto_sha1su1v4si): Ditto.
30259         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
30260         (aarch64_crypto_sha1su0v4si): Ditto.
30261         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
30262         (aarch64_crypto_sha256su0v4si): Ditto.
30263         (aarch64_crypto_sha256su1v4si): Ditto.
30264         (doc/invoke.texi): Document new aes and sha2 options.
30266 2018-01-10  Martin Sebor  <msebor@redhat.com>
30268         PR tree-optimization/83781
30269         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
30270         as string arrays.
30272 2018-01-11  Martin Sebor  <msebor@gmail.com>
30273             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
30275         PR tree-optimization/83501
30276         PR tree-optimization/81703
30278         * tree-ssa-strlen.c (get_string_cst): Rename...
30279         (get_string_len): ...to this.  Handle global constants.
30280         (handle_char_store): Adjust.
30282 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
30283             Jim Wilson  <jimw@sifive.com>
30285         * config/riscv/riscv-protos.h (riscv_output_return): New.
30286         * config/riscv/riscv.c (struct machine_function): New naked_p field.
30287         (riscv_attribute_table, riscv_output_return),
30288         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
30289         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
30290         (riscv_compute_frame_info): Only compute frame->mask if not a naked
30291         function.
30292         (riscv_expand_prologue): Add early return for naked function.
30293         (riscv_expand_epilogue): Likewise.
30294         (riscv_function_ok_for_sibcall): Return false for naked function.
30295         (riscv_set_current_function): New.
30296         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
30297         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
30298         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
30299         * doc/extend.texi (RISC-V Function Attributes): New.
30301 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
30303         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
30304         check for 128-bit long double before checking TCmode.
30305         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
30306         128-bit long doubles before checking TFmode or TCmode.
30307         (FLOAT128_IBM_P): Likewise.
30309 2018-01-10  Martin Sebor  <msebor@redhat.com>
30311         PR tree-optimization/83671
30312         * builtins.c (c_strlen): Unconditionally return zero for the empty
30313         string.
30314         Use -Warray-bounds for warnings.
30315         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
30316         for non-constant array indices with COMPONENT_REF, arrays of
30317         arrays, and pointers to arrays.
30318         (gimple_fold_builtin_strlen): Determine and set length range for
30319         non-constant character arrays.
30321 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
30323         PR middle-end/81897
30324         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
30325         empty blocks.
30327 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
30329         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
30331 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
30333         PR target/83399
30334         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
30335         VECTOR_MEM_ALTIVEC_OR_VSX_P.
30336         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
30337         indexed_or_indirect_operand predicate.
30338         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
30339         (*vsx_le_perm_load_v8hi): Likewise.
30340         (*vsx_le_perm_load_v16qi): Likewise.
30341         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
30342         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
30343         (*vsx_le_perm_store_v8hi): Likewise.
30344         (*vsx_le_perm_store_v16qi): Likewise.
30345         (eight unnamed splitters): Likewise.
30347 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
30349         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
30350         * config/rs6000/emmintrin.h: Likewise.
30351         * config/rs6000/mmintrin.h: Likewise.
30352         * config/rs6000/xmmintrin.h: Likewise.
30354 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
30356         PR c++/43486
30357         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
30358         "public_flag".
30359         * tree.c (tree_nop_conversion): Return true for location wrapper
30360         nodes.
30361         (maybe_wrap_with_location): New function.
30362         (selftest::check_strip_nops): New function.
30363         (selftest::test_location_wrappers): New function.
30364         (selftest::tree_c_tests): Call it.
30365         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
30366         (maybe_wrap_with_location): New decl.
30367         (EXPR_LOCATION_WRAPPER_P): New macro.
30368         (location_wrapper_p): New inline function.
30369         (tree_strip_any_location_wrapper): New inline function.
30371 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
30373         PR target/83735
30374         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
30375         stack_realign_offset for the largest alignment of stack slot
30376         actually used.
30377         (ix86_find_max_used_stack_alignment): New function.
30378         (ix86_finalize_stack_frame_flags): Use it.  Set
30379         max_used_stack_alignment if we don't realign stack.
30380         * config/i386/i386.h (machine_function): Add
30381         max_used_stack_alignment.
30383 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
30385         * config/arm/arm.opt (-mbranch-cost): New option.
30386         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
30387         account.
30389 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
30391         PR target/83629
30392         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
30393         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
30395 2018-01-10  Richard Biener  <rguenther@suse.de>
30397         PR debug/83765
30398         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
30399         early out so it also covers the case where we have a non-NULL
30400         origin.
30402 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
30404         PR tree-optimization/83753
30405         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
30406         for non-strided grouped accesses if the number of elements is 1.
30408 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
30410         PR target/81616
30411         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
30412         * i386.h (TARGET_USE_GATHER): Define.
30413         * x86-tune.def (X86_TUNE_USE_GATHER): New.
30415 2018-01-10  Martin Liska  <mliska@suse.cz>
30417         PR bootstrap/82831
30418         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
30419         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
30420         partitioning.
30421         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
30422         CLEANUP_NO_PARTITIONING is not set.
30424 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
30426         * doc/rtl.texi: Remove documentation of (const ...) wrappers
30427         for vectors, as a partial revert of r254296.
30428         * rtl.h (const_vec_p): Delete.
30429         (const_vec_duplicate_p): Don't test for vector CONSTs.
30430         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
30431         * expmed.c (make_tree): Likewise.
30433         Revert:
30434         * common.md (E, F): Use CONSTANT_P instead of checking for
30435         CONST_VECTOR.
30436         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
30437         checking for CONST_VECTOR.
30439 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30441         PR middle-end/83575
30442         * predict.c (force_edge_cold): Handle in more sane way edges
30443         with no prediction.
30445 2018-01-09  Carl Love  <cel@us.ibm.com>
30447         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
30448         V4SI, V4SF types.
30449         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
30450         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
30451         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
30452         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
30453         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
30454         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
30455         * config/rs6000/rs6000-protos.h: Add extern defition for
30456         rs6000_generate_float2_double_code.
30457         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
30458         function.
30459         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
30460         (float2_v2df): Add define_expand.
30462 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
30464         PR target/83628
30465         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
30466         op_mode in the force_to_mode call.
30468 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30470         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
30471         instead of checking each element individually.
30472         (aarch64_evpc_uzp): Likewise.
30473         (aarch64_evpc_zip): Likewise.
30474         (aarch64_evpc_ext): Likewise.
30475         (aarch64_evpc_rev): Likewise.
30476         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
30477         instead of checking each element individually.  Return true without
30478         generating rtl if
30479         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
30480         whether all selected elements come from the same input, instead of
30481         checking each element individually.  Remove calls to gen_rtx_REG,
30482         start_sequence and end_sequence and instead assert that no rtl is
30483         generated.
30485 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30487         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
30488         order of HIGH and CONST checks.
30490 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
30492         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
30493         if the destination isn't an SSA_NAME.
30495 2018-01-09  Richard Biener  <rguenther@suse.de>
30497         PR tree-optimization/83668
30498         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
30499         move prologue...
30500         (canonicalize_loop_form): ... here, renamed from ...
30501         (canonicalize_loop_closed_ssa_form): ... this and amended to
30502         swap successor edges for loop exit blocks to make us use
30503         the RPO order we need for initial schedule generation.
30505 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
30507         PR tree-optimization/64811
30508         * match.pd: When optimizing comparisons with Inf, avoid
30509         introducing or losing exceptions from comparisons with NaN.
30511 2018-01-09  Martin Liska  <mliska@suse.cz>
30513         PR sanitizer/82517
30514         * asan.c (shadow_mem_size): Add gcc_assert.
30516 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
30518         Don't save registers in main().
30520         PR target/83738
30521         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
30522         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
30523         * config/avr/avr.c (avr_set_current_function): Don't error if
30524         naked, OS_task or OS_main are specified at the same time.
30525         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
30526         OS_main.
30527         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
30528         attribute.
30529         * common/config/avr/avr-common.c (avr_option_optimization_table):
30530         Switch on -mmain-is-OS_task for optimizing compilations.
30532 2018-01-09  Richard Biener  <rguenther@suse.de>
30534         PR tree-optimization/83572
30535         * graphite.c: Include cfganal.h.
30536         (graphite_transform_loops): Connect infinite loops to exit
30537         and remove fake edges at the end.
30539 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30541         * ipa-inline.c (edge_badness): Revert accidental checkin.
30543 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
30545         PR ipa/80763
30546         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
30547         symbols; not inline clones.
30549 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
30551         PR target/83507
30552         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
30553         hard registers.  Formatting fixes.
30555         PR preprocessor/83722
30556         * gcc.c (try_generate_repro): Pass
30557         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
30558         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
30559         do_report_bug.
30561 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
30562             Kito Cheng  <kito.cheng@gmail.com>
30564         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
30565         (riscv_leaf_function_p): Delete.
30566         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
30568 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30570         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
30571         function.
30572         (do_ifelse): New function.
30573         (do_isel): New function.
30574         (do_sub3): New function.
30575         (do_add3): New function.
30576         (do_load_mask_compare): New function.
30577         (do_overlap_load_compare): New function.
30578         (expand_compare_loop): New function.
30579         (expand_block_compare): Call expand_compare_loop() when appropriate.
30580         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
30581         option description.
30582         (-mblock-compare-inline-loop-limit): New option.
30584 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
30586         PR target/83677
30587         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
30588         Reverse order of second and third operands in first alternative.
30589         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
30590         of first and second elements in UNSPEC_VPERMR vector.
30591         (altivec_expand_vec_perm_le): Likewise.
30593 2018-01-08  Jeff Law  <law@redhat.com>
30595         PR rtl-optimizatin/81308
30596         * tree-switch-conversion.c (cfg_altered): New file scoped static.
30597         (process_switch): If group_case_labels makes a change, then set
30598         cfg_altered.
30599         (pass_convert_switch::execute): If a switch is converted, then
30600         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
30602         PR rtl-optimization/81308
30603         * recog.c (split_all_insns): Conditionally cleanup the CFG after
30604         splitting insns.
30606 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
30608         PR target/83663 - Revert r255946
30609         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
30610         generation for cases where splatting a value is not useful.
30611         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
30612         across a vec_duplicate and a paradoxical subreg forming a vector
30613         mode to a vec_concat.
30615 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
30617         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
30618         -march=armv8.3-a variants.
30619         * config/arm/t-multilib: Likewise.
30620         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
30622 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
30624         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
30625         to generate rtl.
30626         (cceq_ior_compare_complement): Give it a name so I can use it, and
30627         change boolean_or_operator predicate to boolean_operator so it can
30628         be used to generate a crand.
30629         (eqne): New code iterator.
30630         (bd/bd_neg): New code_attrs.
30631         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
30632         a single define_insn.
30633         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
30634         decrement (bdnzt/bdnzf/bdzt/bdzf).
30635         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
30636         with the new names of the branch decrement patterns, and added the
30637         names of the branch decrement conditional patterns.
30639 2018-01-08  Richard Biener  <rguenther@suse.de>
30641         PR tree-optimization/83563
30642         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
30643         cache.
30645 2018-01-08  Richard Biener  <rguenther@suse.de>
30647         PR middle-end/83713
30648         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
30650 2018-01-08  Richard Biener  <rguenther@suse.de>
30652         PR tree-optimization/83685
30653         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
30654         references to abnormals.
30656 2018-01-08  Richard Biener  <rguenther@suse.de>
30658         PR lto/83719
30659         * dwarf2out.c (output_indirect_strings): Handle empty
30660         skeleton_debug_str_hash.
30661         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
30663 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30665         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
30666         (emit_store_direct): Likewise.
30667         (arc_trampoline_adjust_address): Likewise.
30668         (arc_asm_trampoline_template): New function.
30669         (arc_initialize_trampoline): Use asm_trampoline_template.
30670         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
30671         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
30672         * config/arc/arc.md (flush_icache): Delete pattern.
30674 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
30676         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
30677         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
30678         munaligned-access.
30680 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30682         PR target/83681
30683         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
30684         by not USED_FOR_TARGET.
30685         (make_pass_resolve_sw_modes): Likewise.
30687 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30689         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
30690         USED_FOR_TARGET.
30692 2018-01-08  Richard Biener  <rguenther@suse.de>
30694         PR middle-end/83580
30695         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
30697 2018-01-08  Richard Biener  <rguenther@suse.de>
30699         PR middle-end/83517
30700         * match.pd ((t * 2) / 2) -> t): Add missing :c.
30702 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
30704         PR middle-end/81897
30705         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
30706         basic blocks with a small number of successors.
30707         (convert_control_dep_chain_into_preds): Improve handling of
30708         forwarder blocks.
30709         (dump_predicates): Split apart into...
30710         (dump_pred_chain): ...here...
30711         (dump_pred_info): ...and here.
30712         (can_one_predicate_be_invalidated_p): Add debugging printfs.
30713         (can_chain_union_be_invalidated_p): Improve check for invalidation
30714         of paths.
30715         (uninit_uses_cannot_happen): Avoid unnecessary if
30716         convert_control_dep_chain_into_preds yielded nothing.
30718 2018-01-06  Martin Sebor  <msebor@redhat.com>
30720         PR tree-optimization/83640
30721         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
30722         subtracting negative offset from size.
30723         (builtin_access::overlap): Adjust offset bounds of the access to fall
30724         within the size of the object if possible.
30726 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
30728         PR rtl-optimization/83699
30729         * expmed.c (extract_bit_field_1): Restrict the vector usage of
30730         extract_bit_field_as_subreg to cases in which the extracted
30731         value is also a vector.
30733         * lra-constraints.c (process_alt_operands): Test for the equivalence
30734         substitutions when detecting a possible reload cycle.
30736 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
30738         PR debug/83480
30739         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
30740         by default if flag_selective_schedling{,2}.  Formatting fixes.
30742         PR rtl-optimization/83682
30743         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
30744         if it has non-VECTOR_MODE element mode.
30745         (vec_duplicate_p): Likewise.
30747         PR middle-end/83694
30748         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
30749         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
30751 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30753         PR target/83604
30754         * config/i386/i386-builtin.def
30755         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
30756         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
30757         Require also OPTION_MASK_ISA_AVX512F in addition to
30758         OPTION_MASK_ISA_GFNI.
30759         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
30760         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
30761         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
30762         to OPTION_MASK_ISA_GFNI.
30763         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
30764         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
30765         OPTION_MASK_ISA_AVX512BW.
30766         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
30767         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
30768         addition to OPTION_MASK_ISA_GFNI.
30769         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
30770         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
30771         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
30772         to OPTION_MASK_ISA_GFNI.
30773         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
30774         a requirement for all ISAs rather than any of them with a few
30775         exceptions.
30776         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
30777         processing.
30778         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
30779         bitmasks to be enabled with 3 exceptions, instead of requiring any
30780         enabled ISA with lots of exceptions.
30781         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
30782         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
30783         Change avx512bw in isa attribute to avx512f.
30784         * config/i386/sgxintrin.h: Add license boilerplate.
30785         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
30786         to __AVX512F__ and __AVX512VL to __AVX512VL__.
30787         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
30788         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
30789         defined.
30790         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
30791         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
30792         temporarily sse2 rather than sse if not enabled already.
30794         PR target/83604
30795         * config/i386/sse.md (VI248_VLBW): Rename to ...
30796         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
30797         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
30798         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
30799         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
30800         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
30801         mode iterator instead of VI248_VLBW.
30803 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
30805         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
30806         (record_modified): Skip clobbers; add debug output.
30807         (param_change_prob): Use sreal frequencies.
30809 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30811         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
30812         punt for user-aligned variables.
30814 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
30816         * tree-chrec.c (chrec_contains_symbols): Return true for
30817         POLY_INT_CST.
30819 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
30821         PR target/82439
30822         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
30823         of (x|y) == x for BICS pattern.
30825 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
30827         PR tree-optimization/83605
30828         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
30829         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
30830         can throw.
30832 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
30834         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
30835         * config/epiphany/rtems.h: New file.
30837 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30838             Uros Bizjak  <ubizjak@gmail.com>
30840         PR target/83554
30841         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
30842         QIreg_operand instead of register_operand predicate.
30843         * config/i386/i386.c (ix86_rop_should_change_byte_p,
30844         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
30845         comments instead of -fmitigate[-_]rop.
30847 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
30849         PR bootstrap/81926
30850         * cgraphunit.c (symbol_table::compile): Switch to text_section
30851         before calling assembly_start debug hook.
30852         * run-rtl-passes.c (run_rtl_passes): Likewise.
30853         Include output.h.
30855 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30857         * tree-vrp.c (extract_range_from_binary_expr_1): Check
30858         range_int_cst_p rather than !symbolic_range_p before calling
30859         extract_range_from_multiplicative_op_1.
30861 2018-01-04  Jeff Law  <law@redhat.com>
30863         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
30864         redundant test in assertion.
30866 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30868         * doc/rtl.texi: Document machine_mode wrapper classes.
30870 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30872         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
30873         using tree_to_uhwi.
30875 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30877         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
30878         the VEC_PERM_EXPR fold to fail.
30880 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30882         PR debug/83585
30883         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
30884         to switched_sections.
30886 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30888         PR target/83680
30889         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
30890         test for d.testing.
30892 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
30894         PR target/83387
30895         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
30896         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
30898 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30900         PR debug/83666
30901         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
30902         is BLKmode and bitpos not zero or mode change is needed.
30904 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
30906         PR target/83675
30907         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
30908         TARGET_VIS2.
30910 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
30912         PR target/83628
30913         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
30914         instead of MULT rtx.  Update all corresponding splitters.
30915         (*saddl_se): Ditto.
30916         (*ssub<modesuffix>): Ditto.
30917         (*ssubl_se): Ditto.
30918         (*cmp_sadd_di): Update split patterns.
30919         (*cmp_sadd_si): Ditto.
30920         (*cmp_sadd_sidi): Ditto.
30921         (*cmp_ssub_di): Ditto.
30922         (*cmp_ssub_si): Ditto.
30923         (*cmp_ssub_sidi): Ditto.
30924         * config/alpha/predicates.md (const23_operand): New predicate.
30925         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
30926         Look for ASHIFT, not MULT inner operand.
30927         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
30929 2018-01-04  Martin Liska  <mliska@suse.cz>
30931         PR gcov-profile/83669
30932         * gcov.c (output_intermediate_file): Add version to intermediate
30933         gcov file.
30934         * doc/gcov.texi: Document new field 'version' in intermediate
30935         file format. Fix location of '-k' option of gcov command.
30937 2018-01-04  Martin Liska  <mliska@suse.cz>
30939         PR ipa/82352
30940         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
30942 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
30944         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
30946 2018-01-03  Martin Sebor  <msebor@redhat.com>
30948         PR tree-optimization/83655
30949         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
30950         checking calls with invalid arguments.
30952 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30954         * tree-vect-stmts.c (vect_get_store_rhs): New function.
30955         (vectorizable_mask_load_store): Delete.
30956         (vectorizable_call): Return false for masked loads and stores.
30957         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
30958         instead of gimple_assign_rhs1.
30959         (vectorizable_load): Handle IFN_MASK_LOAD.
30960         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
30962 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30964         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
30965         split out from..,
30966         (vectorizable_mask_load_store): ...here.
30967         (vectorizable_load): ...and here.
30969 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30971         * tree-vect-stmts.c (vect_build_all_ones_mask)
30972         (vect_build_zero_merge_argument): New functions, split out from...
30973         (vectorizable_load): ...here.
30975 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30977         * tree-vect-stmts.c (vect_check_store_rhs): New function,
30978         split out from...
30979         (vectorizable_mask_load_store): ...here.
30980         (vectorizable_store): ...and here.
30982 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30984         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
30985         split out from...
30986         (vectorizable_mask_load_store): ...here.
30988 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30990         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
30991         (vect_model_store_cost): Take a vec_load_store_type instead of a
30992         vect_def_type.
30993         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
30994         (vect_model_store_cost): Take a vec_load_store_type instead of a
30995         vect_def_type.
30996         (vectorizable_mask_load_store): Update accordingly.
30997         (vectorizable_store): Likewise.
30998         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
31000 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31002         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
31003         IFN_MASK_LOAD calls here rather than...
31004         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
31006 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31007             Alan Hayward  <alan.hayward@arm.com>
31008             David Sherwood  <david.sherwood@arm.com>
31010         * expmed.c (extract_bit_field_1): For vector extracts,
31011         fall back to extract_bit_field_as_subreg if vec_extract
31012         isn't available.
31014 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31015             Alan Hayward  <alan.hayward@arm.com>
31016             David Sherwood  <david.sherwood@arm.com>
31018         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
31019         they are variable or constant sized.
31020         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
31021         slots for constant-sized data.
31023 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31024             Alan Hayward  <alan.hayward@arm.com>
31025             David Sherwood  <david.sherwood@arm.com>
31027         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
31028         handling COND_EXPRs with boolean comparisons, try to find a better
31029         basis for the mask type than the boolean itself.
31031 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31033         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
31034         is calculated and how it can be overridden.
31035         * genmodes.c (max_bitsize_mode_any_mode): New variable.
31036         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
31037         if defined.
31038         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
31039         if nonzero.
31041 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31042             Alan Hayward  <alan.hayward@arm.com>
31043             David Sherwood  <david.sherwood@arm.com>
31045         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
31046         Remove the mode argument.
31047         (aarch64_simd_valid_immediate): Remove the mode and inverse
31048         arguments.
31049         * config/aarch64/iterators.md (bitsize): New iterator.
31050         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
31051         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
31052         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
31053         aarch64_simd_valid_immediate.
31054         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
31055         (aarch64_reg_or_bic_imm): Likewise.
31056         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
31057         with an insn_type enum and msl with a modifier_type enum.
31058         Replace element_width with a scalar_mode.  Change the shift
31059         to unsigned int.  Add constructors for scalar_float_mode and
31060         scalar_int_mode elements.
31061         (aarch64_vect_float_const_representable_p): Delete.
31062         (aarch64_can_const_movi_rtx_p)
31063         (aarch64_simd_scalar_immediate_valid_for_move)
31064         (aarch64_simd_make_constant): Update call to
31065         aarch64_simd_valid_immediate.
31066         (aarch64_advsimd_valid_immediate_hs): New function.
31067         (aarch64_advsimd_valid_immediate): Likewise.
31068         (aarch64_simd_valid_immediate): Remove mode and inverse
31069         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
31070         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
31071         and aarch64_float_const_representable_p on the result.
31072         (aarch64_output_simd_mov_immediate): Remove mode argument.
31073         Update call to aarch64_simd_valid_immediate and use of
31074         simd_immediate_info.
31075         (aarch64_output_scalar_simd_mov_immediate): Update call
31076         accordingly.
31078 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31079             Alan Hayward  <alan.hayward@arm.com>
31080             David Sherwood  <david.sherwood@arm.com>
31082         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
31083         (mode_nunits): Likewise CONST_MODE_NUNITS.
31084         * machmode.def (ADJUST_NUNITS): Document.
31085         * genmodes.c (mode_data::need_nunits_adj): New field.
31086         (blank_mode): Update accordingly.
31087         (adj_nunits): New variable.
31088         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
31089         parameter.
31090         (emit_mode_size_inline): Set need_bytesize_adj for all modes
31091         listed in adj_nunits.
31092         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
31093         listed in adj_nunits.  Don't emit case statements for such modes.
31094         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
31095         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
31096         nothing if adj_nunits is nonnull.
31097         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
31098         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
31099         (emit_mode_fbit): Update use of print_maybe_const_decl.
31100         (emit_move_size): Likewise.  Treat the array as non-const
31101         if adj_nunits.
31102         (emit_mode_adjustments): Handle adj_nunits.
31104 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31106         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
31107         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
31108         (VECTOR_MODES): Use it.
31109         (make_vector_modes): Take the prefix as an argument.
31111 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31112             Alan Hayward  <alan.hayward@arm.com>
31113             David Sherwood  <david.sherwood@arm.com>
31115         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
31116         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
31117         for MODE_VECTOR_BOOL.
31118         * machmode.def (VECTOR_BOOL_MODE): Document.
31119         * genmodes.c (VECTOR_BOOL_MODE): New macro.
31120         (make_vector_bool_mode): New function.
31121         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
31122         MODE_VECTOR_BOOL.
31123         * lto-streamer-in.c (lto_input_mode_table): Likewise.
31124         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
31125         Likewise.
31126         * stor-layout.c (int_mode_for_mode): Likewise.
31127         * tree.c (build_vector_type_for_mode): Likewise.
31128         * varasm.c (output_constant_pool_2): Likewise.
31129         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
31130         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
31131         for MODE_VECTOR_BOOL.
31132         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
31133         of mode class checks.
31134         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
31135         instead of a list of mode class checks.
31136         (expand_vector_scalar_condition): Likewise.
31137         (type_for_widest_vector_mode): Handle BImode as an inner mode.
31139 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31140             Alan Hayward  <alan.hayward@arm.com>
31141             David Sherwood  <david.sherwood@arm.com>
31143         * machmode.h (mode_size): Change from unsigned short to
31144         poly_uint16_pod.
31145         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
31146         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
31147         or if measurement_type is not polynomial.
31148         (fixed_size_mode::includes_p): Check for constant-sized modes.
31149         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
31150         return a poly_uint16 rather than an unsigned short.
31151         (emit_mode_size): Change the type of mode_size from unsigned short
31152         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
31153         (emit_mode_adjustments): Cope with polynomial vector sizes.
31154         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31155         for GET_MODE_SIZE.
31156         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31157         for GET_MODE_SIZE.
31158         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
31159         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
31160         * caller-save.c (setup_save_areas): Likewise.
31161         (replace_reg_with_saved_mem): Likewise.
31162         * calls.c (emit_library_call_value_1): Likewise.
31163         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
31164         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
31165         (gen_lowpart_for_combine): Likewise.
31166         * convert.c (convert_to_integer_1): Likewise.
31167         * cse.c (equiv_constant, cse_insn): Likewise.
31168         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
31169         (cselib_subst_to_values): Likewise.
31170         * dce.c (word_dce_process_block): Likewise.
31171         * df-problems.c (df_word_lr_mark_ref): Likewise.
31172         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
31173         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
31174         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
31175         (rtl_for_decl_location): Likewise.
31176         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
31177         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
31178         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
31179         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
31180         (expand_expr_real_1): Likewise.
31181         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
31182         (pad_below): Likewise.
31183         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31184         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
31185         * ira.c (get_subreg_tracking_sizes): Likewise.
31186         * ira-build.c (ira_create_allocno_objects): Likewise.
31187         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
31188         (ira_sort_regnos_for_alter_reg): Likewise.
31189         * ira-costs.c (record_operand_costs): Likewise.
31190         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
31191         (resolve_simple_move): Likewise.
31192         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
31193         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
31194         (lra_constraints): Likewise.
31195         (CONST_POOL_OK_P): Reject variable-sized modes.
31196         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
31197         (add_pseudo_to_slot, lra_spill): Likewise.
31198         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31199         * optabs-query.c (get_best_extraction_insn): Likewise.
31200         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31201         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
31202         (expand_mult_highpart, valid_multiword_target_p): Likewise.
31203         * recog.c (offsettable_address_addr_space_p): Likewise.
31204         * regcprop.c (maybe_mode_change): Likewise.
31205         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
31206         * regrename.c (build_def_use): Likewise.
31207         * regstat.c (dump_reg_info): Likewise.
31208         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
31209         (find_reloads, find_reloads_subreg_address): Likewise.
31210         * reload1.c (eliminate_regs_1): Likewise.
31211         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
31212         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
31213         (simplify_binary_operation_1, simplify_subreg): Likewise.
31214         * targhooks.c (default_function_arg_padding): Likewise.
31215         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
31216         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
31217         (verify_gimple_assign_ternary): Likewise.
31218         * tree-inline.c (estimate_move_cost): Likewise.
31219         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31220         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
31221         (get_address_cost_ainc): Likewise.
31222         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
31223         (vect_supportable_dr_alignment): Likewise.
31224         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31225         (vectorizable_reduction): Likewise.
31226         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
31227         (vectorizable_operation, vectorizable_load): Likewise.
31228         * tree.c (build_same_sized_truth_vector_type): Likewise.
31229         * valtrack.c (cleanup_auto_inc_dec): Likewise.
31230         * var-tracking.c (emit_note_insn_var_location): Likewise.
31231         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
31232         (ADDR_VEC_ALIGN): Likewise.
31234 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31235             Alan Hayward  <alan.hayward@arm.com>
31236             David Sherwood  <david.sherwood@arm.com>
31238         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
31239         unsigned short.
31240         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
31241         or if measurement_type is polynomial.
31242         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
31243         * combine.c (make_extraction): Likewise.
31244         * dse.c (find_shift_sequence): Likewise.
31245         * dwarf2out.c (mem_loc_descriptor): Likewise.
31246         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
31247         (extract_bit_field, extract_low_bits): Likewise.
31248         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
31249         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
31250         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
31251         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
31252         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
31253         * reload.c (find_reloads): Likewise.
31254         * reload1.c (alter_reg): Likewise.
31255         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
31256         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
31257         * tree-if-conv.c (predicate_mem_writes): Likewise.
31258         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
31259         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
31260         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
31261         * valtrack.c (dead_debug_insert_temp): Likewise.
31262         * varasm.c (mergeable_constant_section): Likewise.
31263         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
31265 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31266             Alan Hayward  <alan.hayward@arm.com>
31267             David Sherwood  <david.sherwood@arm.com>
31269         * expr.c (expand_assignment): Cope with polynomial mode sizes
31270         when assigning to a CONCAT.
31272 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31273             Alan Hayward  <alan.hayward@arm.com>
31274             David Sherwood  <david.sherwood@arm.com>
31276         * machmode.h (mode_precision): Change from unsigned short to
31277         poly_uint16_pod.
31278         (mode_to_precision): Return a poly_uint16 rather than an unsigned
31279         short.
31280         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
31281         or if measurement_type is not polynomial.
31282         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
31283         in which the mode is already known to be a scalar_int_mode.
31284         * genmodes.c (emit_mode_precision): Change the type of mode_precision
31285         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
31286         initializer.
31287         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31288         for GET_MODE_PRECISION.
31289         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31290         for GET_MODE_PRECISION.
31291         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
31292         as polynomial.
31293         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
31294         (expand_field_assignment, make_extraction): Likewise.
31295         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
31296         (get_last_value): Likewise.
31297         * convert.c (convert_to_integer_1): Likewise.
31298         * cse.c (cse_insn): Likewise.
31299         * expr.c (expand_expr_real_1): Likewise.
31300         * lra-constraints.c (simplify_operand_subreg): Likewise.
31301         * optabs-query.c (can_atomic_load_p): Likewise.
31302         * optabs.c (expand_atomic_load): Likewise.
31303         (expand_atomic_store): Likewise.
31304         * ree.c (combine_reaching_defs): Likewise.
31305         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
31306         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
31307         * tree.h (type_has_mode_precision_p): Likewise.
31308         * ubsan.c (instrument_si_overflow): Likewise.
31310 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31311             Alan Hayward  <alan.hayward@arm.com>
31312             David Sherwood  <david.sherwood@arm.com>
31314         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
31315         polynomial numbers of units.
31316         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
31317         (valid_vector_subparts_p): New function.
31318         (build_vector_type): Remove temporary shim and take the number
31319         of units as a poly_uint64 rather than an int.
31320         (build_opaque_vector_type): Take the number of units as a
31321         poly_uint64 rather than an int.
31322         * tree.c (build_vector_from_ctor): Handle polynomial
31323         TYPE_VECTOR_SUBPARTS.
31324         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
31325         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
31326         (build_vector_from_val): If the number of units is variable,
31327         use build_vec_duplicate_cst for constant operands and
31328         VEC_DUPLICATE_EXPR otherwise.
31329         (make_vector_type): Remove temporary is_constant ().
31330         (build_vector_type, build_opaque_vector_type): Take the number of
31331         units as a poly_uint64 rather than an int.
31332         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
31333         VECTOR_CST_NELTS.
31334         * cfgexpand.c (expand_debug_expr): Likewise.
31335         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
31336         (store_constructor, expand_expr_real_1): Likewise.
31337         (const_scalar_mask_from_tree): Likewise.
31338         * fold-const-call.c (fold_const_reduction): Likewise.
31339         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
31340         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
31341         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
31342         (fold_relational_const): Likewise.
31343         (native_interpret_vector): Likewise.  Change the size from an
31344         int to an unsigned int.
31345         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
31346         TYPE_VECTOR_SUBPARTS.
31347         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
31348         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
31349         duplicating a non-constant operand into a variable-length vector.
31350         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
31351         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
31352         * ipa-icf.c (sem_variable::equals): Likewise.
31353         * match.pd: Likewise.
31354         * omp-simd-clone.c (simd_clone_subparts): Likewise.
31355         * print-tree.c (print_node): Likewise.
31356         * stor-layout.c (layout_type): Likewise.
31357         * targhooks.c (default_builtin_vectorization_cost): Likewise.
31358         * tree-cfg.c (verify_gimple_comparison): Likewise.
31359         (verify_gimple_assign_binary): Likewise.
31360         (verify_gimple_assign_ternary): Likewise.
31361         (verify_gimple_assign_single): Likewise.
31362         * tree-pretty-print.c (dump_generic_node): Likewise.
31363         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
31364         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
31365         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
31366         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
31367         (vect_shift_permute_load_chain): Likewise.
31368         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
31369         (expand_vector_condition, optimize_vector_constructor): Likewise.
31370         (lower_vec_perm, get_compute_type): Likewise.
31371         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
31372         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
31373         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
31374         (vect_recog_mask_conversion_pattern): Likewise.
31375         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
31376         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
31377         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
31378         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
31379         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
31380         (vectorizable_shift, vectorizable_operation, vectorizable_store)
31381         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
31382         (supportable_widening_operation): Likewise.
31383         (supportable_narrowing_operation): Likewise.
31384         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
31385         Likewise.
31386         * varasm.c (output_constant): Likewise.
31388 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31389             Alan Hayward  <alan.hayward@arm.com>
31390             David Sherwood  <david.sherwood@arm.com>
31392         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
31393         so that both the length == 3 and length != 3 cases set up their
31394         own permute vectors.  Add comments explaining why we know the
31395         number of elements is constant.
31396         (vect_permute_load_chain): Likewise.
31398 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31399             Alan Hayward  <alan.hayward@arm.com>
31400             David Sherwood  <david.sherwood@arm.com>
31402         * machmode.h (mode_nunits): Change from unsigned char to
31403         poly_uint16_pod.
31404         (ONLY_FIXED_SIZE_MODES): New macro.
31405         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
31406         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
31407         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
31408         New typedefs.
31409         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
31410         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
31411         or if measurement_type is not polynomial.
31412         * genmodes.c (ZERO_COEFFS): New macro.
31413         (emit_mode_nunits_inline): Make mode_nunits_inline return a
31414         poly_uint16.
31415         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
31416         Use ZERO_COEFFS when emitting initializers.
31417         * data-streamer.h (bp_pack_poly_value): New function.
31418         (bp_unpack_poly_value): Likewise.
31419         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
31420         for GET_MODE_NUNITS.
31421         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
31422         for GET_MODE_NUNITS.
31423         * tree.c (make_vector_type): Remove temporary shim and make
31424         the real function take the number of units as a poly_uint64
31425         rather than an int.
31426         (build_vector_type_for_mode): Handle polynomial nunits.
31427         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
31428         * emit-rtl.c (const_vec_series_p_1): Likewise.
31429         (gen_rtx_CONST_VECTOR): Likewise.
31430         * fold-const.c (test_vec_duplicate_folding): Likewise.
31431         * genrecog.c (validate_pattern): Likewise.
31432         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
31433         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
31434         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
31435         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
31436         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
31437         * rtlanal.c (subreg_get_info): Likewise.
31438         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
31439         (vect_grouped_load_supported): Likewise.
31440         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
31441         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
31442         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
31443         (simplify_const_unary_operation, simplify_binary_operation_1)
31444         (simplify_const_binary_operation, simplify_ternary_operation)
31445         (test_vector_ops_duplicate, test_vector_ops): Likewise.
31446         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
31447         instead of CONST_VECTOR_NUNITS.
31448         * varasm.c (output_constant_pool_2): Likewise.
31449         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
31450         explicit-encoded elements in the XVEC for variable-length vectors.
31452 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31454         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
31456 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31457             Alan Hayward  <alan.hayward@arm.com>
31458             David Sherwood  <david.sherwood@arm.com>
31460         * coretypes.h (fixed_size_mode): Declare.
31461         (fixed_size_mode_pod): New typedef.
31462         * builtins.h (target_builtins::x_apply_args_mode)
31463         (target_builtins::x_apply_result_mode): Change type to
31464         fixed_size_mode_pod.
31465         * builtins.c (apply_args_size, apply_result_size, result_vector)
31466         (expand_builtin_apply_args_1, expand_builtin_apply)
31467         (expand_builtin_return): Update accordingly.
31469 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31471         * cse.c (hash_rtx_cb): Hash only the encoded elements.
31472         * cselib.c (cselib_hash_rtx): Likewise.
31473         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
31474         CONST_VECTOR encoding.
31476 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31477             Jeff Law  <law@redhat.com>
31479         PR target/83641
31480         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
31481         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
31482         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
31483         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
31485         PR target/83641
31486         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
31487         explicitly probe *sp in a noreturn function if there were any callee
31488         register saves or frame pointer is needed.
31490 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31492         PR debug/83621
31493         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
31494         BLKmode for ternary, binary or unary expressions.
31496         PR debug/83645
31497         * var-tracking.c (delete_vta_debug_insn): New inline function.
31498         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
31499         insns from get_insns () to NULL instead of each bb separately.
31500         Use delete_vta_debug_insn.  No longer static.
31501         (vt_debug_insns_local, variable_tracking_main_1): Adjust
31502         delete_vta_debug_insns callers.
31503         * rtl.h (delete_vta_debug_insns): Declare.
31504         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
31505         instead of variable_tracking_main.
31507 2018-01-03  Martin Sebor  <msebor@redhat.com>
31509         PR tree-optimization/83603
31510         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
31511         arguments past the endof the argument list in functions declared
31512         without a prototype.
31513         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
31514         Avoid checking when arguments are null.
31516 2018-01-03  Martin Sebor  <msebor@redhat.com>
31518         PR c/83559
31519         * doc/extend.texi (attribute const): Fix a typo.
31520         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
31521         issuing -Wsuggest-attribute for void functions.
31523 2018-01-03  Martin Sebor  <msebor@redhat.com>
31525         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
31526         offset_int::from instead of wide_int::to_shwi.
31527         (maybe_diag_overlap): Remove assertion.
31528         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
31529         * gimple-ssa-sprintf.c (format_directive): Same.
31530         (parse_directive): Same.
31531         (sprintf_dom_walker::compute_format_length): Same.
31532         (try_substitute_return_value): Same.
31534 2018-01-03  Jeff Law  <law@redhat.com>
31536         PR middle-end/83654
31537         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
31538         non-constant residual for zero at runtime and avoid probing in
31539         that case.  Reorganize code for trailing problem to mirror handling
31540         of the residual.
31542 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
31544         PR tree-optimization/83501
31545         * tree-ssa-strlen.c (get_string_cst): New.
31546         (handle_char_store): Call get_string_cst.
31548 2018-01-03  Martin Liska  <mliska@suse.cz>
31550         PR tree-optimization/83593
31551         * tree-ssa-strlen.c: Include tree-cfg.h.
31552         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
31553         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
31554         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
31555         to false.
31556         (strlen_dom_walker::before_dom_children): Call
31557         gimple_purge_dead_eh_edges. Dump tranformation with details
31558         dump flags.
31559         (strlen_dom_walker::before_dom_children): Update call by adding
31560         new argument cleanup_eh.
31561         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
31563 2018-01-03  Martin Liska  <mliska@suse.cz>
31565         PR ipa/83549
31566         * cif-code.def (VARIADIC_THUNK): New enum value.
31567         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
31568         thunks.
31570 2018-01-03  Jan Beulich  <jbeulich@suse.com>
31572         * sse.md (mov<mode>_internal): Tighten condition for when to use
31573         vmovdqu<ssescalarsize> for TI and OI modes.
31575 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31577         Update copyright years.
31579 2018-01-03  Martin Liska  <mliska@suse.cz>
31581         PR ipa/83594
31582         * ipa-visibility.c (function_and_variable_visibility): Skip
31583         functions with noipa attribure.
31585 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
31587         * gcc.c (process_command): Update copyright notice dates.
31588         * gcov-dump.c (print_version): Ditto.
31589         * gcov.c (print_version): Ditto.
31590         * gcov-tool.c (print_version): Ditto.
31591         * gengtype.c (create_file): Ditto.
31592         * doc/cpp.texi: Bump @copying's copyright year.
31593         * doc/cppinternals.texi: Ditto.
31594         * doc/gcc.texi: Ditto.
31595         * doc/gccint.texi: Ditto.
31596         * doc/gcov.texi: Ditto.
31597         * doc/install.texi: Ditto.
31598         * doc/invoke.texi: Ditto.
31600 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31602         * vector-builder.h (vector_builder::m_full_nelts): Change from
31603         unsigned int to poly_uint64.
31604         (vector_builder::full_nelts): Update prototype accordingly.
31605         (vector_builder::new_vector): Likewise.
31606         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
31607         (vector_builder::operator ==): Likewise.
31608         (vector_builder::finalize): Likewise.
31609         * int-vector-builder.h (int_vector_builder::int_vector_builder):
31610         Take the number of elements as a poly_uint64 rather than an
31611         unsigned int.
31612         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
31613         from unsigned int to poly_uint64.
31614         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
31615         (vec_perm_indices::new_vector): Likewise.
31616         (vec_perm_indices::length): Likewise.
31617         (vec_perm_indices::nelts_per_input): Likewise.
31618         (vec_perm_indices::input_nelts): Likewise.
31619         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
31620         number of elements per input as a poly_uint64 rather than an
31621         unsigned int.  Use the original encoding for variable-length
31622         vectors, rather than clamping each individual element.
31623         For the second and subsequent elements in each pattern,
31624         clamp the step and base before clamping their sum.
31625         (vec_perm_indices::series_p): Handle polynomial element counts.
31626         (vec_perm_indices::all_in_range_p): Likewise.
31627         (vec_perm_indices_to_tree): Likewise.
31628         (vec_perm_indices_to_rtx): Likewise.
31629         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
31630         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
31631         (tree_vector_builder::new_binary_operation): Handle polynomial
31632         element counts.  Return false if we need to know the number
31633         of elements at compile time.
31634         * fold-const.c (fold_vec_perm): Punt if the number of elements
31635         isn't known at compile time.
31637 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31639         * vec-perm-indices.h (vec_perm_builder): Change element type
31640         from HOST_WIDE_INT to poly_int64.
31641         (vec_perm_indices::element_type): Update accordingly.
31642         (vec_perm_indices::clamp): Handle polynomial element_types.
31643         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
31644         (vec_perm_indices::all_in_range_p): Likewise.
31645         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
31646         than shwi trees.
31647         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
31648         polynomial vec_perm_indices element types.
31649         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
31650         * fold-const.c (fold_vec_perm): Likewise.
31651         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
31652         * tree-vect-generic.c (lower_vec_perm): Likewise.
31653         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
31654         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
31655         element type to HOST_WIDE_INT.
31657 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31658             Alan Hayward  <alan.hayward@arm.com>
31659             David Sherwood  <david.sherwood@arm.com>
31661         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
31662         rather than an int.  Use plus_constant.
31663         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
31664         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
31666 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31667             Alan Hayward  <alan.hayward@arm.com>
31668             David Sherwood  <david.sherwood@arm.com>
31670         * calls.c (emit_call_1, expand_call): Change struct_value_size from
31671         a HOST_WIDE_INT to a poly_int64.
31673 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31674             Alan Hayward  <alan.hayward@arm.com>
31675             David Sherwood  <david.sherwood@arm.com>
31677         * calls.c (load_register_parameters): Cope with polynomial
31678         mode sizes.  Require a constant size for BLKmode parameters
31679         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
31680         forces a parameter to be padded at the lsb end in order to
31681         fill a complete number of words, require the parameter size
31682         to be ordered wrt UNITS_PER_WORD.
31684 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31685             Alan Hayward  <alan.hayward@arm.com>
31686             David Sherwood  <david.sherwood@arm.com>
31688         * reload1.c (spill_stack_slot_width): Change element type
31689         from unsigned int to poly_uint64_pod.
31690         (alter_reg): Treat mode sizes as polynomial.
31692 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31693             Alan Hayward  <alan.hayward@arm.com>
31694             David Sherwood  <david.sherwood@arm.com>
31696         * reload.c (complex_word_subreg_p): New function.
31697         (reload_inner_reg_of_subreg, push_reload): Use it.
31699 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31700             Alan Hayward  <alan.hayward@arm.com>
31701             David Sherwood  <david.sherwood@arm.com>
31703         * lra-constraints.c (process_alt_operands): Reject matched
31704         operands whose sizes aren't ordered.
31705         (match_reload): Refer to this check here.
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         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
31712         that the mode size is in the set {1, 2, 4, 8, 16}.
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         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
31719         Use plus_constant instead of gen_rtx_PLUS.
31721 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31722             Alan Hayward  <alan.hayward@arm.com>
31723             David Sherwood  <david.sherwood@arm.com>
31725         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
31726         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
31727         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
31728         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
31729         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
31730         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
31731         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
31732         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
31733         * config/i386/i386.c (ix86_push_rounding): ...this new function.
31734         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
31735         a poly_int64.
31736         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
31737         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
31738         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
31739         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
31740         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
31741         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
31742         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
31743         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
31744         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
31745         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
31746         function.
31747         * expr.c (emit_move_resolve_push): Treat the input and result
31748         of PUSH_ROUNDING as a poly_int64.
31749         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
31750         (emit_push_insn): Likewise.
31751         * lra-eliminations.c (mark_not_eliminable): Likewise.
31752         * recog.c (push_operand): Likewise.
31753         * reload1.c (elimination_effects): Likewise.
31754         * rtlanal.c (nonzero_bits1): Likewise.
31755         * calls.c (store_one_arg): Likewise.  Require the padding to be
31756         known at compile time.
31758 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31759             Alan Hayward  <alan.hayward@arm.com>
31760             David Sherwood  <david.sherwood@arm.com>
31762         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
31763         Use plus_constant instead of gen_rtx_PLUS.
31765 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31766             Alan Hayward  <alan.hayward@arm.com>
31767             David Sherwood  <david.sherwood@arm.com>
31769         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
31770         rather than an int.
31772 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31773             Alan Hayward  <alan.hayward@arm.com>
31774             David Sherwood  <david.sherwood@arm.com>
31776         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
31777         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
31778         via stack temporaries.  Treat the mode size as polynomial too.
31780 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31781             Alan Hayward  <alan.hayward@arm.com>
31782             David Sherwood  <david.sherwood@arm.com>
31784         * expr.c (expand_expr_real_2): When handling conversions involving
31785         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
31786         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
31787         as a poly_uint64 too.
31789 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31790             Alan Hayward  <alan.hayward@arm.com>
31791             David Sherwood  <david.sherwood@arm.com>
31793         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
31795 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31796             Alan Hayward  <alan.hayward@arm.com>
31797             David Sherwood  <david.sherwood@arm.com>
31799         * combine.c (can_change_dest_mode): Handle polynomial
31800         REGMODE_NATURAL_SIZE.
31801         * expmed.c (store_bit_field_1): Likewise.
31802         * expr.c (store_constructor): Likewise.
31803         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
31804         and polynomial REGMODE_NATURAL_SIZE.
31805         (gen_lowpart_common): Likewise.
31806         * reginfo.c (record_subregs_of_mode): Likewise.
31807         * rtlanal.c (read_modify_subreg_p): Likewise.
31809 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31810             Alan Hayward  <alan.hayward@arm.com>
31811             David Sherwood  <david.sherwood@arm.com>
31813         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
31814         numbers of elements.
31816 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31817             Alan Hayward  <alan.hayward@arm.com>
31818             David Sherwood  <david.sherwood@arm.com>
31820         * match.pd: Cope with polynomial numbers of vector elements.
31822 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31823             Alan Hayward  <alan.hayward@arm.com>
31824             David Sherwood  <david.sherwood@arm.com>
31826         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
31827         in a POINTER_PLUS_EXPR.
31829 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31830             Alan Hayward  <alan.hayward@arm.com>
31831             David Sherwood  <david.sherwood@arm.com>
31833         * omp-simd-clone.c (simd_clone_subparts): New function.
31834         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
31835         (ipa_simd_modify_function_body): Likewise.
31837 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31838             Alan Hayward  <alan.hayward@arm.com>
31839             David Sherwood  <david.sherwood@arm.com>
31841         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
31842         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
31843         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
31844         (expand_vector_condition, vector_element): Likewise.
31845         (subparts_gt): New function.
31846         (get_compute_type): Use subparts_gt.
31847         (count_type_subparts): Delete.
31848         (expand_vector_operations_1): Use subparts_gt instead of
31849         count_type_subparts.
31851 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31852             Alan Hayward  <alan.hayward@arm.com>
31853             David Sherwood  <david.sherwood@arm.com>
31855         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
31856         (vect_compile_time_alias): ...this new function.  Do the calculation
31857         on poly_ints rather than trees.
31858         (vect_prune_runtime_alias_test_list): Update call accordingly.
31860 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31861             Alan Hayward  <alan.hayward@arm.com>
31862             David Sherwood  <david.sherwood@arm.com>
31864         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
31865         numbers of units.
31866         (vect_schedule_slp_instance): Likewise.
31868 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31869             Alan Hayward  <alan.hayward@arm.com>
31870             David Sherwood  <david.sherwood@arm.com>
31872         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
31873         constant and extern definitions for variable-length vectors.
31874         (vect_get_constant_vectors): Note that the number of units
31875         is known to be constant.
31877 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31878             Alan Hayward  <alan.hayward@arm.com>
31879             David Sherwood  <david.sherwood@arm.com>
31881         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
31882         of units as polynomial.  Choose between WIDE and NARROW based
31883         on multiple_p.
31885 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31886             Alan Hayward  <alan.hayward@arm.com>
31887             David Sherwood  <david.sherwood@arm.com>
31889         * tree-vect-stmts.c (simd_clone_subparts): New function.
31890         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_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-stmts.c (vectorizable_call): Treat the number of
31897         vectors as polynomial.  Use build_index_vector for
31898         IFN_GOMP_SIMD_LANE.
31900 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31901             Alan Hayward  <alan.hayward@arm.com>
31902             David Sherwood  <david.sherwood@arm.com>
31904         * tree-vect-stmts.c (get_load_store_type): Treat the number of
31905         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
31906         for variable-length vectors.
31907         (vectorizable_mask_load_store): Treat the number of units as
31908         polynomial, asserting that it is constant if the condition has
31909         already been enforced.
31910         (vectorizable_store, vectorizable_load): Likewise.
31912 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31913             Alan Hayward  <alan.hayward@arm.com>
31914             David Sherwood  <david.sherwood@arm.com>
31916         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
31917         of units as polynomial.  Punt if we can't tell at compile time
31918         which vector contains the final result.
31920 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31921             Alan Hayward  <alan.hayward@arm.com>
31922             David Sherwood  <david.sherwood@arm.com>
31924         * tree-vect-loop.c (vectorizable_induction): Treat the number
31925         of units as polynomial.  Punt on SLP inductions.  Use an integer
31926         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
31927         cast of such a series for variable-length floating-point
31928         reductions.
31930 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31931             Alan Hayward  <alan.hayward@arm.com>
31932             David Sherwood  <david.sherwood@arm.com>
31934         * tree.h (build_index_vector): Declare.
31935         * tree.c (build_index_vector): New function.
31936         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
31937         of units as polynomial, forcibly converting it to a constant if
31938         vectorizable_reduction has already enforced the condition.
31939         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
31940         to create a {1,2,3,...} vector.
31941         (vectorizable_reduction): Treat the number of units as polynomial.
31942         Choose vectype_in based on the largest scalar element size rather
31943         than the smallest number of units.  Enforce the restrictions
31944         relied on above.
31946 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31947             Alan Hayward  <alan.hayward@arm.com>
31948             David Sherwood  <david.sherwood@arm.com>
31950         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
31951         number of units as polynomial.
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         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
31958         * target.def (autovectorize_vector_sizes): Return the vector sizes
31959         by pointer, using vector_sizes rather than a bitmask.
31960         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
31961         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
31962         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
31963         Likewise.
31964         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
31965         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
31966         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
31967         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
31968         * omp-general.c (omp_max_vf): Likewise.
31969         * omp-low.c (omp_clause_aligned_alignment): Likewise.
31970         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
31971         * tree-vect-loop.c (vect_analyze_loop): Likewise.
31972         * tree-vect-slp.c (vect_slp_bb): Likewise.
31973         * doc/tm.texi: Regenerate.
31974         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
31975         to a poly_uint64.
31976         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
31977         the vector size as a poly_uint64 rather than an unsigned int.
31978         (current_vector_size): Change from an unsigned int to a poly_uint64.
31979         (get_vectype_for_scalar_type): Update accordingly.
31980         * tree.h (build_truth_vector_type): Take the size and number of
31981         units as a poly_uint64 rather than an unsigned int.
31982         (build_vector_type): Add a temporary overload that takes
31983         the number of units as a poly_uint64 rather than an unsigned int.
31984         * tree.c (make_vector_type): Likewise.
31985         (build_truth_vector_type): Take the number of units as a poly_uint64
31986         rather than an unsigned int.
31988 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
31989             Alan Hayward  <alan.hayward@arm.com>
31990             David Sherwood  <david.sherwood@arm.com>
31992         * target.def (get_mask_mode): Take the number of units and length
31993         as poly_uint64s rather than unsigned ints.
31994         * targhooks.h (default_get_mask_mode): Update accordingly.
31995         * targhooks.c (default_get_mask_mode): Likewise.
31996         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
31997         * doc/tm.texi: Regenerate.
31999 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32000             Alan Hayward  <alan.hayward@arm.com>
32001             David Sherwood  <david.sherwood@arm.com>
32003         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
32004         * omp-general.c (omp_max_vf): Likewise.
32005         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
32006         (expand_omp_simd): Handle polynomial safelen.
32007         * omp-low.c (omplow_simd_context): Add a default constructor.
32008         (omplow_simd_context::max_vf): Change from int to poly_uint64.
32009         (lower_rec_simd_input_clauses): Update accordingly.
32010         (lower_rec_input_clauses): Likewise.
32012 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32013             Alan Hayward  <alan.hayward@arm.com>
32014             David Sherwood  <david.sherwood@arm.com>
32016         * tree-vectorizer.h (vect_nunits_for_cost): New function.
32017         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
32018         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
32019         (vect_analyze_slp_cost): Likewise.
32020         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
32021         (vect_model_load_cost): Likewise.
32023 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32024             Alan Hayward  <alan.hayward@arm.com>
32025             David Sherwood  <david.sherwood@arm.com>
32027         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
32028         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
32029         from an unsigned int * to a poly_uint64_pod *.
32030         (calculate_unrolling_factor): New function.
32031         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
32033 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32034             Alan Hayward  <alan.hayward@arm.com>
32035             David Sherwood  <david.sherwood@arm.com>
32037         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
32038         from an unsigned int to a poly_uint64.
32039         (_loop_vec_info::slp_unrolling_factor): Likewise.
32040         (_loop_vec_info::vectorization_factor): Change from an int
32041         to a poly_uint64.
32042         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
32043         (vect_get_num_vectors): New function.
32044         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
32045         (vect_get_num_copies): Use vect_get_num_vectors.
32046         (vect_analyze_data_ref_dependences): Change max_vf from an int *
32047         to an unsigned int *.
32048         (vect_analyze_data_refs): Change min_vf from an int * to a
32049         poly_uint64 *.
32050         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
32051         than an unsigned HOST_WIDE_INT.
32052         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
32053         (vect_analyze_data_ref_dependence): Change max_vf from an int *
32054         to an unsigned int *.
32055         (vect_analyze_data_ref_dependences): Likewise.
32056         (vect_compute_data_ref_alignment): Handle polynomial vf.
32057         (vect_enhance_data_refs_alignment): Likewise.
32058         (vect_prune_runtime_alias_test_list): Likewise.
32059         (vect_shift_permute_load_chain): Likewise.
32060         (vect_supportable_dr_alignment): Likewise.
32061         (dependence_distance_ge_vf): Take the vectorization factor as a
32062         poly_uint64 rather than an unsigned HOST_WIDE_INT.
32063         (vect_analyze_data_refs): Change min_vf from an int * to a
32064         poly_uint64 *.
32065         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
32066         vfm1 as a poly_uint64 rather than an int.  Make the same change
32067         for the returned bound_scalar.
32068         (vect_gen_vector_loop_niters): Handle polynomial vf.
32069         (vect_do_peeling): Likewise.  Update call to
32070         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
32071         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
32072         be constant.
32073         * tree-vect-loop.c (vect_determine_vectorization_factor)
32074         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
32075         (vect_get_known_peeling_cost): Likewise.
32076         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
32077         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
32078         (vect_transform_loop): Likewise.  Use the lowest possible VF when
32079         updating the upper bounds of the loop.
32080         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
32081         rather than an int.
32082         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
32083         polynomial unroll factors.
32084         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
32085         (vect_make_slp_decision): Likewise.
32086         (vect_supported_load_permutation_p): Likewise, and polynomial
32087         vf too.
32088         (vect_analyze_slp_cost): Handle polynomial vf.
32089         (vect_slp_analyze_node_operations): Likewise.
32090         (vect_slp_analyze_bb_1): Likewise.
32091         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
32092         than an unsigned HOST_WIDE_INT.
32093         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
32094         (vectorizable_load): Handle polynomial vf.
32095         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
32096         a poly_uint64.
32097         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
32099 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32100             Alan Hayward  <alan.hayward@arm.com>
32101             David Sherwood  <david.sherwood@arm.com>
32103         * match.pd: Handle bit operations involving three constants
32104         and try to fold one pair.
32106 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
32108         * tree-vect-loop-manip.c: Include gimple-fold.h.
32109         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
32110         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
32111         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
32112         Add a path that uses a step of VF instead of 1, but disable it
32113         for now.
32114         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
32115         and niters_no_overflow parameters.  Update calls to
32116         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
32117         Create a new SSA name if the latter choses to use a ste other
32118         than zero, and return it via niters_vector_mult_vf_var.
32119         * tree-vect-loop.c (vect_transform_loop): Update calls to
32120         vect_do_peeling, vect_gen_vector_loop_niters and
32121         slpeel_make_loop_iterate_ntimes.
32122         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
32123         (vect_gen_vector_loop_niters): Update declarations after above changes.
32125 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
32127         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
32128         128-bit round to integer instructions.
32129         (ceil<mode>2): Likewise.
32130         (btrunc<mode>2): Likewise.
32131         (round<mode>2): Likewise.
32133 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32135         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
32136         unaligned VSX load/store on P8/P9.
32137         (expand_block_clear): Allow the use of unaligned VSX
32138         load/store on P8/P9.
32140 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32142         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
32143         New function.
32144         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
32145         swap associated with both a load and a store.
32147 2018-01-02  Andrew Waterman  <andrew@sifive.com>
32149         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
32150         * config/riscv/riscv.md (clear_cache): Use it.
32152 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
32154         * web.c: Remove out-of-date comment.
32156 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32158         * expr.c (fixup_args_size_notes): Check that any existing
32159         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
32160         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
32161         (emit_single_push_insn): ...here.
32163 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32165         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
32166         (const_vector_encoded_nelts): New function.
32167         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
32168         (const_vector_int_elt, const_vector_elt): Declare.
32169         * emit-rtl.c (const_vector_int_elt_1): New function.
32170         (const_vector_elt): Likewise.
32171         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
32172         of CONST_VECTOR_ELT.
32174 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32176         * expr.c: Include rtx-vector-builder.h.
32177         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
32178         directly on the tree encoding.
32179         (const_vector_from_tree): Likewise.
32180         * optabs.c: Include rtx-vector-builder.h.
32181         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
32182         sequence of "u" values.
32183         * vec-perm-indices.c: Include rtx-vector-builder.h.
32184         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
32185         directly on the vec_perm_indices encoding.
32187 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32189         * doc/rtl.texi (const_vector): Describe new encoding scheme.
32190         * Makefile.in (OBJS): Add rtx-vector-builder.o.
32191         * rtx-vector-builder.h: New file.
32192         * rtx-vector-builder.c: Likewise.
32193         * rtl.h (rtx_def::u2): Add a const_vector field.
32194         (CONST_VECTOR_NPATTERNS): New macro.
32195         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
32196         (CONST_VECTOR_DUPLICATE_P): Likewise.
32197         (CONST_VECTOR_STEPPED_P): Likewise.
32198         (CONST_VECTOR_ENCODED_ELT): Likewise.
32199         (const_vec_duplicate_p): Check for a duplicated vector encoding.
32200         (unwrap_const_vec_duplicate): Likewise.
32201         (const_vec_series_p): Check for a non-duplicated vector encoding.
32202         Say that the function only returns true for integer vectors.
32203         * emit-rtl.c: Include rtx-vector-builder.h.
32204         (gen_const_vec_duplicate_1): Delete.
32205         (gen_const_vector): Call gen_const_vec_duplicate instead of
32206         gen_const_vec_duplicate_1.
32207         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
32208         (gen_const_vec_duplicate): Use rtx_vector_builder.
32209         (gen_const_vec_series): Likewise.
32210         (gen_rtx_CONST_VECTOR): Likewise.
32211         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
32212         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32213         Build a new vector rather than modifying a CONST_VECTOR in-place.
32214         (handle_special_swappables): Update call accordingly.
32215         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
32216         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
32217         Build a new vector rather than modifying a CONST_VECTOR in-place.
32218         (handle_special_swappables): Update call accordingly.
32220 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32222         * simplify-rtx.c (simplify_const_binary_operation): Use
32223         CONST_VECTOR_ELT instead of XVECEXP.
32225 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32227         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
32228         the selector elements to be different from the data elements
32229         if the selector is a VECTOR_CST.
32230         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
32231         ssizetype for the selector.
32233 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32235         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
32236         before testing each element individually.
32237         * tree-vect-generic.c (lower_vec_perm): Likewise.
32239 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32241         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
32242         * selftest-run-tests.c (selftest::run_tests): Call it.
32243         * vector-builder.h (vector_builder::operator ==): New function.
32244         (vector_builder::operator !=): Likewise.
32245         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
32246         (vec_perm_indices::all_from_input_p): New function.
32247         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
32248         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
32249         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
32250         instead of reading the VECTOR_CST directly.  Detect whether both
32251         vector inputs are the same before constructing the vec_perm_indices,
32252         and update the number of inputs argument accordingly.  Use the
32253         utility functions added above.  Only construct sel2 if we need to.
32255 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32257         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
32258         the broadcast of the low byte.
32259         (expand_mult_highpart): Use an explicit encoding for the permutes.
32260         * optabs-query.c (can_mult_highpart_p): Likewise.
32261         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
32262         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32263         (vectorizable_bswap): Likewise.
32264         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
32265         explicit encoding for the power-of-2 permutes.
32266         (vect_permute_store_chain): Likewise.
32267         (vect_grouped_load_supported): Likewise.
32268         (vect_permute_load_chain): Likewise.
32270 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32272         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
32273         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
32274         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
32275         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
32276         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
32277         (vect_gen_perm_mask_any): Likewise.
32279 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32281         * int-vector-builder.h: New file.
32282         * vec-perm-indices.h: Include int-vector-builder.h.
32283         (vec_perm_indices): Redefine as an int_vector_builder.
32284         (auto_vec_perm_indices): Delete.
32285         (vec_perm_builder): Redefine as a stand-alone class.
32286         (vec_perm_indices::vec_perm_indices): New function.
32287         (vec_perm_indices::clamp): Likewise.
32288         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
32289         (vec_perm_indices::new_vector): New function.
32290         (vec_perm_indices::new_expanded_vector): Update for new
32291         vec_perm_indices class.
32292         (vec_perm_indices::rotate_inputs): New function.
32293         (vec_perm_indices::all_in_range_p): Operate directly on the
32294         encoded form, without computing elided elements.
32295         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
32296         encoding.  Update for new vec_perm_indices class.
32297         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
32298         the given vec_perm_builder.
32299         (expand_vec_perm_var): Update vec_perm_builder constructor.
32300         (expand_mult_highpart): Use vec_perm_builder instead of
32301         auto_vec_perm_indices.
32302         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
32303         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
32304         or double series encoding as appropriate.
32305         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
32306         vec_perm_indices instead of auto_vec_perm_indices.
32307         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32308         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32309         (vect_permute_store_chain): Likewise.
32310         (vect_grouped_load_supported): Likewise.
32311         (vect_permute_load_chain): Likewise.
32312         (vect_shift_permute_load_chain): Likewise.
32313         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32314         (vect_transform_slp_perm_load): Likewise.
32315         (vect_schedule_slp_instance): Likewise.
32316         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32317         (vectorizable_mask_load_store): Likewise.
32318         (vectorizable_bswap): Likewise.
32319         (vectorizable_store): Likewise.
32320         (vectorizable_load): Likewise.
32321         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
32322         vec_perm_indices instead of auto_vec_perm_indices.  Use
32323         tree_to_vec_perm_builder to read the vector from a tree.
32324         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
32325         vec_perm_builder instead of a vec_perm_indices.
32326         (have_whole_vector_shift): Use vec_perm_builder and
32327         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
32328         truncation to calc_vec_perm_mask_for_shift.
32329         (vect_create_epilog_for_reduction): Likewise.
32330         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
32331         from auto_vec_perm_indices to vec_perm_indices.
32332         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32333         instead of changing individual elements.
32334         (aarch64_vectorize_vec_perm_const): Use new_vector to install
32335         the vector in d.perm.
32336         * config/arm/arm.c (expand_vec_perm_d::perm): Change
32337         from auto_vec_perm_indices to vec_perm_indices.
32338         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
32339         instead of changing individual elements.
32340         (arm_vectorize_vec_perm_const): Use new_vector to install
32341         the vector in d.perm.
32342         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
32343         Update vec_perm_builder constructor.
32344         (rs6000_expand_interleave): Likewise.
32345         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
32346         (rs6000_expand_interleave): Likewise.
32348 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32350         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
32351         to qimode could truncate the indices.
32352         * optabs.c (expand_vec_perm_var): Likewise.
32354 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32356         * Makefile.in (OBJS): Add vec-perm-indices.o.
32357         * vec-perm-indices.h: New file.
32358         * vec-perm-indices.c: Likewise.
32359         * target.h (vec_perm_indices): Replace with a forward class
32360         declaration.
32361         (auto_vec_perm_indices): Move to vec-perm-indices.h.
32362         * optabs.h: Include vec-perm-indices.h.
32363         (expand_vec_perm): Delete.
32364         (selector_fits_mode_p, expand_vec_perm_var): Declare.
32365         (expand_vec_perm_const): Declare.
32366         * target.def (vec_perm_const_ok): Replace with...
32367         (vec_perm_const): ...this new hook.
32368         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
32369         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
32370         * doc/tm.texi: Regenerate.
32371         * optabs.def (vec_perm_const): Delete.
32372         * doc/md.texi (vec_perm_const): Likewise.
32373         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
32374         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
32375         expand_vec_perm for constant permutation vectors.  Assert that
32376         the mode of variable permutation vectors is the integer equivalent
32377         of the mode that is being permuted.
32378         * optabs-query.h (selector_fits_mode_p): Declare.
32379         * optabs-query.c: Include vec-perm-indices.h.
32380         (selector_fits_mode_p): New function.
32381         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
32382         is defined, instead of checking whether the vec_perm_const_optab
32383         exists.  Use targetm.vectorize.vec_perm_const instead of
32384         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
32385         fit in the vector mode before using a variable permute.
32386         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
32387         vec_perm_indices instead of an rtx.
32388         (expand_vec_perm): Replace with...
32389         (expand_vec_perm_const): ...this new function.  Take the selector
32390         as a vec_perm_indices rather than an rtx.  Also take the mode of
32391         the selector.  Update call to shift_amt_for_vec_perm_mask.
32392         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
32393         Use vec_perm_indices::new_expanded_vector to expand the original
32394         selector into bytes.  Check whether the indices fit in the vector
32395         mode before using a variable permute.
32396         (expand_vec_perm_var): Make global.
32397         (expand_mult_highpart): Use expand_vec_perm_const.
32398         * fold-const.c: Includes vec-perm-indices.h.
32399         * tree-ssa-forwprop.c: Likewise.
32400         * tree-vect-data-refs.c: Likewise.
32401         * tree-vect-generic.c: Likewise.
32402         * tree-vect-loop.c: Likewise.
32403         * tree-vect-slp.c: Likewise.
32404         * tree-vect-stmts.c: Likewise.
32405         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
32406         Delete.
32407         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
32408         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
32409         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
32410         (aarch64_vectorize_vec_perm_const): ...this new function.
32411         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32412         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32413         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
32414         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
32415         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32416         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32417         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
32418         into...
32419         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
32420         check for NEON modes.
32421         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
32422         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
32423         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
32424         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
32425         into...
32426         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
32427         the old VEC_PERM_CONST conditions.
32428         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
32429         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
32430         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
32431         (ia64_vectorize_vec_perm_const_ok): Merge into...
32432         (ia64_vectorize_vec_perm_const): ...this new function.
32433         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
32434         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
32435         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
32436         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
32437         * config/mips/mips.c (mips_expand_vec_perm_const)
32438         (mips_vectorize_vec_perm_const_ok): Merge into...
32439         (mips_vectorize_vec_perm_const): ...this new function.
32440         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
32441         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
32442         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
32443         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
32444         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
32445         (rs6000_expand_vec_perm_const): Delete.
32446         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
32447         Delete.
32448         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32449         (altivec_expand_vec_perm_const_le): Take each operand individually.
32450         Operate on constant selectors rather than rtxes.
32451         (altivec_expand_vec_perm_const): Likewise.  Update call to
32452         altivec_expand_vec_perm_const_le.
32453         (rs6000_expand_vec_perm_const): Delete.
32454         (rs6000_vectorize_vec_perm_const_ok): Delete.
32455         (rs6000_vectorize_vec_perm_const): New function.
32456         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32457         an element count and rtx array.
32458         (rs6000_expand_extract_even): Update call accordingly.
32459         (rs6000_expand_interleave): Likewise.
32460         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
32461         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
32462         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
32463         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
32464         (rs6000_expand_vec_perm_const): Delete.
32465         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
32466         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32467         (altivec_expand_vec_perm_const_le): Take each operand individually.
32468         Operate on constant selectors rather than rtxes.
32469         (altivec_expand_vec_perm_const): Likewise.  Update call to
32470         altivec_expand_vec_perm_const_le.
32471         (rs6000_expand_vec_perm_const): Delete.
32472         (rs6000_vectorize_vec_perm_const_ok): Delete.
32473         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
32474         reference to the SPE evmerge intructions.
32475         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
32476         an element count and rtx array.
32477         (rs6000_expand_extract_even): Update call accordingly.
32478         (rs6000_expand_interleave): Likewise.
32479         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
32480         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
32481         new function.
32482         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
32484 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32486         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
32487         vector mode and that that mode matches the mode of the data
32488         being permuted.
32489         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
32490         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
32491         directly using expand_vec_perm_1 when forcing selectors into
32492         registers.
32493         (expand_vec_perm_var): New function, split out from expand_vec_perm.
32495 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32497         * optabs-query.h (can_vec_perm_p): Delete.
32498         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
32499         * optabs-query.c (can_vec_perm_p): Split into...
32500         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
32501         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
32502         particular selector is valid.
32503         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
32504         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
32505         (vect_grouped_load_supported): Likewise.
32506         (vect_shift_permute_load_chain): Likewise.
32507         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
32508         (vect_transform_slp_perm_load): Likewise.
32509         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
32510         (vectorizable_bswap): Likewise.
32511         (vect_gen_perm_mask_checked): Likewise.
32512         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
32513         implementations of variable permutation vectors into account
32514         when deciding which selector to use.
32515         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
32516         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
32517         with a false third argument.
32518         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
32519         to test whether the constant selector is valid and can_vec_perm_var_p
32520         to test whether a variable selector is valid.
32522 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32524         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
32525         * optabs-query.c (can_vec_perm_p): Likewise.
32526         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
32527         instead of vec_perm_indices.
32528         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
32529         (vect_gen_perm_mask_checked): Likewise,
32530         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
32531         (vect_gen_perm_mask_checked): Likewise,
32533 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
32535         * optabs-query.h (qimode_for_vec_perm): Declare.
32536         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
32537         (qimode_for_vec_perm): ...this new function.
32538         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
32540 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
32542         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
32543         does not have a conditional at the top.
32545 2018-01-02  Richard Biener  <rguenther@suse.de>
32547         * ipa-inline.c (big_speedup_p): Fix expression.
32549 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32551         PR target/81616
32552         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
32553         for generic 4->6.
32555 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
32557         PR target/81616
32558         Generic tuning.
32559         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
32560         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
32561         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
32562         cond_taken_branch_cost 3->4.
32564 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
32566         PR tree-optimization/83581
32567         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
32568         TODO_cleanup_cfg if any changes have been made.
32570         PR middle-end/83608
32571         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
32572         convert_modes if target mode has the right side, but different mode
32573         class.
32575         PR middle-end/83609
32576         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
32577         last argument when extracting from CONCAT.  If either from_real or
32578         from_imag is NULL, use expansion through memory.  If result is not
32579         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
32580         the parts directly to inner mode, if even that fails, use expansion
32581         through memory.
32583         PR middle-end/83623
32584         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
32585         check for bswap in mode rather than HImode and use that in expand_unop
32586         too.
32588 Copyright (C) 2018 Free Software Foundation, Inc.
32590 Copying and distribution of this file, with or without modification,
32591 are permitted in any medium without royalty provided the copyright
32592 notice and this notice are preserved.