Daily bump.
[official-gcc.git] / gcc / ChangeLog
blobb43cc388fa8187523e15638d45579e68c170230a
1 2018-10-01  Carl Love  <cel@us.ibm.com>
3         PR 69431
4         * config/rs6000/rs6000-builtin.def (__builtin_mffsl): New.
5         (__builtin_mtfsb0): New.
6         (__builtin_mtfsb1): New.
7         ( __builtin_set_fpscr_rn): New.
8         (__builtin_set_fpscr_drn): New.
9         * config/rs6000/rs6000.c (rs6000_expand_mtfsb_builtin): Add.
10         (rs6000_expand_set_fpscr_rn_builtin): Add.
11         (rs6000_expand_set_fpscr_drn_builtin): Add.
12         (rs6000_expand_builtin): Add case statement entries for
13         RS6000_BUILTIN_MTFSB0, RS6000_BUILTIN_MTFSB1,
14         RS6000_BUILTIN_SET_FPSCR_RN, RS6000_BUILTIN_SET_FPSCR_DRN,
15         RS6000_BUILTIN_MFFSL.
16         (rs6000_init_builtins): Add ftype initialization and def_builtin
17         calls for __builtin_mffsl, __builtin_mtfsb0, __builtin_mtfsb1,
18         __builtin_set_fpscr_rn, __builtin_set_fpscr_drn.
19         * config/rs6000.md (rs6000_mtfsb0, rs6000_mtfsb1, rs6000_mffscrn,
20         rs6000_mffscdrn): Add define_insn.
21         (rs6000_set_fpscr_rn, rs6000_set_fpscr_drn): Add define_expand.
22         * doc/extend.texi: Add documentation for the builtins.
24 2018-10-01  Richard Biener  <rguenther@suse.de>
26         PR tree-optimization/87465
27         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Fix typo
28         causing branch miscounts.
30 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
32         * common/config/aarch64/aarch64-common.c (TARGET_OPTION_DEFAULT_PARAM,
33         aarch64_option_default_param):  New.
34         (params.h): Include.
35         (TARGET_OPTION_VALIDATE_PARAM, aarch64_option_validate_param): New.
36         * config/aarch64/aarch64.c (aarch64_override_options_internal): Simplify
37         stack-clash protection validation code.
39 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
41         * params.c (validate_param): New.
42         (add_params): Use it.
43         (set_param_value): Refactor param validation into validate_param.
44         (diagnostic.h): Include.
45         * diagnostic.h (diagnostic_ready_p): New.
47 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
49         * params.c (set_param_value):
50         Add index of parameter being validated.
51         * common/common-target.def (option_validate_param): New.
52         * common/common-targhooks.h (default_option_validate_param): New.
53         * common/common-targhooks.c (default_option_validate_param): New.
54         * doc/tm.texi.in (TARGET_OPTION_VALIDATE_PARAM): New.
55         * doc/tm.texi: Regenerate.
57 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
59         PR target/86486
60         * config/aarch64/aarch64.c (aarch64_override_options_internal):
61         Add validation for stack-clash parameters and set defaults.
63 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
65         PR target/86486
66         * configure.ac: Add stack-clash-protection-guard-size.
67         * doc/install.texi: Document it.
68         * config.in (DEFAULT_STK_CLASH_GUARD_SIZE): New.
69         * params.def: Update comment for guard-size.
70         (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE,
71         PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Update description.
72         * configure: Regenerate.
74 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
76         PR target/86486
77         * config/aarch64/aarch64.h (STACK_CLASH_MIN_BYTES_OUTGOING_ARGS,
78         STACK_DYNAMIC_OFFSET): New.
79         * config/aarch64/aarch64.c (aarch64_layout_frame):
80         Update outgoing args size.
81         (aarch64_stack_clash_protection_alloca_probe_range,
82         TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
84 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
86         PR target/86486
87         * explow.c (anti_adjust_stack_and_probe_stack_clash): Support custom
88         probe ranges.
89         * target.def (stack_clash_protection_alloca_probe_range): New.
90         (stack_clash_protection_final_dynamic_probe): Remove.
91         * targhooks.h (default_stack_clash_protection_alloca_probe_range) New.
92         (default_stack_clash_protection_final_dynamic_probe): Remove.
93         * targhooks.c: Likewise.
94         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_ALLOCA_PROBE_RANGE): New.
95         (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE): Remove.
96         * doc/tm.texi: Regenerate.
98 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
100         PR target/86486
101         * config/aarch64/aarch64-protos.h (aarch64_output_probe_sve_stack_clash): New.
102         * config/aarch64/aarch64.c (aarch64_output_probe_sve_stack_clash,
103         aarch64_clamp_to_uimm12_shift): New.
104         (aarch64_allocate_and_probe_stack_space): Add SVE specific section.
105         * config/aarch64/aarch64.md (probe_sve_stack_clash): New.
107 2018-10-01  Tamar Christina  <tamar.christina@arm.com>
109         PR target/86486
110         * config/aarch64/aarch64.c (aarch64_layout_frame): Add assert.
112 2018-10-01  Jeff Law  <law@redhat.com>
113             Richard Sandiford <richard.sandiford@linaro.org>
114             Tamar Christina  <tamar.christina@arm.com>
116         PR target/86486
117         * config/aarch64/aarch64.md
118         (probe_stack_range): Add k (SP) constraint.
119         * config/aarch64/aarch64.h (STACK_CLASH_CALLER_GUARD,
120         STACK_CLASH_MAX_UNROLL_PAGES): New.
121         * config/aarch64/aarch64.c (aarch64_output_probe_stack_range): Emit
122         stack probes for stack clash.
123         (aarch64_allocate_and_probe_stack_space): New.
124         (aarch64_expand_prologue): Use it.
125         (aarch64_expand_epilogue): Likewise and update IP regs re-use criteria.
126         (aarch64_sub_sp): Add emit_move_imm optional param.
128 2018-10-01  MCC CS <deswurstes@users.noreply.github.com>
130         PR tree-optimization/87261
131         * match.pd: Remove trailing whitespace.
132         Add (x & y) | ~(x | y) -> ~(x ^ y),
133         (~x | y) ^ (x ^ y) -> x | ~y and (x ^ y) | ~(x | y) -> ~(x & y)
135 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
137         * config/arc/arc.md (*add_n): Clean up pattern, update instruction
138         constraints.
139         (ashlsi3_insn): Update instruction constraints.
140         (ashrsi3_insn): Likewise.
141         (rotrsi3): Likewise.
142         (add_shift): Likewise.
143         * config/arc/constraints.md (Csz): New 32 bit constraint. It
144         avoids placing in the limm field small constants which, otherwise,
145         could end into a small instruction.
147 2018-10-01  Claudiu Zissulescu  <claziss@synopsys.com>
149         * config/arc/arc.md (maddsidi4_split): Don't use dmac if the
150         destination register is not odd-even.
151         (umaddsidi4_split): Likewise.
153 2018-10-01  Richard Biener  <rguenther@suse.de>
155         * tree-inline.c (expand_call_inline): Store origin of fn
156         in BLOCK_ABSTRACT_ORIGIN for the inline BLOCK.
157         * tree.c (block_ultimate_origin): Simplify and do some
158         checking.
160 2018-09-30  Uros Bizjak  <ubizjak@gmail.com>
162         * config/i386/mmx.md (EMMS): New int iterator.
163         (emms): New int attribute.
164         (mmx_<emms>): Macroize insn from *mmx_emms and *mmx_femms using
165         EMMS int iterator.  Explicitly declare clobbers.
166         (mmx_emms): Remove expander.
167         (mmx_femms): Ditto.
168         * config/i386/predicates.md (emms_operation): Remove predicate.
169         (vzeroall_pattern): New predicate.
170         (vzeroupper_pattern): Rename from vzeroupper_operation.
171         * config/i386/i386.c (ix86_avx_u128_mode_after): Use
172         vzeroupper_pattern and vzeroall_pattern predicates.
174 2018-09-30  Peter Bergner  <bergner@linux.ibm.com>
176         PR rtl-optimization/86939
177         * ira-lives.c (make_hard_regno_born): Rename from this...
178         (make_hard_regno_live): ... to this.  Remove update to conflict
179         information.  Update function comment.
180         (make_hard_regno_dead): Add conflict information update.  Update
181         function comment.
182         (make_object_born): Rename from this...
183         (make_object_live): ... to this.  Remove update to conflict information.
184         Update function comment.
185         (make_object_dead):  Add conflict information update.  Update function
186         comment.
187         (mark_pseudo_regno_live): Call make_object_live.
188         (mark_pseudo_regno_subword_live): Likewise.
189         (mark_hard_reg_dead): Update function comment.
190         (mark_hard_reg_live): Call make_hard_regno_live.
191         (process_bb_node_lives): Likewise.
192         * lra-lives.c (make_hard_regno_born): Rename from this...
193         (make_hard_regno_live): ... to this.  Remove update to conflict
194         information.  Remove now uneeded check_pic_pseudo_p argument.
195         Update function comment.
196         (make_hard_regno_dead): Add check_pic_pseudo_p argument and add update
197         to conflict information.  Update function comment.
198         (mark_pseudo_live): Remove update to conflict information.  Update
199         function comment.
200         (mark_pseudo_dead): Add conflict information update.
201         (mark_regno_live): Call make_hard_regno_live.
202         (mark_regno_dead): Call make_hard_regno_dead with new arguement.
203         (process_bb_lives): Call make_hard_regno_live and make_hard_regno_dead.
205 2018-09-29  H.J. Lu  <hongjiu.lu@intel.com>
207         PR target/87370
208         * config/i386/i386.c (construct_container): Use TImode for
209         BLKmode values in 2 integer registers.
211 2018-09-29  Jeff Law  <law@redhat.com>
213         * builtins.c (unterminated_array): Pass in c_strlen_data * to
214         c_strlen rather than just a tree *.
215         (c_strlen): Change NONSTR argument to a c_strlen_data pointer.
216         Update recursive calls appropriately.  If caller did not provide a
217         suitable data pointer, create a local one.  When a non-terminated
218         string is discovered, bubble up information about the string via the
219         c_strlen_data object.
220         * builtins.h (c_strlen): Update prototype.
221         (c_strlen_data): New structure.
222         * gimple-fold.c (get_range_strlen): Update calls to c_strlen.
223         For a type 2 call, if c_strlen indicates a non-terminated string
224         use the length of the non-terminated string.
225         (gimple_fold_builtin_stpcpy): Update calls to c_strlen.
227 2018-09-29  Jakub Jelinek  <jakub@redhat.com>
229         PR target/87467
230         * config/i386/avx512fintrin.h (_mm512_abs_pd, _mm512_mask_abs_pd): Use
231         __m512d type for __A argument rather than __m512.
233 2018-09-28  John David Anglin  <danglin@gcc.gnu.org>
235         * match.pd (simple_comparison): Don't optimize if either operand is
236         a function pointer when target needs function pointer canonicalization.
238 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
240         * config/rs6000/driver-rs6000.c (asm_names): Adjust the entries for
241         power5 .. power9 to remove indirection.
242         * config/rs6000/rs6000.h (ASM_CPU_POWER5_SPEC, ASM_CPU_POWER6_SPEC,
243         ASM_CPU_POWER7_SPEC, ASM_CPU_POWER8_SPEC, ASM_CPU_POWER9_SPEC,
244         ASM_CPU_476_SPEC): Delete.
245         (ASM_CPU_SPEC): Adjust.
246         (EXTRA_SPECS): Delete asm_cpu_power5, asm_cpu_power6, asm_cpu_power7,
247         asm_cpu_power8, asm_cpu_power9, asm_cpu_476.
249 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
251         * config.in: Delete HAVE_AS_DCI.
252         * config/powerpcspe/powerpcspe.h: Treat HAVE_AS_DCI as always true.
253         * config/rs6000/rs6000.h: Ditto.
254         * configure.ac: Delete HAVE_AS_DCI.
255         * configure: Regenerate.
257 2018-09-28  Segher Boessenkool  <segher@kernel.crashing.org>
259         * config.in (HAVE_AS_LWSYNC): Delete.
260         * config/powerpcspe/powerpcspe.h (TARGET_LWSYNC_INSTRUCTION): Delete.
261         * config/powerpcspe/sync.md (*lwsync): Always generate lwsync, never
262         do it as a .long .
263         * config/rs6000/rs6000.h (TARGET_LWSYNC_INSTRUCTION): Delete.
264         * config/rs6000/sync.md (*lwsync): Always generate lwsync, never do it
265         as a .long .
266         * configure.ac: Delete HAVE_AS_LWSYNC.
267         * configure: Regenerate.
269 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
270             Pierre-Marie de Rodat  <derodat@adacore.com>
272         * calls.c (expand_call): Try to do a tail call for thunks at -O0 too.
273         * cgraph.h (struct cgraph_thunk_info): Add indirect_offset.
274         (cgraph_node::create_thunk): Add indirect_offset parameter.
275         (thunk_adjust): Likewise.
276         * cgraph.c (cgraph_node::create_thunk): Add indirect_offset parameter
277         and initialize the corresponding field with it.
278         (cgraph_node::dump): Dump indirect_offset field.
279         * cgraphclones.c (duplicate_thunk_for_node): Deal with indirect_offset.
280         * cgraphunit.c (cgraph_node::analyze): Be prepared for external thunks.
281         (thunk_adjust): Add indirect_offset parameter and deal with it.
282         (cgraph_node::expand_thunk): Deal with the indirect_offset field and
283         pass it to thunk_adjust.  Do not call the target hook if it's non-zero
284         or if the thunk is external or local.  Fix formatting.  Do not chain
285         the RESULT_DECL to BLOCK_VARS.  Pass the static chain to the target,
286         if any, in the GIMPLE representation.
287         * ipa-icf.c (sem_function::equals_wpa): Deal with indirect_offset.
288         * lto-cgraph.c (lto_output_node): Write indirect_offset field.
289         (input_node): Read indirect_offset field.
290         * tree-inline.c (expand_call_inline): Pass indirect_offset field in the
291         call to thunk_adjust.
292         * tree-nested.c (struct nesting_info): Add thunk_p field.
293         (create_nesting_tree): Set it.
294         (convert_all_function_calls): Copy static chain from targets to thunks.
295         (finalize_nesting_tree_1): Return early for thunks.
296         (unnest_nesting_tree_1): Do not finalize thunks.
297         (gimplify_all_functions): Do not gimplify thunks.
299 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
301         * opt-suggestions.c (option_proposer::build_option_suggestions):
302         Release "option_values".
304 2018-09-28  David Malcolm  <dmalcolm@redhat.com>
306         * coverage.c (get_coverage_counts): Convert problem-reporting dump
307         messages from MSG_OPTIMIZED_LOCATIONS to MSG_MISSED_OPTIMIZATION.
308         * dumpfile.c (kind_as_string): New function.
309         (dump_loc): Rather than a hardcoded prefix of "note: ", use
310         kind_as_string to vary the prefix based on dump_kind.
311         (selftest::test_capture_of_dump_calls): Update for above.
313 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
315         * config/i386/i386.h (SSE_REGNO): Fix check for FIRST_REX_SSE_REG.
316         (GET_SSE_REGNO): Rename from SSE_REGNO.  Update all uses for rename.
318 2018-09-28  Uros Bizjak  <ubizjak@gmail.com>
320         * config/i386/i386.h (CC_REGNO): Remove FPSR_REG.
321         * config/i386/i386.c (ix86_fixed_condition_code_regs): Use
322         INVALID_REGNUM instead of FPSR_REG.
323         (ix86_md_asm_adjust): Do not clobber FPSR_REG.
324         * config/i386/i386.md: Update comment of FP compares.
325         (fldenv): Do not clobber FPSR_REG.
327 2018-09-28  Richard Biener  <rguenther@suse.de>
329         * tree.h (BLOCK_ORIGIN): New.
330         * omp-expand.c (grid_expand_target_grid_body): Assign
331         BLOCK_ORIGIN to BLOCK_ABSTRACT_ORIGIN.
332         * tree-inline.c (remap_block): Likewise.
333         * auto-profile.c (get_function_decl_from_block): Simplify
334         by eliding the BLOCK_ABSTRACT_ORIGIN chasing.
335         * langhooks.c (lhd_print_error_function): Likewise.
336         * optinfo-emit-json.cc (optrecord_json_writer::inlining_chain_to):
337         Likewise.
338         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
339         * tree.c (block_nonartificial_location): Likewise.
340         (block_ultimate_origin): Likewise.
341         * tree-pretty-print.c (percent_K_format): Likewise.  Remove
342         no longer needed LTO case.
344 2018-09-28  Andrew Stubbs  <ams@codesourcery.com>
345             Jan Hubicka  <jh@suse.cz>
346             Martin Jambor  <mjambor@suse.cz>
348         * simplify-rtx.c (simplify_merge_mask): New function.
349         (simplify_ternary_operation): Use it, also see if VEC_MERGEs with the
350         same masks are used in op1 or op2.
351         (test_vec_merge): New function.
352         (test_vector_ops): Call test_vec_merge.
354 2018-09-28  Eric Botcazou  <ebotcazou@adacore.com>
356         * config/sparc/sparc-protos.h (sparc_branch_cost): Declare.
357         * config/sparc/sparc.h (BRANCH_COST): Call sparc_branch_cost.
358         * config/sparc/sparc.c (struct processor_costs): Add branch_cost field.
359         (cypress_costs): Set it.
360         (supersparc_costs): Likewise.
361         (hypersparc_costs): Likewise.
362         (leon_cost): Likewise.
363         (leon3_costs): Likewise.
364         (sparclet_costs): Likewise.
365         (ultrasparc_costs): Likewise.
366         (ultrasparc_costs): Likewise.
367         (niagara_costs): Likewise.
368         (niagara2_costs): Likewise.
369         (niagara3_costs): Likewise.
370         (niagara4_costs): Likewise.
371         (niagara7_costs): Likewise.
372         (m8_costs): Likewise.
373         (TARGET_CAN_FOLLOW_JUMP): Define.
374         (pass_work_around_errata::gate): Minor tweak.
375         (sparc_option_override): Remove MASK_FSMULD mask for V7 processors.
376         Do not set both MASK_VIS4 and MASK_VIS4B for M8 processor.
377         Automaitcally clear MASK_FSMULD mask for V7 processors.
378         (sparc_can_follow_jump): New static function.
379         (output_ubranch): Deal with CROSSING_JUMP_P.
380         (sparc_use_sched_lookahead): Rewrite using switch statement.
381         (sparc_issue_rate): Reorder.
382         (sparc_branch_cost): New function.
384 2018-09-27  Martin Sebor  <msebor@redhat.com>
386         * tree.h (tree_to_shwi): Add attribute nonnull and pure.
387         (tree_to_poly_int64, tree_to_uhwi, tree_to_poly_uint64): Same.
388         (int_fits_type_p): Same.
390 2018-09-27  Uros Bizjak  <ubizjak@gmail.com>
392         * config/i386/i386.md (FPCR_REG): Remove.
393         (UNSPEC_FLDCW): Remove.
394         (x86_fnstcw_1): Use (const_int 0) instead of FPCR_REG.
395         (x86_fldcw_1): Remove insn pattern.
396         (fnstenv): Do not clobber FPCR_REG.
397         (fldenv): Ditto.
398         * config/i386/i386.h (FIXED_REGISTERS) Remove fpsr register.
399         (CALL_USED_REGISTERS): Ditto.
400         (REG_ALLOC_ORDER): Ditto.
401         (REG_CLASS_CONTENTS): Ditto.
402         (HI_REGISTER_NAMES): Ditto.
403         (ADDITIONAL_REGISTER_NAMES): Use defines instead
404         of numerical constants.
405         * config/i386/i386.c (regclass_map): Remove fpsr register.
406         (dbx_register_map): Ditto.
407         (dbx64_register_map): Ditto.
408         (svr4_dbx_register_map): Ditto.
409         (print_reg): Do not handle FPCR_REG.
411 2018-09-27  Segher Boessenkool  <segher@kernel.crashing.org>
413         PR target/87149
414         * config.in (HAVE_AS_CMPB, HAVE_AS_DFP, HAVE_AS_FPRND, HAVE_AS_MFPGPR,
415         HAVE_AS_POPCNTB, HAVE_AS_POPCNTD, HAVE_AS_POWER8, HAVE_AS_POWER9):
416         Delete, always treat as true.
417         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
418         Ditto.  Simplify remaining code.
419         * config/powerpcspe/powerpcspe.h: Ditto.
420         * config/rs6000/rs6000.c (rs6000_option_override_internal): Ditto.
421         Simplify remaining code.
422         (rs6000_expand_builtin): Ditto.
423         * config/rs6000/rs6000.h: Ditto.
424         * configure.ac: Ditto.
425         * configure: Regenerate.
427 2018-09-27  Martin Liska  <mliska@suse.cz>
429         * coverage.c (get_coverage_counts): Revert the formatting
430         of missing profile opt info.
432 2018-09-27  Richard Biener  <rguenther@suse.de>
434         PR debug/37801
435         PR debug/87440
436         * dwarf2out.c (set_block_origin_self): Do not mark outermost
437         block as we do not output that.
438         (gen_inlined_subroutine_die): Elide the originally outermost
439         block, matching what we do for concrete instances.
440         (decls_for_scope): Add parameter specifying whether to recurse
441         to subblocks.
443 2018-09-27  Andrew Stubbs  <ams@codesourcery.com>
444             Tom de Vries  <tom@codesourcery.com>
446         PR 82089
448         * expmed.c (emit_cstore): Fix handling of result_mode == BImode and
449         STORE_FLAG_VALUE == 1.
451 2018-09-27  Andreas Krebbel  <krebbel@linux.ibm.com>
453         * config/s390/s390.md (PPA_TX_ABORT, PPA_OOO_BARRIER): New
454         constant definitions.
455         ("tx_assist"): Replace magic number with PPA_TX_ABORT.
456         ("*ppa"): Enable pattern also for -march=zEC12 -mno-htm.
457         ("speculation_barrier"): New expander definition.
459 2018-09-26  Indu Bhagat  <indu.bhagat@oracle.com>
461         PR gcov-profile/86957
462         * common.opt: New warning option -Wmissing-profile.
463         * coverage.c (get_coverage_counts): Add warning for missing .gcda file.
464         * doc/invoke.texi: Document -Wmissing-profile.
466 2018-09-26  Jim Wilson  <jimw@sifive.com>
468         * config/riscv/riscv.md (subsi3_extended2): Add J constraint.
469         (negdi2, negsi2, negsi2_extended, negsi2_extended2): New.
471 2018-09-26  Martin Sebor  <msebor@redhat.com>
473         * tree.c (zerop): Change return type to bool.
474         (integer_zerop, integer_onep, integer_each_onep): Same.
475         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
476         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
477         (real_onep, real_minus_onep, chain_index): Same.
478         (print_type_hash_statistics, type_list_equal): Same.
479         * tree.h (zerop): Same.
480         (zerop, integer_zerop, integer_onep, integer_each_onep): Same.
481         (integer_all_onesp, integer_minus_onep, integer_pow2p): Same.
482         (integer_nonzerop, integer_truep, tree_ctz, real_zerop): Same.
483         (real_onep, real_minus_onep, chain_index): Same.
484         (print_type_hash_statistics, type_list_equal): Same.
486 2018-09-26  Jim Wilson  <jimw@sifive.com>
488         * config/riscv/riscv.h (FUNCTION_ARG_REGNO_P): Fix comment.
490 2018-09-26  Jakub Jelinek  <jakub@redhat.com>
492         PR target/87414
493         * config/i386/i386.c: Include debug.h and dwarf2out.h.
494         (output_indirect_thunk): Emit DW_CFA_def_cfa_offset after the
495         call.
497 2018-09-25  Andrew Stubbs  <ams@codesourcery.com>
499         * builtins.c (get_builtin_sync_mem): Force address mode conversion.
501 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
503         * config/i386/i386.h (enum reg_class): Remove FP_TOP_SSE_REGS
504         and FP_SECOND_SSE_REGS.
505         (REG_CLASS_NAMES): Ditto.
506         (REG_CLASS_CONTENTS): Ditto.
507         * config/i386/i386.c (ix86_preferred_reload_class) Do not handle
508         FP_TOP_SSE_REGS and FP_SECOND_SSE_REGS classes.
509         (ix86_preferred_output_reload_class): Ditto.
510         * config/i386/i386.md (fix_trunc<mode>_i387_fisttp): Change "=&1f"
511         clobber constraint to "=&f".
512         (fix_truncdi_i387): Ditto.
513         (lrintxfdi2): Ditto.
514         (fistdi2_<rounding>): Ditto.
515         (fpremxf4_i387): Change "=u" constraint to "=f".
516         (fprem1xf4_i387): Ditto.
517         (sincosxf3): Ditto.
518         (fptanxf4_i387): Ditto.
519         (fxtractxf3_i387): Ditto.
520         (fscalexf4_i387): Ditto.
521         (atan2xf3): Change "u" constraint to "f".
522         (fyl2xxf3_i387): Ditto.
523         (fyl2xp1xf3_i387): Ditto.
525 2018-09-26  Uros Bizjak  <ubizjak@gmail.com>
527         PR target/87439
528         * config/i386/i386.h (NUM_MODES_FOR_MODE_SWITCHING): Update
529         for removed I387_MASK_PM entity.
532 2018-09-26  Jeff Law  <law@redhat.com>
533         Revert
534         2018-09-26  Alexey Neyman  <stilor@att.net>
536         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
537         headers are no longer pulled in by <isl/val.h>.
539 2018-09-26  Richard Biener  <rguenther@suse.de>
541         PR debug/87443
542         * dwarf2out.c (gen_lexical_block_die): Do not equate inline
543         or concrete instance DIE to the tree.  Create abstract origin
544         attributes also for concrete instances.
546 2018-09-26  Alexey Neyman  <stilor@att.net>
548         * graphite.h: Include <isl/id.h> and <isl/space.h>; these
549         headers are no longer pulled in by <isl/val.h>.
551 2018-09-26  Matthew Malcomson  <matthew.malcomson@arm.com>
553         * config/arm/arm.c (arm_split_compare_and_swap, arm_split_atomic_op):
554         Use new helper functions.
555         * config/arm/sync.md (atomic_load<mode>, atomic_store<mode>):
556         Use new helper functions.
557         * config/arm/aarch-common-protos.h (aarch_mm_needs_acquire,
558         aarch_mm_needs_release): New declarations.
559         * config/arm/aarch-common.c (aarch_mm_needs_acquire,
560         aarch_mm_needs_release): New.
562 2018-09-26  Eric Botcazou  <ebotcazou@adacore.com>
564         * config/arm/arm.c (arm_reorg): Skip Thumb reorg pass for thunks.
565         (arm32_output_mi_thunk): Deal with long calls.
567 2018-09-26  Richard Biener  <rguenther@suse.de>
569         PR debug/87428
570         PR debug/87362
571         * tree-inline.c (expand_call_inline): When the location
572         of the call is UNKNOWN_LOCATION use DECL_SOURCE_LOCATION
573         or BUILTINS_LOCATION for the BLOCK_SOURCE_LOCATION of
574         the inserted BLOCK to make inlined_function_outer_scope_p
575         recognize it.
576         * dwarf2out.c (add_call_src_coords_attributes): Do not add
577         coords for reserved locations.
579 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
581         * config/rs6000/rs6000.md (*movcc_internal1): Use set_attr_alternative.
582         (*call_indirect_nonlocal_sysv<mode>): Ditto.
583         (*call_value_indirect_nonlocal_sysv<mode>): Ditto.
584         (*sibcall_nonlocal_sysv<mode>): Ditto.
585         (*sibcall_value_nonlocal_sysv<mode>): Ditto.
586         (<bd>_<mode>): Ditto.
587         (<bd>tf_<mode>): Ditto.
589 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
591         * config/rs6000/altivec.md (*altivec_mov<mode>): Write the output
592         control string as a list of templates instead of as C code.
593         (*altivec_movti): Ditto.
594         * config/rs6000/darwin.md (movdf_low_di): Ditto.
596 2018-09-25  Jim Wilson  <jimw@sifive.com>
598         * config/riscv/riscv.c (riscv_split_symbol): Mark auipc label as weak
599         when target symbol is weak.
601 2018-09-25  Bernd Edlinger  <bernd.edlinger@hotmail.de>
603         PR c/87387
604         * builtins.c (unterminated_array): Simplify.
605         * expr.c (string_constant): Handle SSA_NAME.  Add more exceptions
606         where pointer arithmetic is safe.
608 2018-09-25  Segher Boessenkool  <segher@kernel.crashing.org>
610         PR target/86987
611         * config/rs6000/altivec.md (altivec_vspltb): Use
612         const_0_to_15_operand instead of u5bit_cint_operand.
613         (*altivec_vspltb_internal): Ditto.
614         (altivec_vspltb_direct): Ditto.
615         (altivec_vsplth): Use const_0_to_7_operand instead of
616         u5bit_cint_operand.
617         (*altivec_vsplth_internal): Ditto.
618         (altivec_vsplth_direct): Ditto.
619         (altivec_vspltw): Use const_0_to_3_operand instead of
620         u5bit_cint_operand.
621         (*altivec_vspltw_internal): Ditto.
622         (altivec_vspltw_direct): Ditto.
623         (altivec_vspltsf): Ditto.
624         (*altivec_vspltsf_internal): Ditto.
625         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Handle the
626         various splats with the proper size immediate.  Reorder the various
627         cases by ascending size of immediate, and put all such together.
629 2018-09-25  Richard Biener  <rguenther@suse.de>
631         PR debug/83941
632         * dwarf2out.c (add_AT_external_die_ref): Remove now redundant
633         GC-ification.
634         (maybe_create_die_with_external_ref): Do not create
635         DW_TAG_imported_unit here.
636         (add_abstract_origin_attribute): Handle external BLOCK refs.
637         (dwarf2out_abstract_function): Simplify LTO case.
638         (dwarf2out_early_finish): Create DW_TAG_imported_unit explicitely
639         rather than using maybe_create_die_with_external_ref.
641 2018-09-25  Uros Bizjak  <ubizjak@gmail.com>
643         PR target/71278
644         * config/i386/i386.md (frndintxf2_mask_pm): Remove.
645         (frndintxf2_mask_pm_i387): Ditto.
646         (nearbyintxf2): Rewrite expander pattern to match rintxf2.
647         Enable for !flag_trapping_math.
648         (nearbyint<mode>2): Enable x87 modes for !flag_trapping_math.
649         Enable SSE modes for TARGET_SSE4_1 and expand them with round insn.
650         Change operand 1 predicate to nonimmediate_operand.
651         (attr "i387_cw"): Remove mask_pm.
652         * config/i386/i386.h (enum ix86_stack_slot): Remove SLOT_CW_MASK_PM.
653         (enum ix86_entity): Remove I387_MASK_PM.
654         * config/i386/i386.c (ix86_i387_mode_needed): Do not
655         handle I387_MASK_PM.
656         (ix86_mode_needed): Ditto.
657         (ix86_mode_after): Ditto.
658         (ix86_mode_entry): Ditto.
659         (ix86_mode_exit): Ditto.
660         (emit_i387_cw_initialization): Do not handle I387_CW_MASK_PM.
662 2018-09-25  Jakub Jelinek  <jakub@redhat.com>
664         * vr-values.c (vr_values::vr_values): Initialize to_remove_edges and
665         to_update_switch_stmts to vNULL instead of calling create on them
666         immediately.
668 2018-09-25  Richard Biener  <rguenther@suse.de>
670         PR tree-optimization/87402
671         * tree-ssa-sccvn.c (SSA_VISITED): Remove unused function.
672         (visit_phi): Re-instantiate handling of supposed to be VARYING
673         but non-VARYING backedge value.
675 2018-09-25  Richard Biener  <rguenther@suse.de>
677         PR debug/83941
678         * dwarf2out.c (struct sym_off_pair): New.
679         (external_die_map): New global.
680         (lookup_decl_die): When in LTO create DIEs lazily from the
681         external_die_map.
682         (lookup_block_die): New function, create DIEs lazily in LTO.
683         (equate_block_to_die): New function.
684         (dwarf2out_die_ref_for_decl): During WPA get the association
685         from the external DIE map.
686         (dwarf2out_register_external_die): Record mapping into the
687         external DIE map.
688         (maybe_create_die_with_external_ref): New function split out from
689         DIE generation part of old dwarf2out_register_external_die.
690         (add_abstract_origin_attribute): Do not return the DIE.  When
691         in LTO reference externals directly.
692         (dwarf2out_abstract_function): When in LTO ignore calls for
693         decls with external DIEs (already present abstract instances).
694         (gen_call_site_die): Adjust.
695         (add_high_low_attributes): Likewise.
696         (gen_lexical_block_die): Likewise.
697         (gen_inlined_subroutine_die): Likewie.
698         (gen_block_die): Likewise.
699         (dwarf2out_inline_entry): Likewise.
700         (dwarf2out_early_finish): In LTRANS phase create DW_TAG_imported_unit
701         DIEs.
703 2018-09-25  Martin Liska  <mliska@suse.cz>
705         * ipa-fnsummary.c (estimate_node_size_and_time): Scale by two
706         integers and not by a float value.
708 2018-09-25  Martin Liska  <mliska@suse.cz>
710         PR fortran/87394
711         * dbgcnt.c (dbg_cnt_process_single_pair): Return false
712         instead of NULL.
713         * dumpfile.c (dump_enable_all): Remove extra parenthesis.
714         * gcov-tool.c: Declare the function with ATTRIBUTE_NORETURN.
715         * godump.c (go_format_type): Remove extra parenthesis.
717 2018-09-25  Martin Liska  <mliska@suse.cz>
719         * alias.c (set_dest_equal_p): Remove unused function.
720         * config/i386/i386.c (def_builtin_pure2): Likewise.
721         * diagnostic-show-locus.c (class layout): Remove
722         unused field.
723         (layout::layout): Likewise here.
724         * dump-context.h (class temp_dump_context): Likewise.
725         * dwarf2out.c (add_AT_fde_ref): Remove unused function.
726         (add_AT_loclistsptr): Likewise.
727         (add_AT_offset): Likewise.
728         (get_AT_hi_pc): Likewise.
729         (is_comdat_die): Likewise.
730         (type_is_enum): Likewise.
731         (ceiling): Likewise.
732         (add_AT_vms_delta): Likewise.
733         (is_class_die): Likewise.
734         * edit-context.c (class line_event): Remove unused field.
735         * graphite-sese-to-poly.c (tree_int_to_gmp): Remove
736         unused function.
737         * ipa-cp.c (ipa_get_vr_lat): Likewise.
738         * lra-constraints.c (ok_for_index_p_nonstrict): Likewise.
739         (ok_for_base_p_nonstrict): Likewise.
740         * tree-chrec.c (is_not_constant_evolution): Likewise.
741         (chrec_fold_poly_cst): Likewise.
742         * tree-if-conv.c (has_pred_critical_p): Likewise.
743         * tree-ssa-coalesce.c (print_exprs): Likewise.
744         * tree-ssa-pre.c (bitmap_set_contains_expr): Likewise.
745         * tree-ssa-uninit.c (is_and_or_or_p): Likewise.
746         * tree-vrp.c (value_ranges_intersect_p): Likewise.
747         (value_range_nonnegative_p): Likewise.
749 2018-09-25  Martin Liska  <mliska@suse.cz>
751         * config/powerpcspe/powerpcspe.c (rs6000_output_function_epilogue):
752         Do not handle "GNU Pascal".
753         * config/rs6000/rs6000.c (rs6000_output_function_epilogue):
754         Likewise.
755         * config/sparc/sparc.c (sparc_pass_by_reference): Remove Pascal
756         from documentation. Likewise.
757         * dbxout.c (dbxout_range_type): Likewise.
758         * doc/cpp.texi: Likewise.
759         * doc/extend.texi: Likewise.
760         * doc/frontends.texi: Likewise.
761         * doc/invoke.texi: Remove Pascal entry.
762         * tree.def (CLEANUP_POINT_EXPR): Likewise.
763         * doc/rtl.texi (MODE_FUNCTION): Remove not used entry.
765 2018-09-25  Martin Liska  <mliska@suse.cz>
767         PR middle-end/86078
768         * doc/invoke.texi: Document all parameters and remove default
769         of the parameters.
771 2018-09-25  Ilya Leoshkevich  <iii@linux.ibm.com>
773         PR bootstrap/87417
774         * rtl.c (rtx_code_size): Take into account that EQ_ATTR_ALT
775         contains HOST_WIDE_INTs when computing its size.
777 2018-09-24  Jim Wilson  <jimw@sifive.com>
779         PR target/87391
780         * config/riscv/riscv.h (STACK_BOUNDARY): Test riscv_abi == ABI_ILP32E
781         not TARGET_RVE.
782         (ABI_STACK_BOUNDARY, MAX_ARGS_IN_REGISTERS): Likewise.
784 2018-09-24  Andrew Pinski  <apinski@marvell.com>
786         *  config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Don't
787         access prev before checking it for NULLness in the
788         AARCH64_FUSE_CMP_BRANCH case.
790 2018-09-24  H.J. Lu  <hongjiu.lu@intel.com>
792         PR target/82699
793         * config/i386/i386.c (rest_of_insert_endbranch): Set
794         endbr_queued_at_entrance to true and don't insert ENDBR if
795         x86_function_profiler will be called.
796         (x86_function_profiler): Insert ENDBR if endbr_queued_at_entrance
797         is true.
798         * config/i386/i386.h (machine_function): Add
799         endbr_queued_at_entrance.
801 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
803         * genattrtab.c (mk_attr_alt): Use alternative_mask.
804         (attr_rtx_1): Adjust caching to match the new EQ_ATTR_ALT field
805         types.
806         (check_attr_test): Use alternative_mask.
807         (get_attr_value): Likewise.
808         (compute_alternative_mask): Use alternative_mask and XWINT.
809         (make_alternative_compare): Use alternative_mask.
810         (attr_alt_subset_p): Use XWINT.
811         (attr_alt_subset_of_compl_p): Likewise.
812         (attr_alt_intersection): Use alternative_mask and XWINT.
813         (attr_alt_union): Likewise.
814         (attr_alt_complement): Use HOST_WIDE_INT and XWINT.
815         (mk_attr_alt): Use alternative_mask and HOST_WIDE_INT.
816         (simplify_test_exp): Use alternative_mask and XWINT.
817         (write_test_expr): Use alternative_mask and XWINT, adjust bit
818         number calculation to support 64 bits.  Generate code that
819         checks 64-bit masks.
820         (main): Use alternative_mask.
821         * rtl.def (EQ_ATTR_ALT): Change field types from ii to ww.
823 2018-09-24  Ilya Leoshkevich  <iii@linux.ibm.com>
825         PR target/80080
826         * config/s390/s390.c (s390_emit_epilogue): Do not use PARALLEL
827         RETURN+USE when returning via %r14.
829 2018-09-24  Martin Liska  <mliska@suse.cz>
831         * gcov.c (output_lines): Print colorization legend
832         for both flag_use_colors and flag_use_hotness_colors.
833         Reword the help.
835 2018-09-24  Martin Liska  <mliska@suse.cz>
837         * coverage.c (get_coverage_counts): Use warning_at
838         with current_function_decl location. Use %qD in warning
839         message.
841 2018-09-24  Martin Liska  <mliska@suse.cz>
843         * memory-block.h (memory_block_pool::release): Annotate with
844         valgrind that the memory is not accessible.
846 2018-09-24  Martin Liska  <mliska@suse.cz>
848         PR sanitizer/85774
849         * asan.c: Make asan_handled_variables extern.
850         * asan.h: Likewise.
851         * cfgexpand.c (expand_stack_vars): Make sure
852         a representative is unpoison if another
853         variable in the partition is handled by
854         use-after-scope sanitization.
856 2018-09-24  Richard Biener  <rguenther@suse.de>
858         PR tree-optimization/63155
859         * tree-ssa-propagate.c (add_ssa_edge): Avoid adding PHIs to
860         the worklist when the edge of the respective argument isn't
861         executable.
863 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
865         * config/i386/i386.h (enum reg_class): Rename MASK_REGS to
866         ALL_MASK_REGS and MASK_EVEX_REGS to MASK_REGS.
867         (MASK_CLASS_P): Update for rename.
868         (MAYBE_MASK_CLASS_P): Ditto.
869         (REG_CLASS_NAMES): Update.
870         (REG_CLASS_CONTENT): Update.
871         * config/i386/i386.c (regclass_map): Update for MASK_REG
872         and ALL_MASK_REGS rename.
873         * config/i386/constraints.md (Yk): Update for rename.
874         (k): Ditto.
876 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
878         * config/i386/i386.h (enum reg_class): Remove
879         EVEX_SSE_REGS and MOD4_SSE_REGS.
880         (REG_CLASS_NAMES): Update.
881         (REG_CLASS_CONTENT): Update.
882         * config/i386/i386.c (regclass_map): Declare AVX-512 SSE
883         registers as ALL_SSE_REGS.
884         (ix86_additional_allocno_class_p): Remove.
885         (TARGET_ADDITIONAL_ALLOCNO_CLASS_P): Remove.
886         (ix86_register_priority): Lower priority of EVEX SSE registers.
887         Use IN_RANGE macro where appropriate.
888         (ix86_hard_regno_mode_ok): Merge AVX-5124FMAPS and
889         AVX-5124VNNIW checks.
890         (ix86_modes_tieable_p): Tie 512-bit SSE modes.
891         * config/i386/sse.md (avx5124fmaddps_4fmaddps)
892         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddps_maskz)
893         (avx5124fmaddps_4fmaddss, avx5124fmaddps_4fmaddss_mask)
894         (avx5124fmaddps_4fmaddss_maskz, avx5124fmaddps_4fnmaddps)
895         (avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddps_maskz)
896         (avx5124fmaddps_4fnmaddss, avx5124fmaddps_4fnmaddss_mask)
897         (avx5124fmaddps_4fnmaddss_maskz, avx5124vnniw_vp4dpwssd)
898         (avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssd_maskz)
899         (avx5124vnniw_vp4dpwssds, avx5124vnniw_vp4dpwssds_mask)
900         (avx5124vnniw_vp4dpwssds_maskz): Use "v" instead of "Yh" constraint.
901         * config/i386/constraints.md (Yh): Remove.
903 2018-09-23   Uros Bizjak  <ubizjak@gmail.com>
905         * config/i386/i386.c (regclass_map): Declare integer REX registers
906         as GENERAL_REGS.
908 2018-09-23  Gerald Pfeifer  <gerald@pfeifer.com>
910         * doc/service.texi (Service): Switch the fsf.org link to https.
912 2018-09-22  Chung-Ju Wu  <jasonwucj@gmail.com>
914         PR target/86798
915         * config/nds32/nds32.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
916         Define to speculation_safe_value_not_needed.
918 2018-09-21  Florian Weimer  <fweimer@redhat.com>
920         PR middle-end/81035
921         * doc/extend.texi (Common Function Attributes): Mention that
922         noreturn suppresses tail call optimization.
924 2018-09-21  Jeff Law  <law@redhat.com>
926         * gimple-ssa-evrp.c (evrp_dom_walker::cleanup): Call
927         vr_values::cleanup_edges_and_switches.
928         * tree-vrp.c (to_remove_edges, to_update_switch_stmts): Moved into
929         vr_values class.
930         (identify_jump_threads): Remove EDGE_IGNORE handling.
931         (execute_vrp): Move handling of to_remove_edges and
932         to_update_switch_stmts into vr_values class member functions.
933         * tree-vrp.h (switch_update, to_remove_edges): Remove declarations.
934         (to_update_switch_stmts): Likewise.
935         * vr-values.c: Include cfghooks.h.
936         (vr_values::vr_values): Initialize to_remove_edges and
937         to_update_switch_stmts.
938         (vr_values::~vr_values): Verify to_remove_edges and
939         to_update_switch_stmts are empty.
940         (vr_values::simplify_switch_using_ranges): Set EDGE_IGNORE as needed.
941         (vr_values::cleanup_edges_and_switches): New member function.
942         * vr-values.h (vr_values): Add cleanup_edges_and_switches member
943         function.  Add new data members.
945 2018-09-21  David Malcolm  <dmalcolm@redhat.com>
947         PR tree-optimization/87309
948         * dumpfile.c (dump_context::begin_scope): Filter the dump_loc
949         calls with pflags and alt_flags.
950         (selftest::test_capture_of_dump_calls): Add test of interaction of
951         MSG_OPTIMIZED_LOCATIONS with AUTO_DUMP_SCOPE.
953 2018-09-21  Olivier Hainque  <hainque@adacore.com>
955         * config.gcc: Factorize and comment inclusion of vxworks-dummy.h.
957 2018-09-21  Olivier Hainque  <hainque@adacore.com>
959         * config/vxworks.h (CLEAR_INSN_CACHE): #define to 1.
961 2018-09-21  Olivier Hainque  <hainque@adacore.com>
963         * config/vxworks.h (VXWORKS_LIBGCC_SPEC): Remove -lc_internal.
964         Merge block comment with the one ahead of VXWORKS_LIBS_RTP. Then:
965         (VXWORKS_LIBS_RTP): Minor reordering.
967 2018-09-21  Olivier Hainque  <hainque@adacore.com>
969         * config/vxworks.h (STARTFILE_PREFIX_SPEC): Define.
970         (VXWORKS_LIBS_DIR_RTP): Remove definition and use.
972 2018-09-21  Olivier Hainque  <hainque@adacore.com>
974         * config/vxworks.h (SIZE_TYPE): Account for TARGET_VXWORKS64.
975         (PTRDIFF_TYPE): Likewise.
977 2018-09-21  Olivier Hainque  <hainque@adacore.com>
979         * config.gcc: Enforce def of TARGET_VXWORKS64 to 1 from
980         triplet, similar to support for VxWorks7.
981         * config/vxworks-dummy.h: Provide a default definition
982         of TARGET_VXWORKS64 to 0.
984 2018-09-21  Olivier Hainque  <hainque@adacore.com>
986         * config/vxworks.h (TARGET_VXWORKS7): Move default definition ...
987         * config/vxworks-dummy.h: here.
989 2018-09-21  Olivier Hainque  <hainque@adacore.com>
991         * config.gcc: Prepend vxworks-dummy.h to tm_file for powerpc*
993 2018-09-21  Shaokun Zhang  <zhangshaokun@hisilicon.com>
994             Bo Zhou  <zbo.zhou@hisilicon.com>
996         * config/aarch64/aarch64-cores.def (tsv110): New CPU.
997         * config/aarch64/aarch64-tune.md: Regenerated.
998         * doc/invoke.texi (AArch64 Options/-mtune): Add "tsv110".
999         * config/aarch64/aarch64.c (tsv110_tunings): New tuning table.
1000         * config/aarch64/aarch64-cost-tables.h: Add "tsv110" extra costs.
1002 2018-09-21  Andrew Stubbs  <ams@codesourcery.com>
1003             Julian Brown  <julian@codesourcery.com>
1005         * builtins.c (get_builtin_sync_mem): Handle address spaces.
1007 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
1009         * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Return false
1010         if the call takes a static chain.
1012 2018-09-21  Martin Liska  <mliska@suse.cz>
1014         * auto-profile.c (autofdo_source_profile::read): Do not
1015         set sum_all.
1016         (read_profile): Do not add working sets.
1017         (read_autofdo_file): Remove sum_all.
1018         (afdo_callsite_hot_enough_for_early_inline): Remove const
1019         qualifier.
1020         * coverage.c (struct counts_entry): Remove gcov_summary.
1021         (read_counts_file): Read new GCOV_TAG_OBJECT_SUMMARY,
1022         do not support GCOV_TAG_PROGRAM_SUMMARY.
1023         (get_coverage_counts): Remove summary and expected
1024         arguments.
1025         * coverage.h (get_coverage_counts): Likewise.
1026         * doc/gcov-dump.texi: Remove -w option.
1027         * gcov-dump.c (dump_working_sets): Remove.
1028         (main): Do not support '-w' option.
1029         (print_usage): Likewise.
1030         (tag_summary): Likewise.
1031         * gcov-io.c (gcov_write_summary): Do not dump
1032         histogram.
1033         (gcov_read_summary): Likewise.
1034         (gcov_histo_index): Remove.
1035         (gcov_histogram_merge): Likewise.
1036         (compute_working_sets): Likewise.
1037         * gcov-io.h (GCOV_TAG_OBJECT_SUMMARY): Mark
1038         it not obsolete.
1039         (GCOV_TAG_PROGRAM_SUMMARY): Mark it obsolete.
1040         (GCOV_TAG_SUMMARY_LENGTH): Adjust.
1041         (GCOV_HISTOGRAM_SIZE): Remove.
1042         (GCOV_HISTOGRAM_BITVECTOR_SIZE): Likewise.
1043         (struct gcov_summary): Simplify rapidly just
1044         to runs and sum_max fields.
1045         (gcov_histo_index): Remove.
1046         (NUM_GCOV_WORKING_SETS): Likewise.
1047         (compute_working_sets): Likewise.
1048         * gcov-tool.c (print_overlap_usage_message): Remove
1049         trailing empty line.
1050         * gcov.c (read_count_file): Read GCOV_TAG_OBJECT_SUMMARY.
1051         (output_lines): Remove program related line.
1052         * ipa-profile.c (ipa_profile): Do not consider GCOV histogram.
1053         * lto-cgraph.c (output_profile_summary): Do not stream GCOV
1054         histogram.
1055         (input_profile_summary): Do not read it.
1056         (merge_profile_summaries): And do not merge it.
1057         (input_symtab): Do not call removed function.
1058         * modulo-sched.c (sms_schedule): Do not print sum_max.
1059         * params.def (HOT_BB_COUNT_FRACTION): Reincarnate param that was
1060         removed when histogram method was invented.
1061         (HOT_BB_COUNT_WS_PERMILLE): Mention that it's used only in LTO
1062         mode.
1063         * postreload-gcse.c (eliminate_partially_redundant_load): Fix
1064         GCOV coding style.
1065         * predict.c (get_hot_bb_threshold): Use HOT_BB_COUNT_FRACTION
1066         and dump selected value.
1067         * profile.c (add_working_set): Remove.
1068         (get_working_sets): Likewise.
1069         (find_working_set): Likewise.
1070         (get_exec_counts): Do not work with working sets.
1071         (read_profile_edge_counts): Do not inform as sum_max is removed.
1072         (compute_branch_probabilities): Likewise.
1073         (compute_value_histograms): Remove argument for call of
1074         get_coverage_counts.
1075         * profile.h: Do not make gcov_summary const.
1077 2018-09-21  Monk Chiang  <sh.chiang04@gmail.com>
1079         * config.gcc (nds32*-*-*): Set TARGET_DEFAULT_TLSDESC_TRAMPOLINE=0.
1081 2018-09-21  Eric Botcazou  <ebotcazou@adacore.com>
1083         PR tree-optimization/86990
1084         * gimple-ssa-store-merging.c (imm_store_chain_info:coalesce_immediate):
1085         Check that the entire merged store group is made of constants only for
1086         overlapping stores.
1088 2018-09-20  Allan Sandfeld Jensen  <allan.jensen@qt.io>
1090         * gcc.c (LINK_COMMAND_SPEC): Handle -r like -nostdlib.
1091         (VTABLE_VERIFICATION_SPEC): Likewise.
1092         (SANITIZER_EARLY_SPEC): Likewise.
1093         (SANITIZER_SPEC): Likewise.
1094         * config/darwin.h (LINK_COMMAND_SPEC): Likewise.
1095         * doc/invoke.texi (Link Options): Document -r.
1097 2018-09-20  Richard Biener <rguenther@suse.de>
1099         PR middle-end/87054
1100         * gimplify.c (gimplify_expr): Retain alignment of
1101         addressable lvalue in dereference.
1103 2018-09-20  Alexandre Oliva <aoliva@redhat.com>
1105         PR bootstrap/87013
1106         * configure.ac: Check for .loc is_stmt support.
1107         * configure, config.in: Rebuilt.
1108         * dwarf2out.c (dwarf2out_source_line): Skip is_stmt
1109         if not supported.
1111 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
1113         * config/rs6000/rs6000.opt (misel=no, misel=yes): Delete.
1114         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -misel=yes and
1115         -misel=no.
1117 2018-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
1119         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
1120         VECTOR_OTHER.
1121         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete
1122         case VECTOR_OTHER.
1124 2018-09-20  Marek Polacek  <polacek@redhat.com>
1126         * doc/invoke.texi: Add -Wno-init-list-lifetime to C++ Language Options.
1128 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
1130         PR tree-optimization/87288
1131         * tree-vect-loop.c (vect_analyze_loop_2): Take PEELING_FOR_GAPS
1132         into account when determining PEELING_FOR_NITERS.
1134 2018-09-20  Richard Sandiford  <richard.sandiford@arm.com>
1136         PR tree-optimization/86877
1137         * tree-vect-loop.c (vect_analyze_loop_2): Call
1138         vect_verify_datarefs_alignment.
1140 2018-09-19  Marek Polacek  <polacek@redhat.com>
1142         * doc/invoke.texi: Document -Wclass-conversion.
1144 2018-09-19  John David Anglin  <danglin@gcc.gnu.org>
1146         * config/pa/pa.c (pa_adjust_priority): Delete.
1147         (TARGET_SCHED_ADJUST_PRIORITY): Delete define.
1149         * config/pa/pa.md (atomic_storeqi): Restore deleted expander.
1150         (atomic_storehi): Likewise.
1151         (atomic_storesi): Likewise.
1152         (atomic_loaddi): Restore compare and swap exchange loop code.
1154 2018-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
1156         PR rtl-optimization/86902
1157         * combine.c (try_combine): When changing the CC mode used, don't change
1158         an unrelated mode in other_insn to that new CC mode.
1160 2018-09-19  David Malcolm  <dmalcolm@redhat.com>
1162         * tree-data-ref.c (runtime_alias_check_p): Use formatted printing
1163         with %T in place of calls to dump_generic_expr.
1164         (prune_runtime_alias_test_list): Likewise.
1165         (create_runtime_alias_checks): Likewise.
1166         * tree-vect-data-refs.c (vect_check_nonzero_value): Likewise.
1167         (vect_analyze_data_ref_dependence): Likewise.
1168         (vect_slp_analyze_data_ref_dependence): Likewise.
1169         (vect_record_base_alignment): Likewise.  Use %G in place of call
1170         to dump_gimple_stmt.
1171         (vect_compute_data_ref_alignment): Likewise.
1172         (verify_data_ref_alignment): Likewise.
1173         (vect_find_same_alignment_drs): Likewise.
1174         (vect_analyze_group_access_1): Likewise.
1175         (vect_analyze_data_ref_accesses): Likewise.
1176         (dependence_distance_ge_vf): Likewise.
1177         (dump_lower_bound): Likewise.
1178         (vect_prune_runtime_alias_test_list): Likewise.
1179         (vect_find_stmt_data_reference): Likewise.
1180         (vect_analyze_data_refs): Likewise.
1181         (vect_create_addr_base_for_vector_ref): Likewise.
1182         (vect_create_data_ref_ptr): Likewise.
1183         * tree-vect-loop-manip.c (vect_set_loop_condition): Likewise.
1184         (vect_can_advance_ivs_p): Likewise.
1185         (vect_update_ivs_after_vectorizer): Likewise.
1186         (vect_gen_prolog_loop_niters): Likewise.
1187         (vect_prepare_for_masked_peels): Likewise.
1188         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
1189         (vect_determine_vectorization_factor): Likewise.
1190         (vect_is_simple_iv_evolution): Likewise.
1191         (vect_analyze_scalar_cycles_1): Likewise.
1192         (vect_analyze_loop_operations): Likewise.
1193         (report_vect_op): Likewise.
1194         (vect_is_slp_reduction): Likewise.
1195         (check_reduction_path): Likewise.
1196         (vect_is_simple_reduction): Likewise.
1197         (vect_create_epilog_for_reduction): Likewise.
1198         (vect_finalize_reduction:): Likewise.
1199         (vectorizable_induction): Likewise.
1200         (vect_transform_loop_stmt): Likewise.
1201         (vect_transform_loop): Likewise.
1202         (optimize_mask_stores): Likewise.
1203         * tree-vect-patterns.c (vect_pattern_detected): Likewise.
1204         (vect_split_statement): Likewise.
1205         (vect_recog_over_widening_pattern): Likewise.
1206         (vect_recog_average_pattern): Likewise.
1207         (vect_determine_min_output_precision_1): Likewise.
1208         (vect_determine_precisions_from_range): Likewise.
1209         (vect_determine_precisions_from_users): Likewise.
1210         (vect_mark_pattern_stmts): Likewise.
1211         (vect_pattern_recog_1): Likewise.
1212         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
1213         (vect_record_max_nunits): Likewise.
1214         (vect_build_slp_tree_1): Likewise.
1215         (vect_build_slp_tree_2): Likewise.
1216         (vect_print_slp_tree): Likewise.
1217         (vect_analyze_slp_instance): Likewise.
1218         (vect_detect_hybrid_slp_stmts): Likewise.
1219         (vect_detect_hybrid_slp_1): Likewise.
1220         (vect_slp_analyze_operations): Likewise.
1221         (vect_slp_analyze_bb_1): Likewise.
1222         (vect_transform_slp_perm_load): Likewise.
1223         (vect_schedule_slp_instance): Likewise.
1224         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
1225         (vect_mark_stmts_to_be_vectorized): Likewise.
1226         (vect_init_vector_1): Likewise.
1227         (vect_get_vec_def_for_operand): Likewise.
1228         (vect_finish_stmt_generation_1): Likewise.
1229         (vect_check_load_store_mask): Likewise.
1230         (vectorizable_call): Likewise.
1231         (vectorizable_conversion): Likewise.
1232         (vectorizable_operation): Likewise.
1233         (vectorizable_load): Likewise.
1234         (vect_analyze_stmt): Likewise.
1235         (vect_is_simple_use): Likewise.
1236         (vect_get_vector_types_for_stmt): Likewise.
1237         (vect_get_mask_type_for_stmt): Likewise.
1238         * tree-vectorizer.c (increase_alignment): Likewise.
1240 2018-09-19  Andrew Stubbs  <ams@codesourcery.com>
1242         * doc/rtl.texi: Adjust vec_select description.
1243         * simplify-rtx.c (simplify_binary_operation_1): Allow VEC_SELECT to use
1244         non-constant selectors.
1246 2018-09-19  Matthew Malcomson  <matthew.malcomson@arm.com>
1248         * config/aarch64/aarch64-protos.h
1249         (aarch64_offset_9bit_signed_unscaled_p): New declaration.
1250         * config/aarch64/aarch64.md (arches): New "rcpc8_4" attribute value.
1251         (arch_enabled): Add check for "rcpc8_4" attribute value of "arch".
1252         * config/aarch64/aarch64.h (AARCH64_FL_RCPC8_4): New bitfield.
1253         (AARCH64_FL_FOR_ARCH8_4): Include AARCH64_FL_RCPC8_4.
1254         (AARCH64_FL_PROFILE): Move index so flags are ordered.
1255         (AARCH64_ISA_RCPC8_4): New flag.
1256         * config/aarch64/aarch64.c (offset_9bit_signed_unscaled_p): Renamed
1257         to aarch64_offset_9bit_signed_unscaled_p.
1258         * config/aarch64/atomics.md (atomic_store<mode>): Allow offset
1259         and use stlur.
1260         * config/aarch64/constraints.md (Ust): New constraint.
1261         * config/aarch64/predicates.md.
1262         (aarch64_9bit_offset_memory_operand): New predicate.
1263         (aarch64_rcpc_memory_operand): New predicate.
1265 2018-09-19  Eric Botcazou  <ebotcazou@adacore.com>
1267         PR rtl-optimization/87361
1268         * rtlanal.c (nonzero_bits1): Revert accidental change.
1270 2018-09-19  Richard Biener  <rguenther@suse.de>
1272         PR tree-optimization/87349
1273         PR tree-optimization/87342
1274         * tree-ssa-sccvn.c (do_rpo_vn): Iterate max_rpo computation.
1276 2018-09-18  Marek Polacek  <polacek@redhat.com>
1278         P1064R0 - Allowing Virtual Function Calls in Constant Expressions
1279         * gimple-fold.c (gimple_get_virt_method_for_vtable): Adjust assert.
1281 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
1283         * config/rs6000/rs6000.md: Remove old "Cygnus sibcall" comment.
1285 2018-09-18  Segher Boessenkool  <segher@kernel.crashing.org>
1287         PR rtl-optimization/86882
1288         * rtlanal.c (reg_overlap_mentioned_p): Handle CLOBBER.
1290 2018-09-18  Uros Bizjak  <ubizjak@gmail.com>
1292         * config/i386/i386.md (*<code>extend<mode>xf2): Macroize insn from
1293         *<code>extendsfxf2 and *<code>extenddfxf2 using MODEF mode iterator.
1295 2018-09-18  Jonathan Wakely  <jwakely@redhat.com>
1297         PR other/87353
1298         * doc/invoke.texi (Link Options): Fix formatting and grammar.
1300 2018-09-18  Richard Biener  <rguenther@suse.de>
1302         PR middle-end/63155
1303         * tree-ssa-coalesce.c (tree_int_map_hasher): Remove.
1304         (compute_samebase_partition_bases): Likewise.
1305         (coalesce_ssa_name): Always use compute_optimized_partition_bases.
1306         (gimple_can_coalesce_p): Simplify.
1308 2018-09-18  Hans-Peter Nilsson  <hp@bitrange.com>
1310         Handle a library implementation of ffs calling __builtin_ffs.
1311         * config/mmix/mmix.c (TARGET_INIT_LIBFUNCS): Override with...
1312         (mmix_init_libfuncs): New function: make __builtin_ffs expand
1313         to __ffsdi2.
1315 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
1317         * diagnostic-show-locus.c (class layout_range): Add field
1318         "m_original_idx".
1319         (layout_range::layout_range): Add "original_idx" param and use it
1320         to initialize new field.
1321         (make_range): Use 0 for original_idx.
1322         (layout::layout): Pass in index to calls to
1323         maybe_add_location_range.
1324         (layout::maybe_add_location_range): Add param "original_idx" and
1325         pass it on to layout_range.
1326         (layout::print_any_labels): Pass on range->m_original_idx to
1327         get_text call.
1328         (gcc_rich_location::add_location_if_nearby): Use 0 for
1329         original_idx.
1330         * gcc-rich-location.h (text_range_label::get_text): Update for new
1331         param.
1332         (range_label_for_type_mismatch::get_text): Likewise.
1334 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
1336         * config/i386/i386.c (ix86_emit_i387_log1p): Emit fldln2 earlier.
1338 2018-09-17  David Malcolm  <dmalcolm@redhat.com>
1340         * gimple-ssa-sprintf.c (fmtwarn): Update for introduction of
1341         format_string_diagnostic_t.
1342         (fmtwarn_n): Likewise.
1343         * substring-locations.c
1344         (format_string_diagnostic_t::format_string_diagnostic_t) New ctor.
1345         (format_warning_n_va): Convert to...
1346         (format_string_diagnostic_t::emit_warning_n_va): ...this.
1347         (format_warning_va): Convert to...
1348         (format_string_diagnostic_t::emit_warning_va): ...this.
1349         (format_warning_at_substring): Convert to...
1350         (format_string_diagnostic_t::emit_warning): ...this.
1351         (format_warning_at_substring_n): Convert to...
1352         (format_string_diagnostic_t::emit_warning_n): ...this.
1353         * substring-locations.h (class format_string_diagnostic_t): New
1354         class.
1355         (format_warning_va): Convert to
1356         format_string_diagnostic_t::emit_warning_va.
1357         (format_warning_n_va): Convert to
1358         format_string_diagnostic_t::emit_warning_n_va.
1359         (format_warning_at_substring): Convert to
1360         format_string_diagnostic_t::emit_warning.
1361         (format_warning_at_substring_n): Convert to
1362         format_string_diagnostic_t::emit_warning_n.
1364 2018-09-17  Cesar Philippidis  <cesar@codesourcery.com>
1365             Bernd Schmidt <bernds_cb1@t-online.de>
1367         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with
1368         SImode args.
1370 2018-09-17  Uros Bizjak  <ubizjak@gmail.com>
1372         * config/i386/i386.md (truncxf<mode>2_i387_noop_unspec): Change
1373         operand 0 predicate to nonimmediate operand.
1374         (rint<mode>2_frndint): Remove insn pattern.
1375         (rint<mode>2): Change operand 1 predicate to general_operand.
1376         Extend operand 1 to XFmode and generate rintxf2 insn.
1377         (frndintxf2_<rounding>): Rename from frndint<mode>2_<rounding>.
1378         Do not use X87MODEF mode macro.
1379         (frndintxf2_<rounding>_i387): Rename from
1380         frndint<mode>2_<rounding>_i387.  Do not use X87MODEF mode macro.
1381         (<rounding_insn><mode>2): For non-SSE modes, extend operand 1
1382         to XFmode and generate significandxf3 insn.
1384 2018-09-17  Richard Biener  <rguenther@suse.de>
1386         PR tree-optimization/87328
1387         * tree-ssa-sccvn.c (process_bb): Remove assertion about not
1388         visiting unexecutable backedges when not iterating.
1389         (do_rpo_vn): Mark all edges not executable even when not
1390         iterating.
1392 2018-09-17  Martin Jambor  <mjambor@suse.cz>
1394         PR c/63886
1395         * doc/invoke.texi (Warning Options): Likewise.
1397 2018-09-17  Richard Biener  <rguenther@suse.de>
1399         PR tree-optimization/87301
1400         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup): Properly
1401         clean EH info from leftover copy assignments.
1403 2018-09-17  Martin Liska  <mliska@suse.cz>
1405         PR gcov-profile/85871
1406         * gcov.c (output_intermediate_file): Fix out of bounds
1407         access.
1409 2018-09-17  Vineet Gupta  <vgupta@synopsys.com>
1411         * config/arc/arc.c: Object attributes for core4 not reflected
1412         correctly.
1413         * config/arc/arc.h: Don't restrict DBNZ to core3 (core4 includes
1414         core3).
1416 2018-09-17  Alexey Brodkin  <abrodkin@synopsys.com>
1418         * config/arc/linux.h (LINK_EH_SPEC): Add missing space.
1420 2018-09-17  Martin Liska  <mliska@suse.cz>
1422         * doc/gcov.texi: Document new option --use-hotness-colors.
1423         * gcov.c (struct source_info): Declare new field.
1424         (source_info::source_info): Set default for maximum_count.
1425         (print_usage): Add new -q option.
1426         (process_args): Process it.
1427         (accumulate_line_info): Save src->maximum_count.
1428         (output_line_beginning): Make color line number if
1429         flag_use_hotness_colors is set.
1430         (output_line_details): Pass default argument value.
1431         (output_lines): Pass src->maximum_count.
1433 2018-09-17  Martin Liska  <mliska@suse.cz>
1435         * common/config/i386/i386-common.c (ix86_get_valid_option_values):
1436         Use processor_names table.
1437         * config/i386/i386.c (ix86_default_align): Use
1438         processor_cost_table for alignment values.
1439         (ix86_option_override_internal): Use processor_names.
1440         (ix86_function_specific_print): Likewise.
1441         * config/i386/i386.h (struct processor_costs):
1442         Add alignment values.
1443         (struct ptt): Remove and replace with const char *.
1444         * config/i386/x86-tune-costs.h (struct processor_costs):
1445         Declare default alignments for all costs.
1447 2018-09-17  Aldy Hernandez  <aldyh@redhat.com>
1449         * tree-vrp.c (extract_range_from_unary_expr): Do not special case
1450         symbolics or VR_VARYING ranges for ABS_EXPR.
1451         * wide-int-range.cc (wide_int_range_abs): Return positive numbers
1452         when range will wrap.
1454 2018-09-15  Eric Botcazou  <ebotcazou@adacore.com>
1456         PR middle-end/86864
1457         * cfgexpand.c (expand_gimple_basic_block): Be prepared for a BARRIER
1458         before and after a JUMP_TABLE_DATA.
1460 2018-09-14  John David Anglin  <danglin@gcc.gnu.org>
1462         PR middle-end/87188
1463         * dojump.c (do_compare_and_jump): Canonicalize function pointers
1464         when one operand is a function pointer.  Use POINTER_TYPE_P and
1465         FUNC_OR_METHOD_TYPE_P.
1466         * expr.c (do_store_flag): Use POINTER_TYPE_P and FUNC_OR_METHOD_TYPE_P.
1467         * fold-const.c (build_range_check): Likewise.
1468         * match.pd (simple_comparison): Likewise.
1470 2018-09-14  David Malcolm  <dmalcolm@redhat.com>
1472         PR c/82967
1473         * spellcheck.c (get_edit_distance_cutoff): New function.
1474         (selftest::test_edit_distance_unit_test_oneway): Rename to...
1475         (selftest::test_get_edit_distance_one_way): ...this.
1476         (selftest::test_get_edit_distance_unit): Rename to...
1477         (selftest::test_get_edit_distance_both_ways): ...this.
1478         (selftest::test_edit_distances): Move tests to this new function,
1479         and test some more pairs of strings.  Update for above renaming.
1480         (selftest::get_old_cutoff): New function.
1481         (selftest::test_get_edit_distance_cutoff): New function.
1482         (selftest::assert_suggested_for): New function.
1483         (ASSERT_SUGGESTED_FOR): New macro.
1484         (selftest::assert_not_suggested_for): New function.
1485         (ASSERT_NOT_SUGGESTED_FOR): New macro.
1486         (selftest::test_suggestions): New function.
1487         (selftest::spellcheck_c_tests): Move test_get_edit_distance_unit
1488         tests to selftest::test_edit_distances and call it.  Add calls to
1489         selftest::test_get_edit_distance_cutoff and
1490         selftest::test_suggestions.
1491         * spellcheck.h (get_edit_distance_cutoff): New function declaration.
1492         (best_match::consider): Replace hard-coded cutoff calculation with
1493         a call to...
1494         (best_match::get_cutoff): New declaration.
1495         (best_match::get_best_meaningful_candidate): Likewise.
1497 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1499         * builtins.c (fold_builtin_strlen): Remove TODO comment.
1501 2018-09-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1503         revert:
1504         2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1506         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
1507         terminated string literal.
1509 2018-09-14  Martin Sebor  <msebor@redhat.com>
1511         * builtins.c (unterminated_array): Handle ARRAY_REF.
1512         (expand_builtin_stpcpy_1): Detect unterminated char arrays.
1513         * builtins.h (unterminated_array): Declare extern.
1514         * gimple-fold.c (gimple_fold_builtin_stpcpy): Detect unterminated
1515         arrays.
1516         (gimple_fold_builtin_sprintf): Propagate NO_WARNING to transformed
1517         calls.
1519 2018-09-14  Martin Sebor  <msebor@redhat.com>
1520             Jeff Law  <law@redhat.com>
1522         * builtins.c (unterminated_array): New.
1523         (expand_builtin_strcpy): Adjust.
1524         (expand_builtin_strcpy_args): Detect unterminated arrays.
1525         * gimple-fold.c (get_maxval_strlen): Add argument.  Detect
1526         unterminated arrays.
1527         * gimple-fold.h (get_maxval_strlen): Add argument.
1528         (gimple_fold_builtin_strcpy): Detec unterminated arrays.
1530         * gimple-fold.c (get_range_strlen): Add argument.
1531         (get_maxval_strlen): Adjust.
1532         * gimple-fold.h (get_range_strlen): Add argument.
1534 2018-09-14  Wei Xiao  <wei3.xiao@intel.com>
1536         * config/i386/movdirintrin.h: Fix copyright year.
1538 2018-09-14  Uros Bizjak  <ubizjak@gmail.com>
1540         * reg-stack.c: Include regs.h.
1541         (replace_reg): Assert that mode is MODE_FLOAT or MODE_COMPLEX_FLOAT.
1542         (emit_pop_insn): Default pop insn mode to the reg_raw_mode of
1543         FIRST_STACK_REG, not DFmode.
1544         (emit_swap_insn): Default swap insn mode to the reg_raw_mode of
1545         FIRST_STACK_REG, not XFmode.  Explicitly construct swap RTX.
1546         (change stack): Default register mode to the reg_raw_mode of
1547         FIRST_STACK_REG, not DFmode.
1548         * config/i386/i386.md (*swap<mode>): Remove insn pattern.
1549         (*swapxf): Rename from swapxf.
1551 2018-09-14  Carl Love  <cel@us.ibm.com>
1553         * config/rs6000/emmintrin.h: Add _MM_SHUFFLE2.
1554         * config/rs6000/xmmintrin.h: Add _MM_SHUFFLE.
1556 2018-09-14  Segher Boessenkool  <segher@kernel.crashing.org>
1558         PR target/87224
1559         * config/rs6000/rs6000.md (*mov<mode>_hardfloat64): Add Z to the Y
1560         alternatives.
1562 2018-09-14  Sam Tebbs  <sam.tebbs@arm.com>
1564         PR target/85628
1565         * config/aarch64/aarch64.md (*aarch64_bfxilsi_uxtw): Define.
1567 2018-09-14  Jason Merrill  <jason@redhat.com>
1569         Fix --enable-gather-detailed-mem-stats.
1570         * hash-table.c (hash_table_usage): Change from variable to function.
1571         * hash-table.h: Adjust.
1572         * Makefile.in: Add missing dependencies on hash-table.h.
1574 2018-09-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1576         PR tree-optimization/87259
1577         PR lto/87283
1578         (pass_cse_reciprocals::execute): Run optimize_recip_sqrt after
1579         execute_cse_reciprocals_1 has tried transforming.
1581 2018-09-14  Aldy Hernandez  <aldyh@redhat.com>
1583         * tree-vrp.c (extract_range_from_binary_expr_1): Normalize
1584         VR_VARYING for PLUS/MINUS_EXPR.
1586 2018-09-14  Ilya Leoshkevich  <iii@linux.ibm.com>
1588         * config/s390/s390-passes.def (INSERT_PASS_BEFORE): Improve
1589         formatting.
1591 2018-09-14  Richard Biener  <rguenther@suse.de>
1593         PR middle-end/63155
1594         * tree-ssa-coalesce.c (ssa_conflicts_merge): Remove conflict
1595         bits for the merged partition.
1597 2018-09-13  Martin Sebor  <msebor@redhat.com>
1598             Bernd Edlinger  <bernd.edlinger@hotmail.de>
1600         * builtins.h (c_srlen): Add argument.
1601         * builtins.c (warn_string_no_nul): New function.
1602         (c_strlen): Add argument and use it.  Update recursive calls.
1603         Pass DECL argument to string_constant to get info on non
1604         terminated strings.  Update *NONSTR as needed.
1605         (fold_builtin_strlen): Add argument to calls to c_strlen.
1606         Warn for unterminated arrays.
1607         (warn_string_no_null): Add prototype.
1608         * expr.c (string_constant): Update arguments.  Update recursive
1609         calls appropriately.  Detect missing NUL terminator and outermost
1610         declaration its missing in.
1611         Improve checks for arrays with nonzero lower bound or elements
1612         that are not a single byte.  Simplify offset computation.
1613         Simplify checks for non-NUL terminated strings.
1614         * gimple-fold.c (get_range_strlen): Add argument to c_strlen call.
1615         * gimple-ssa-sprintf.c (get_string_length): Remove unnecessary code.
1617 2018-09-13  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1619         * builtins.c (c_strlen): Handle not zero terminated STRING_CSTs
1620         correctly.
1621         * fold-const.c (c_getstr): Fix function comment.  Remove unused third
1622         argument.  Fix range checks.
1623         * fold-const.h (c_getstr): Adjust protoype.
1624         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid folding when
1625         string is constant but contains no NUL byte.
1627         * expr.c (string_constant): Adjust function comment.
1628         Remove bogus check for zero termination.
1630         * fold-const.c (c_getstr): Clamp STRING_LENGTH to STRING_SIZE.
1632         * varasm.c (compare_constant): Compare type size of STRING_CSTs.
1633         (get_constant_size): Don't make STRING_CSTs larger than they are.
1634         (check_string_literal): New check function for STRING_CSTs.
1635         (output_constant): Use it.
1637 2018-09-13  Eric Botcazou  <ebotcazou@adacore.com>
1639         PR target/86812
1640         * config/visium/visium.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define.
1642 2018-09-13  Richard Biener  <rguenther@suse.de>
1644         PR tree-optimization/87263
1645         * tree-ssa-sccvn.c (visit_phi): Revert some earlier changes.
1646         (struct unwind_state): Add max_rpo field.
1647         (do_rpo_vn): Allow up-to-date loop state to be used when not iterating.
1648         Compute max_rpo, the max RPO number a block can be backwards reached
1649         from.  Re-write non-iterating mode to a RPO ordered worklist approach,
1650         separating it from the iterating mode.
1652 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
1654         * haifa-sched.c (rank_for_schedule): Schedule by INSN_COST.
1655         (rfs_decision): New scheduling decision.
1657 2018-09-13  Richard Biener  <rguenther@suse.de>
1659         PR bootstrap/87134
1660         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Fix assert.
1661         (vn_nary_op_insert_pieces_predicated): Do not write useless
1662         valid_dominated_by_p entry outside of the allocated storage.
1664 2018-09-13  Omar Sandoval  <osandov@osandov.com>
1665             Tom de Vries  <tdevries@suse.de>
1667         PR debug/86985
1668         * dwarf2out.c (is_c): New function.
1669         (add_subscript_info): Add DW_AT_count of 0 for C zero-length arrays.
1671 2018-09-13  Sam Tebbs  <sam.tebbs@arm.com>
1673         PR target/85628
1674         * config/aarch64/aarch64.md (*aarch64_bfxil):
1675         Define.
1676         * config/aarch64/constraints.md (Ulc): Define.
1677         * config/aarch64/aarch64-protos.h (aarch64_high_bits_all_ones_p):
1678         Define.
1679         * config/aarch64/aarch64.c (aarch64_high_bits_all_ones_p):
1680         New function.
1682 2018-09-13  Vlad Lazar  <vlad.lazar@arm.com>
1684         * config/aarch64/aarch64.h (TARGET_COMPUTE_FRAME_LAYOUT): Define.
1685         * config/aarch64/aarch64.c (aarch64_expand_prologue): Remove
1686         aarch64_layout_frame call.
1687         (aarch64_expand_epilogue): Likewise.
1688         (aarch64_initial_elimination_offset): Likewise.
1689         (aarch64_get_separate_components): Likewise.
1690         (aarch64_use_return_insn_p): Likewise.
1691         (aarch64_layout_frame): Remove unneeded check.
1693 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
1695         * configure.ac: Only append
1696         " : (reconfigured) $TOPLEVEL_CONFIGURE_ARGUMENTS" to
1697         gcc_config_arguments if it was never reconfigured or last reconfigure
1698         was with different arguments.
1699         * configure: Regenerated.
1701 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
1702             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1704         PR middle-end/87290
1705         * expr.c (maybe_optimize_pow2p_mod_cmp): New function.
1706         (maybe_optimize_mod_cmp): Use it if integer_pow2p treeop1.
1708 2018-09-13  Jakub Jelinek  <jakub@redhat.com>
1710         PR tree-optimization/87287
1711         * fold-const.c (fold_binary_loc) <case EQ_EXPR>: Move signed modulo
1712         X % C == 0 to X % (unsigned) C == 0 optimization to ...
1713         * match.pd (X % C == 0): ... here.  New optimization.
1715 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
1717         PR middle-end/82853
1718         * expr.h (maybe_optimize_mod_cmp): Declare.
1719         * expr.c (mod_inv): New function.
1720         (maybe_optimize_mod_cmp): New function.
1721         (do_store_flag): Use it.
1722         * cfgexpand.c (expand_gimple_cond): Likewise.
1724 2018-09-09  Cesar Philippidis  <cesar@codesourcery.com>
1725             Julian Brown  <julian@codesourcery.com>
1727         PR middle-end/86336
1728         * gimplify.c (gimplify_scan_omp_clauses): Set
1729         target_firstprivatize_array_bases in OpenACC parallel and kernels
1730         region contexts.  Remove GOMP_MAP_FIRSTPRIVATE_REFERENCE clauses from
1731         OpenACC data regions.
1733 2018-09-12  Uros Bizjak  <ubizjak@gmail.com>
1735         * config/i386/i386.md (sqrt_extend<mode>xf3_i387): Remove.
1736         (sqrt<mode>2): Extend operand 1 to XFmode and generate sqrtxf3 insn.
1738 2018-09-12  Richard Biener  <rguenther@suse.de>
1740         PR tree-optimization/87280
1741         * tree-ssa-sccvn.c (process_bb): Handle the case of executable
1742         edge but unreachable target.
1743         (do_rpo_vn): For conservatively handling a PHI only mark
1744         the backedge executable but not the block reachable.
1746 2018-09-12  Richard Biener  <rguenther@suse.de>
1748         PR tree-optimization/87266
1749         * tree-ssa-sccvn.c (do_rpo_vn): Always iterate to not yet
1750         visited blocks.
1752 2018-09-12  Andreas Krebbel  <krebbel@linux.ibm.com>
1754         * config/s390/s390.md (PFPO_RND_MODE_DFP, PFPO_RND_MODE_BFP): New
1755         constants.
1756         ("trunc<BFP:mode><DFP_ALL:mode>2")
1757         ("trunc<DFP_ALL:mode><BFP:mode>2")
1758         ("extend<BFP:mode><DFP_ALL:mode>2")
1759         ("extend<DFP_ALL:mode><BFP:mode>2"): Set proper rounding mode
1760         according to the target operand type.
1762 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
1763             Andreas Krebbel  <krebbel@linux.ibm.com>
1765         PR tree-optimization/86844
1766         * gimple-ssa-store-merging.c
1767         (imm_store_chain_info::coalesce_immediate): For overlapping stores, if
1768         there are any overlapping stores in between them, make sure they are
1769         also coalesced or we give up completely.
1771 2018-09-12  Jakub Jelinek  <jakub@redhat.com>
1773         PR middle-end/87248
1774         * fold-const.c (fold_ternary_loc) <case COND_EXPR>: Verify also that
1775         BIT_AND_EXPR's second operand is a power of two.  Formatting fix.
1777 2018-09-12  Tom de Vries  <tdevries@suse.de>
1779         * common.opt (gdescribe-dies): Add option.
1780         * dwarf2out.c (add_name_and_src_coords_attributes): Add description
1781         attribute for artifical and nameless decls.
1782         (dwarf2out_register_external_die): Add description attribute to
1783         external reference die.
1784         (add_desc_attribute): New functions.
1785         (gen_subprogram_die): Add description attribute to
1786         DW_TAG_call_site_parameter.
1787         * tree-pretty-print.c (print_generic_expr_to_str): New function.
1788         * tree-pretty-print.h (print_generic_expr_to_str): Declare.
1789         * doc/invoke.texi (@item Debugging Options): Add -gdescribe-dies and
1790         -gno-describe-dies.
1791         (@item -gdescribe-dies): Add.
1793 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
1795         * tree-vrp.c (vrp_shift_undefined_p): Remove.
1796         (extract_range_from_binary_expr_1: Call
1797         wide_int_range_shift_undefined_p instead of vrp_shift_undefined_p.
1798         * wide-int-range.h (wide_int_range_shift_undefined_p): Do not
1799         depend on sign.
1801 2018-09-12  Aldy Hernandez  <aldyh@redhat.com>
1803         * gimple-ssa-warn-alloca.c
1804         (alloca_type_and_limit::alloca_type_and_limit): Initialize limit
1805         field for ALLOCA_BOUND_*_LARGE.
1807 2018-09-11  Nathan Sidwell  <nathan@acm.org>
1809         * gcc.c (load_specs, execute, run_attempt): Use %qs not '%s'.
1811 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
1813         * reg-stack.c (subst_asm_stack_regs): Call replace_reg also
1814         for clobbers.  Remove obsolete comment.
1816 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
1818         * config/i386/i386.md (define_attr "type"): Remove mpxmov, mpxmk,
1819         mpxchk, mpxld and mpxst types.
1820         (define_attr length_immediate): Remove all processing of mpx types.
1821         (define_attr prefix_0f): Ditto.
1822         (define_attr memory): Ditto.
1824 2018-09-11  Uros Bizjak  <ubizjak@gmail.com>
1826         * config/i386/i386.md (fyl2x_extend<mode>xf3_i387): Remove.
1827         (log<mode>2): Change operand 1 predicate to general_operand.
1828         Extend operand 1 to XFmode and generate logxf3 insn.
1829         (log10<mode>2): Change operand 1 predicate to general_operand.
1830         Extend operand 1 to XFmode and generate log10xf3 insn.
1831         (log2<mode>2): Change operand 1 predicate to general_operand.
1832         Extend operand 1 to XFmode and generate log2xf3 insn.
1833         (fyl2xp1_extend<mode>xf3_i387): Remove.
1834         (log1p<mode>2): Change operand 1 predicate to general_operand.
1835         Extend operand 1 to XFmode and generate log1pxf3 insn.
1836         (fxtract_extend<mode>xf3_i387): Remove.
1837         (logb<mode>2): Change operand 1 predicate to general_operand.
1838         Extend operand 1 to XFmode and generate logbxf3 insn.
1839         (ilogb<mode>2): Change operand 1 predicate to general_operand.
1840         Extend operand 1 to XFmode and generate fxtractxf3_i387 insn.
1841         (significand<mode>2): Change operand 1 predicate to general_operand.
1842         Extend operand 1 to XFmode and generate significandxf3 insn.
1844 2018-09-11  Nathan Sidwell  <nathan@acm.org>
1846         * gcc.c (perror_with_name, pfatal_with_name): Delete.
1847         (load_specs): Use fatal_error.
1848         (DELETE_IF_ORDINARY, process_command): Use error.
1849         (execute, run_attempt): Use fatal_error.
1851 2018-09-11  Andrew Stubbs  <ams@codesourcery.com>
1853         * diagnostic-core.h (sorry_at): New prototype.
1854         * diagnostic.c (sorry_at): New function.
1856 2018-09-11  Aldy Hernandez  <aldyh@redhat.com>
1858         * tree-vrp.c (extract_range_from_binary_expr_1): Treat all divisions
1859         by zero as VR_UNDEFINED.
1861 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
1863         * config/i386/i386.md (<sincos>xf2): Rename from *<sincos>xf2_i387.
1864         (*<sincos>_extend<mode>xf2_i387): Remove insn pattern.
1865         (<sincos>mode2): New expander.
1866         (sincos_extend<mode>xf3_i387): Remove insn pattern.
1867         (sincos -> sin, cos splitters): Remove splitter patterns.
1868         (sincos<mode>3): Change operand 2 predicate to general_operand.
1869         Extend operand 2 to XFmode and generate sincosxf3 insn.
1870         (fptanxf4_i387): Change mode of operands 0 and 3 to SFmode.
1871         Change operand 3 predicate to const1_operand.
1872         (fptan_extend<mode>xf4_i387): Remove insn pattern.
1873         (tanxf2): Update operands in the call to fptanxf4_i387.
1874         (tan<mode>2): Change operand 1 predicate to general_operand.
1875         Extend operand 1 to XFmode and generate tanxf3 insn.
1876         (atan2xf3): Rename from *fpatanxf3_i387.
1877         (fpatan_extend<mode>xf3_i387): Remove insn pattern.
1878         (atan2xf3): Remove expander.
1879         (atan2<mode<3):  Change operand 1 and 2 predicates to general_operand.
1880         Extend operands 1 and 2 to XFmode and generate atan2xf3 insn.
1881         (atan<mode>2): Change operand 1 predicate to general_operand.
1882         Extend operand 1 to XFmode and generate atanxf3 insn.
1884 2018-09-10  Uros Bizjak  <ubizjak@gmail.com>
1886         * config/i386/i386.md (x87/SSE constant load splitter): Use
1887         memory_operand instead of nonimmediate_operand for input operand
1888         predicate.
1890 2018-09-09  Uros Bizjak  <ubizjak@gmail.com>
1892         * config/i386/i386.md (float partial SSE register stall splitter): Move
1893         splitter near its instruction pattern.
1894         (float_extend partial SSE register stall splitter): Ditto.
1895         (float_truncate partial SSE register stall splitter): Ditto.
1897 2018-09-09  Hans-Peter Nilsson  <hp@bitrange.com>
1899         PR target/86794
1900         * config/mmix/mmix.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
1901         to speculation_safe_value_not_needed.
1903         PR target/85666
1904         * config/mmix/mmix.c (mmix_assemble_integer): Handle byte-size
1905         non-CONST_INT rtx:es using assemble_integer_with_op ".byte".
1906         (MMIX_CFUN_NEEDS_SAVED_EH_RETURN_ADDRESS): Don't call
1907         leaf_function_p, instead use has_hard_reg_initial_val.
1909 2018-09-09  Nathan Sidwell  <nathan@acm.org>
1911         * gcc.h (pfatal_with_name): Don't declare here.
1912         * gcc.c (pfatal_with_name): Make static.
1914 2018-09-09  Xianmiao Qu  <xianmiao_qu@c-sky.com>
1916         * config/csky/csky.md (*cskyv2_adddi3, *cskyv2_subdi3): Add
1917         earlyclobber.
1919 2018-09-08  John David Anglin  <danglin@gcc.gnu.org>
1921         PR rtl-optimization/85458
1922         * sel-sched.c (sel_target_adjust_priority): Allow backend adjust
1923         priority hook to reduce the priority of EXPR.
1925 2018-09-07  Uros Bizjak  <ubizjak@gmail.com>
1927         * config/i386/i386.md (float<SWI48x:mode><MODEF:mode>2) Enable
1928         DImode for x87 on 32bit targets.  Conditionally disable x87 modes
1929         with X87_ENABLE_FLOAT.  Remove preparation code.
1930         (*float<SWI48:mode><MODEF:mode>2): Rename from
1931         *float<SWI48:mode><MODEF:mode>2_mixed.  Handle x87, SSE and mixed
1932         math using "enabled" attribute.
1933         (*floatdi<MODEF:mode>2_i387): Rename from
1934         *float<SWI48x:mode><MODEF:mode>2_i387.  Handle only DImode and
1935         enable for 32bit targets only.
1936         (floatdi<X87MODEF:mode>2_i387_with_xmm pre-reload splitter): New
1937         splitter.
1938         (floatdi<X87MODEF:mode>2_i387_with_xmm): Use register_operand
1939         as operand 1 predicate.  Rewrite as define_insn_and_split.
1940         (floatdi<X87MODEF:mode>2_i387_with_xmm memory input splitter): Remove.
1942 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
1944         * reg-stack.c (get_true_reg) <case FLOAT_TRUNCATE>: Reorder
1945         to fallthru to FLOAT case.
1947 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
1949         PR target/86731
1950         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Update logic
1951         around folding of vec_sl to handle out of range shift values.
1953 2018-09-06  Uros Bizjak  <ubizjak@gmail.com>
1955         * config/i386/i386.md (fix_trunc<mode>_fisttp_i387_1): Remove.
1956         Update callers to gen_fix_trunc<mode>_i387_fisttp
1957         (fix_trunc<mode>_i387_fisttp): Change operand 0 predicate to
1958         nonimmediate_operand.
1959         (fix_trunc<mode>_i387_fisttp_with_temp): Remove insn pattern
1960         and corresponding splitters.
1961         (*fix_trunc<mode>_i387_1): Always expand with fix_trunc<mode>_i387.
1962         (fix_truncdi_i387): Change operand 0 predicate to nonimmediate_operand.
1963         (fix_truncdi_i387_with_temp): Remove insn pattern
1964         and corresponding splitters.
1965         (fix_trunc<mode>_i387): Change operand 0 predicate to
1966         nonimmediate_operand.
1967         (fix_trunc<mode>_i387_with_temp): Remove insn pattern
1968         and corresponding splitters.
1969         (*fistdi2_1): Remove.
1970         (fistdi2): Ditto.
1971         (fistdi2_with_temp): Remove insn pattern and corresponding splitters.
1972         (lrintxfdi2): Remove expander.  Reimplement as define_insn.
1973         (*fist<mode>2_1): Remove.
1974         (fist<mode>2): Ditto.
1975         (fist<mode>2_with_temp): Remove insn pattern and corresponding
1976         splitters.
1977         (lrintxf<mode>2): Remove expander.  Reimplement as define_insn.
1978         (*fist<mode>2_<rounding>_1): Always expand with fist<mode>2_<rounding>.
1979         (fistdi2_<rounding>): Change operand 0 predicate to
1980         nonimmediate_operand.
1981         (fistdi2_<rounding>_with_temp): Remove insn pattern
1982         and corresponding splitters.
1983         (fist<mode>2_<rounding>): Change operand 0 predicate to
1984         nonimmediate_operand.
1985         (fist<mode>2_<rounding>_with_temp): Remove insn pattern
1986         and corresponding splitters.
1988         (*fixuns_trunc<mode>si2_avx512f_zext): Depend on TARGET_SSE_MATH.
1990 2018-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
1992         * varasm.c (output_constructor_regular_field): Check TYPE_SIZE_UNIT of
1993         the init value.
1995 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
1997         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
1998         early gimple folding of vec_splat().
1999         * tree-vect-generic.c: Remove static from tree_vec_extract() definition.
2000         * gimple-fold.h: Add an extern define for tree_vec_extract().
2002 2018-09-06  Will Schmidt  <will_schmidt@vnet.ibm.com>
2004         * config/rs6000/rs6000.c (fold_mergehl_helper): Add types_compatible_p
2005         wrappers around TREE_TYPE comparisons.
2007 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2009         PR target/80080
2010         * config/s390/predicates.md: Add nonsym_memory_operand.
2011         * config/s390/s390.c (s390_legitimize_cs_operand): If operand
2012         contains a SYMBOL_REF, load it into an intermediate pseudo.
2013         (s390_emit_compare_and_swap): Legitimize operand.
2014         * config/s390/s390.md: Use the new nonsym_memory_operand
2015         with UNSPECV_CAS patterns.
2017 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2019         PR target/80080
2020         * config/s390/s390-passes.def: New file.
2021         * config/s390/s390-protos.h (class rtl_opt_pass): Add forward
2022         declaration.
2023         (make_pass_s390_early_mach): Add declaration.
2024         * config/s390/s390.c (make_pass_s390_early_mach):
2025         (s390_option_override): Remove dynamic registration.
2026         * config/s390/t-s390: Add s390-passes.def.
2028 2018-09-06  Ilya Leoshkevich  <iii@linux.ibm.com>
2030         * config/s390/s390.c (s390_decompose_constant_pool_ref):
2031         Remove UNSPEC_LTREL_BASE check.
2032         (annotate_constant_pool_refs): Likewise.
2033         (find_constant_pool_ref): Likewise.
2034         (find_ltrel_base): Removed.
2035         (replace_ltrel_base): Removed.
2036         (s390_mainpool_finish): Remove replace_ltrel_base call.
2037         (s390_chunkify_start): Remove pending LTREL_BASE logic.
2038         (s390_chunkify_finish): Remove replace_ltrel_base call.
2039         * config/s390/s390.md: Remove UNSPEC_LTREL_BASE.
2041 2018-09-06  Hans-Peter Nilsson  <hp@axis.com>
2043         PR target/86779
2044         * config/cris/cris.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
2045         to speculation_safe_value_not_needed.
2047 2018-09-05  Cesar Philippidis  <cesar@codesourcery.com>
2048             Bernd Schmidt  <bernds_cb1@t-online.de>
2050         * config/nvptx/nvptx-opts.h: New file.
2051         * config/nvptx/nvptx.c (nvptx_file_start): Print the correct .target.
2052         * config/nvptx/nvptx.h: Include "nvptx-opts.h".
2053         (ASM_SPEC): Define.
2054         (TARGET_SM35): New macro.
2055         * config/nvptx/nvptx.md (atomic_fetch_<logic><mode>): Enable with the
2056         correct predicate.
2057         * config/nvptx/nvptx.opt (ptx_isa, sm_30, sm_35): New enum and its
2058         values.
2059         (misa=): New option.
2060         * doc/invoke.texi (Nvidia PTX Options): Document -misa.
2062 2018-09-05  Uros Bizjak  <ubizjak@gmail.com>
2064         * config/i386/i386.md (truncdfsf2): Remove expander.
2065         (truncdfsf2_with_temp): Ditto.
2066         (truncxf<mode>2): Ditto.
2067         (*truncdfsf_fast_mixed): Remove insn pattern.
2068         (*truncdfsf_fast_i387): Ditto.
2069         (*truncdfsf_mixed): Ditto.
2070         (*truncdfsf_i387): Ditto.
2071         (*truncdfsf2_i387_1): Ditto.
2072         (*truncxfsf2_mixed): Ditto.
2073         (*truncxfdf2_mixed): Ditto.
2074         (*truncxf<mode>2_i387_noop): Ditto. Update callers
2075         to call gen_truncxf<mode>2 instead.
2076         (*truncxf<mode>2_i387): Remove.
2077         (reg->reg splitters): Remove splitter pattern.
2078         (reg->mem splitters): Ditto.
2080         (truncdfsf2): New insn pattern.
2081         (truncxf<mode>2): Ditto.
2083 2018-09-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2085         * tree-ssa-math-opts.c (is_mult_by): New function.
2086         (is_square_of): Use the above.
2087         (optimize_recip_sqrt): New function.
2088         (pass_cse_reciprocals::execute): Use the above.
2090 2018-09-05  Richard Biener  <rguenther@suse.de>
2092         PR bootstrap/87134
2093         * tree-ssa-sccvn.c (rpo_elim::eliminate_push_avail): Make sure
2094         to zero-init the emplaced vec.
2096 2018-09-05  Martin Liska  <mliska@suse.cz>
2098         PR tree-optimization/87205
2099         * tree-switch-conversion.c (pass_lower_switch::execute):
2100         Group cases for switch statements.
2102 2018-09-05  Richard Biener  <rguenther@suse.de>
2104         PR tree-optimization/87217
2105         * tree-ssa-sccvn.c (vuse_valueize): New.
2106         (vn_reference_lookup_pieces): Use it.
2107         (vn_reference_lookup): Likewise.
2109 2018-09-05  Nathan Sidwell  <nathan@acm.org>
2111         PR c++/87137
2112         * stor-layout.c (place_field): Scan forwards to check last
2113         bitfield when ms_bitfield_placement is in effect.
2115 2018-09-05  Richard Biener  <rguenther@suse.de>
2117         PR bootstrap/87225
2118         * tree-vect-stmts.c (vectorizable_simd_clone_call): Fix bogus
2119         return.
2121 2018-09-05  Siddhesh Poyarekar  <siddhesh@sourceware.org>
2122             Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
2124         * config/aarch64/falkor-tag-collision-avoidance.c: New file.
2125         * config.gcc (extra_objs): Build it.
2126         * config/aarch64/t-aarch64 (falkor-tag-collision-avoidance.o):
2127         Likewise.
2128         * config/aarch64/aarch64-passes.def
2129         (pass_tag_collision_avoidance): New pass.
2130         * config/aarch64/aarch64.c (qdf24xx_tunings): Add
2131         AARCH64_EXTRA_TUNE_RENAME_LOAD_REGS to tuning_flags.
2132         (aarch64_classify_address): Remove static qualifier.
2133         (aarch64_address_info, aarch64_address_type): Move to...
2134         * config/aarch64/aarch64-protos.h: ... here.
2135         (make_pass_tag_collision_avoidance): New function.
2136         * config/aarch64/aarch64-tuning-flags.def (rename_load_regs):
2137         New tuning flag.
2139 2018-09-05  Martin Liska  <mliska@suse.cz>
2141         * doc/gcov.texi: Update documentation of humar
2142         readable mode.
2143         * gcov.c (format_count): Print one decimal place, it provides
2144         more fine number of situations like '1G' vs. '1.4G'.
2146 2018-09-05  Martin Liska  <mliska@suse.cz>
2148         PR target/87164
2149         * config/rs6000/rs6000.opt: Mark the option as Deprecated.
2150         * optc-gen.awk: Allow 'Var' for Deprecated options in order
2151         to generate a MASK value.
2153 2018-09-04  H.J. Lu  <hongjiu.lu@intel.com>
2155         PR debug/86593
2156         * dwarf2out.c (based_loc_descr): Allow hard frame pointer even
2157         if frame pointer isn't used.
2158         (compute_frame_pointer_to_fb_displacement): Likewise.
2160 2018-09-04  Jakub Jelinek  <jakub@redhat.com>
2162         PR target/87198
2163         * common/config/i386/i386-common.c (OPTION_MASK_ISA_XSAVEOPT_SET,
2164         OPTION_MASK_ISA_XSAVES_SET, OPTION_MASK_ISA_XSAVEC_SET): Use
2165         OPTION_MASK_ISA_XSAVE_SET instead of OPTION_MASK_ISA_XSAVE.
2166         (OPTION_MASK_ISA_XSAVE_UNSET): Add OPTION_MASK_ISA_XSAVES_UNSET
2167         and OPTION_MASK_ISA_XSAVEC_UNSET.
2169 2018-09-04  Max Filippov  <jcmvbkbc@gmail.com>
2171         * config/xtensa/xtensa.c (xtensa_expand_atomic): Reorder AND and
2172         XOR operations in NAND case.
2174 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
2176         * wide-int-range.cc (wide_int_range_convert): New.
2177         * wide-int-range.h (wide_int_range_convert): New.
2178         * tree-vrp.c (extract_range_from_unary_expr): Abstract wide int
2179         code into wide_int_range_convert.
2180         (extract_range_into_wide_ints): Do not munge anti range constants
2181         into the entire domain.  Just return the range back.
2183 2018-09-04  Martin Liska  <mliska@suse.cz>
2185         * genmatch.c (output_line_directive): Add new argument
2186         fnargs.
2187         (dt_simplify::gen_1): Encapsulate dump within __builtin_expect.
2189 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
2191         * doc/invoke.texi (Option Summary): Add whitespace.
2193         * doc/invoke.texi (Option Summary): Add -Waligned-new.
2195 2018-09-04  Richard Biener  <rguenther@suse.de>
2197         PR tree-optimization/87211
2198         * tree-ssa-sccvn.c (visit_phi): When value-numbering to a
2199         backedge value we're supposed to treat as VARYING also number
2200         the PHI to VARYING in case it got a different value-number already.
2202 2018-09-04  Aldy Hernandez  <aldyh@redhat.com>
2204         * tree-vrp.c (vrp_can_optimize_bit_op): Remove.
2205         (extract_range_from_binary_expr_1): Do not call
2206         vrp_can_optimize_bit_op.
2207         * wide-int-range.cc (wide_int_range_can_optimize_bit_op): Make
2208         static.
2209         (wide_int_range_get_mask_and_bounds): New.
2210         (wide_int_range_optimize_bit_op): New.
2211         (wide_int_range_bit_ior): Call wide_int_range_optimize_bit_op.
2212         (wide_int_range_bit_and): Same.
2213         * wide-int-range.h (wide_int_range_can_optimize_bit_op): Remove.
2214         (wide_int_range_optimize_bit_op): New.
2215         (wide_int_range_get_mask_and_bounds): New.
2217 2018-09-04  Richard Biener  <rguenther@suse.de>
2219         PR tree-optimization/87176
2220         * tree-ssa-sccvn.c (visit_phi): Remove redundant allsame
2221         variable.  When value-numbering a virtual PHI node make sure
2222         to not value-number to the backedge value.
2224 2018-09-04  Jonathan Wakely  <jwakely@redhat.com>
2226         * doc/extend.texi (Long Long, Hex Floats): Document support for
2227         long long and hex floats in more recent versions of ISO C++.
2229 2018-09-03  Richard Biener  <rguenther@suse.de>
2231         PR tree-optimization/87177
2232         * tree-ssa-sccvn.c (vuse_ssa_val): Revert previous change, keep
2233         cleanup.
2235 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
2237         * bb-reorder.c (edge_order): Convert to C-qsort-style
2238         tri-state comparator.
2239         (reorder_basic_blocks_simple): Change std::stable_sort to
2240         gcc_stablesort.
2242 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
2244         * tree-loop-distribution.c (offset_cmp): Convert to C-qsort-style
2245         tri-state comparator.
2246         (fuse_memset_builtins): Change std::stable_sort to gcc_stablesort.
2248 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
2250         * sort.cc (struct sort_ctx): New field 'nlim'.  Use it...
2251         (mergesort): ... here as maximum count for using netsort.
2252         (gcc_qsort): Set nlim to 3 if stable sort is requested.
2253         (gcc_stablesort): New.
2254         * system.h (gcc_stablesort): Declare.
2256 2018-09-03  Alexander Monakov  <amonakov@ispras.ru>
2258         * sort.cc (gcc_qsort) [CHECKING_P]: Call qsort_chk.
2259         * system.h (qsort): Always redirect to gcc_qsort.  Update comment.
2260         * vec.c (qsort_chk): Do not call gcc_qsort.  Update comment.
2262 2018-09-03  Segher Boessenkool  <segher@kernel.crashing.org>
2264         * config/rs6000/rs6000.md (*mov<mode>_hardfloat32): Remove %U from the
2265         lxsdx and stxsdx alternatives.
2266         (*mov<mode>_hardfloat64): Ditto.
2267         * config/rs6000/vsx.md (*vsx_extract_<mode>_store): Ditto.
2269 2018-09-03  Richard Biener  <rguenther@suse.de>
2271         PR tree-optimization/87200
2272         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Valueize a
2273         simplify result.
2275 2018-09-03  Martin Liska  <mliska@suse.cz>
2277         PR tree-optimization/87201
2278         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
2279         Fix parenthesis in an expression.
2281 2018-09-03  Richard Biener  <rguenther@suse.de>
2283         PR tree-optimization/87197
2284         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Mark the new def
2285         visited.  CSE the VN_INFO hashtable lookup.
2287         PR tree-optimization/87169
2288         * tree-ssa-sccvn.c (do_rpo_vn): When marking loops for not
2289         iterating make sure there's no extra backedges from irreducible
2290         regions feeding the header.  Mark the destination block
2291         executable.
2293 2018-09-03  Martin Liska  <mliska@suse.cz>
2295         PR driver/83193
2296         * common/common-target.def: Add TARGET_GET_VALID_OPTION_VALUES.
2297         * common/common-targhooks.c (default_get_valid_option_values):
2298         New function.
2299         * common/common-targhooks.h (default_get_valid_option_values):
2300         Likewise.
2301         * common/config/i386/i386-common.c: Move processor_target_table
2302         from i386.c.
2303         (ix86_get_valid_option_values): New function.
2304         (TARGET_GET_VALID_OPTION_VALUES): New macro.
2305         * config/i386/i386.c (struct ptt): Move to i386-common.c.
2306         (PTA_*): Move all defined masks into i386-common.c.
2307         (ix86_function_specific_restore): Use new processor_cost_table.
2308         * config/i386/i386.h (struct ptt): Moved from i386.c.
2309         (struct pta): Likewise.
2310         * doc/tm.texi: Document new TARGET_GET_VALID_OPTION_VALUES.
2311         * doc/tm.texi.in: Likewise.
2312         * opt-suggestions.c (option_proposer::suggest_option):
2313         Pass prefix to build_option_suggestions.
2314         (option_proposer::get_completions): Likewise.
2315         (option_proposer::build_option_suggestions): Use the new target
2316         hook.
2317         * opts.c (struct option_help_tuple): New struct.
2318         (print_filtered_help): Use the new target hook.
2320 2018-09-03  Martin Liska  <mliska@suse.cz>
2322         PR middle-end/59521
2323         * predict.c (set_even_probabilities): Add likely_edges
2324         argument and handle cases where we have precisely one
2325         likely edge.
2326         (combine_predictions_for_bb): Catch also likely_edges.
2327         (tree_predict_by_opcode): Handle gswitch statements.
2328         * tree-cfg.h (find_case_label_for_value): New declaration.
2329         (find_taken_edge_switch_expr): Likewise.
2330         * tree-switch-conversion.c (switch_decision_tree::balance_case_nodes):
2331         Find pivot in decision tree based on probabily, not by number of
2332         nodes.
2334 2018-09-02  Gerald Pfeifer  <gerald@pfeifer.com>
2336         * doc/standards.texi (Standards): Update Objective-C reference.
2338 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
2340         * doc/install.texi (Prerequisites): Update link for MPC.
2342 2018-09-01  Michael Matz  <matz@suse.de>
2344         PR tree-optimization/87074
2345         * gimple-loop-jam.c (unroll_jam_possible_p): Check loop exit
2346         PHIs for outer-loop uses.
2348 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
2350         * doc/generic.texi (OpenMP): Adjust link to openmp.org.
2351         * doc/invoke.texi (C Dialect Options): Ditto.
2353 2018-09-01  Gerald Pfeifer  <gerald@pfeifer.com>
2355         * doc/install.texi (Prerequisites): Adjust link mpfr.org.
2357 2018-08-31  Richard Biener  <rguenther@suse.de>
2359         PR tree-optimization/87168
2360         * tree-ssa-sccvn.c (SSA_VAL): Add visited output parameter.
2361         (rpo_elim::eliminate_avail): When OP was not visited it must
2362         be available.
2364 2018-08-31  David Malcolm  <dmalcolm@redhat.com>
2366         * tree-vrp.c (copy_value_range): Convert param "from" from
2367         "value_range *" to "const value_range *".
2368         (range_is_null): Likewise for param "vr".
2369         (range_int_cst_p): Likewise.
2370         (range_int_cst_singleton_p): Likewise.
2371         (symbolic_range_p): Likewise.
2372         (value_ranges_intersect_p): Likewise for both params.
2373         (value_range_nonnegative_p): Likewise for param "vr".
2374         (value_range_constant_singleton): Likewise.
2375         (vrp_set_zero_nonzero_bits): Likewise for param "ar".
2376         (extract_range_into_wide_ints): Likewise for param "vr".
2377         (extract_range_from_multiplicative_op): Likewise for params "vr0"
2378         and "vr1".
2379         (vrp_can_optimize_bit_op): Likewise.
2380         (extract_range_from_binary_expr_1): Likewise for params "vr0_" and
2381         "vr1_".
2382         (extract_range_from_unary_expr): Likewise.
2383         (debug_value_range): Likewise for param "vr".
2384         (value_range::dump): Add "const" qualifier.
2385         (vrp_prop::check_array_ref): Convert local "vr" from
2386         "value_range *" to "const value_range *".
2387         (vrp_prop::check_mem_ref): Likewise.
2388         (vrp_prop::visit_stmt): Likewise for local "old_vr".
2389         (vrp_intersect_ranges_1): Likewise for param "vr_1".
2390         (vrp_intersect_ranges): Likewise.
2391         (simplify_stmt_for_jump_threading): Likewise for local "vr".
2392         (vrp_prop::vrp_finalize): Likewise.
2393         * tree-vrp.h (value_range::dump): Add "const" qualifier.
2394         (vrp_intersect_ranges): Add "const" qualifier to params as above.
2395         (extract_range_from_unary_expr): Likewise.
2396         (value_range_constant_singleton): Likewise.
2397         (symbolic_range_p): Likewise.
2398         (copy_value_range): Likewise.
2399         (extract_range_from_binary_expr_1): Likewise.
2400         (range_int_cst_p): Likewise.
2401         (vrp_set_zero_nonzero_bits): Likewise.
2402         (range_int_cst_singleton_p): Likewise.
2404 2018-08-31  Vlad Lazar  <vlad.lazar@arm.com>
2406         * config/aarch64/arm_neon.h (vabsd_s64): New.
2407         (vnegd_s64): Likewise.
2409 2018-08-31  Martin Jambor  <mjambor@suse.cz>
2411         * ipa-cp.c (estimate_local_effects): Replace wrong MAX with MIN.
2413 2018-08-31  Martin Liska  <mliska@suse.cz>
2415         * ipa-icf.c (sem_item::add_type): Use
2416         sem_item::m_type_hash_cache.
2417         * ipa-icf.h: Move the cache from sem_item_optimizer
2418         to sem_item.
2420 2018-08-31  Nathan Sidwell  <nathan@acm.org>
2422         * doc/extend.texi (Backwards Compatibility): Remove implicit
2423         extern C leeway of () being (...).
2425 2018-08-31  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2427         * ipa-inline.c (can_inline_edge_by_limits_p): Fix typos in comment.
2429 2018-08-31  Segher Boessenkool  <segher@kernel.crashing.org>
2431         PR target/86684
2432         PR target/87149
2433         * config/rs6000/rs6000.md (lround<mode>di2): Gate on TARGET_FPRND.
2435 2018-08-31  Jakub Jelinek  <jakub@redhat.com>
2437         PR middle-end/87138
2438         * expmed.c (expand_mult_const): Use immed_wide_int_const instead of
2439         gen_int_mode.  Formatting fixes.
2441 2018-08-30  Sandra Loosemore  <sandra@codesourcery.com>
2443         * target.def (custom_function_descriptors): Improve documentation.
2444         * doc/tm.texi.in (Trampolines): Expand discussion of function
2445         descriptors and move TARGET_CUSTOM_FUNCTION_DESCRIPTORS to the
2446         beginning of the section.
2447         * doc/tm.texi: Regenerated.
2449 2018-08-30  Jose E. Marchesi  <jose.marchesi@oracle.com>
2451         * cfg.h (class auto_edge_flag): Spell out the template-id of the
2452         base class in the initializer list.  This is a workaround for
2453         building with older GCC.
2454         (class auto_bb_flag): Likewise.
2456 2018-08-30  Aaron Sawdey  <acsawdey@linux.ibm.com>
2458         * config/rs6000/altivec.md (altivec_eq<mode>): Remove star.
2459         (altivec_vcmpequ<VI_char>_p): Remove star.
2460         * config/rs6000/rs6000-string.c (do_load_for_compare): Support
2461         vector load modes.
2462         (expand_strncmp_vec_sequence): New function.
2463         (emit_final_str_compare_vec): New function.
2464         (expand_strn_compare): Add support for vector strncmp.
2465         * config/rs6000/rs6000.opt (-mstring-compare-inline-limit): Change
2466         length specification to bytes.
2467         * config/rs6000/vsx.md (vsx_ld_elemrev_v16qi_internal): Remove star.
2468         (vcmpnezb_p): New pattern.
2469         * doc/invoke.texi (RS/6000 and PowerPC Options): Update documentation
2470         for option -mstring-compare-inline-limit.
2472 2018-08-30  Thiago Macieira  <thiago.macieira@intel.com>
2474         * config/i386/i386.c (PTA_WESTMERE): Remove PTA_AES.
2475         (PTA_SKYLAKE): Add PTA_AES.
2476         (PTA_GOLDMONT): Likewise.
2478 2018-08-29  Jan Hubicka  <jh@suse.cz>
2480         PR lto/86517
2481         * lto-opts.c (lto_write_options): Always stream PIC/PIE mode.
2482         * lto-wrapper.c (merge_and_complain): Fix merging of PIC/PIE.
2484 2018-08-29  Jan Hubicka  <jh@suse.cz>
2486         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not follow
2487         TYPE_STUB_DECL.
2488         (hash_tree): Do not visit TYPE_STUB_DECL.
2489         * tree-streamer-out.c (write_ts_type_common_tree_pointers): Do not
2490         stream TYPE_STUB_DECL.
2491         * tree-streamer-in.c (lto_input_ts_type_common_tree_pointers): Likewise.
2492         * ipa-utils.h (type_with_linkage_p): Do not rely on TYPE_STUB_DECL
2493         after free_lang_data.
2494         (type_in_anonymous_namespace_p): Likewise.
2496 2018-08-29  Jan Hubicka  <jh@suse.cz>
2498         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
2499         comment that it has to be even number.
2500         (class sreal): Change m_sig type to int32_t.
2501         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
2502         int64_t for temporary calculations.
2503         (sreal_verify_basics): Drop one bit from minimum and maximum.
2505 2018-08-30  Richard Biener  <rguenther@suse.de>
2507         PR tree-optimization/87147
2508         * tree-ssa-sccvn.c (SSA_VISITED): New function.
2509         (visit_phi): When the degenerate result is from the backedge and
2510         we didn't visit its definition yet drop to VARYING.
2511         (do_rpo_vn): Properly mark blocks with incoming backedges as executable.
2513 2018-08-29  Jan Hubicka  <jh@suse.cz>
2515         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not walk
2516         DECL_VINDEX.
2517         (hash_tree): Likewise.
2519 2018-08-29  Jan Hubicka  <jh@suse.cz>
2521         * tree.c (find_decls_types_r): Walk also TYPE_NEXT_PTR_TO
2522         and TYPE_NEXT_REF_TO.
2524 2018-08-29  Jan Hubicka  <jh@suse.cz>
2526         * sreal.h (SREAL_PART_BITS): Change to 31; remove seemingly unnecessary
2527         comment that it has to be even number.
2528         (class sreal): Change m_sig type to int32_t.
2529         * sreal.c (sreal::dump, sreal::to_int, opreator+, operator-): Use
2530         int64_t for temporary calculations.
2531         (sreal_verify_basics): Drop one bit from minimum and maximum.
2533 2018-08-30  Tamar Christina  <tamar.christina@arm.com>
2535         * config/aarch64/aarch64.c (aarch64_expand_movmem): Set TImode max.
2537 2018-08-30  Vlad Lazar  <vlad.lazar@arm.com>
2539         PR middle-end/86995
2540         * expmed.c (canonicalize_comparison): Use wi::sub instead of wi::add
2541         if to_add is negative.
2543 2018-08-29  Bernd Edlinger  <bernd.edlinger@hotmail.de>
2545         PR middle-end/87053
2546         * builtins.c (c_strlen): Improve range checks.
2548 2018-08-29  Martin Sebor  <msebor@redhat.com>
2549             Jeff Law  <law@redhat.com>
2551         PR tree-optimization/86714
2552         PR tree-optimization/86711
2553         * builtins.c (c_strlen): Add arguments to call to string_constant.
2554         * expr.c (string_constant): Add argument.  Detect missing nul
2555         terminator and outermost declaration it's missing in.
2556         * expr.h (string_constant): Add argument.
2557         * fold-const.c (read_from_constant_string): Add arguments to call to
2558         string_constant.
2559         (c_getstr): Likewise.
2560         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
2561         to string_constant.
2562         * tree-ssa-strlen.c (get_stridx): Likewise.
2564 2018-08-29  Jan Hubicka  <jh@suse.cz>
2566         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers):
2567         Do not stream DECL_VINDEX.
2568         * tree-streamer-out.c (write_ts_function_decl_tree_pointers): Likewise.
2569         * tree.c (free_lang_data_in_decl): Clear DECL_VINDEX.
2570         (decl_function_context): Use DECL_VIRTUAL_P rather than DECL_VINDEX.
2572 2018-08-29  Richard Biener  <rguenther@suse.de>
2574         * tree-ssa-sccvn.c (vuse_ssa_val): Return NULL for unvisited
2575         virtual operands that are not default defs to honor region
2576         boundaries.
2577         (rpo_vn_valueize): Remove ineffective code here.
2579 2018-08-29  Richard Biener  <rguenther@suse.de>
2581         PR tree-optimization/87132
2582         * tree-ssa-alias.c (get_continuation_for_phi): Do not translate
2583         when skipping defs reachable over backedges.
2585 2018-08-29  Richard Biener  <rguenther@suse.de>
2587         * tree-core.h: Document use of deprecated_flag in SSA_NAME.
2588         * tree.h (SSA_NAME_POINTS_TO_READONLY_MEMORY): Define.
2589         * tree-into-ssa.c (pass_build_ssa::execute): Initialize
2590         function parameters SSA_NAME_POINTS_TO_READONLY_MEMORY from fnspec.
2591         * tree-ssa-sccvn.c (const_parms, init_const_parms): Remove.
2592         (vn_reference_lookup_3): Remove use of const_parms.
2593         (free_rpo_vn): Do not free const_parms.
2594         (do_rpo_vn): Do not call init_const_parms.
2595         * tree-ssa-alias.c (refs_may_alias_p_1): Honor
2596         SSA_NAME_POINTS_TO_READONLY_MEMORY.
2597         (call_may_clobber_ref_p_1): Likewise.
2599 2018-08-29  Alexander Monakov  <amonakov@ispras.ru>
2601         PR other/86726
2602         * invoke.texi (Optimization Options): List -ftree-scev-cprop.
2603         (-O): Ditto.
2604         (-ftree-scev-cprop): Document.
2606 2018-08-29  Jan Hubicka  <jh@suse.cz>
2608         * sreal.h (normalize, normalize_up, normalize_down): Add new_sig/new_exp
2609         parameters.
2610         (sreal constructor): Update.
2611         * sreal.c (sreal:operator+, sreal:operator-, sreal:operator*,
2612         sreal:operator/): Update.
2614 2018-08-29  Martin Liska  <mliska@suse.cz>
2616         * tree-switch-conversion.c (switch_conversion::expand):
2617         Strenghten assumption about gswitch statements.
2619 2018-08-29  Richard Biener  <rguenther@suse.de>
2621         PR tree-optimization/87117
2622         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_stmt): Only
2623         re-value-number released SSA VDEFs.
2625 2018-08-29  Richard Biener  <rguenther@suse.de>
2627         PR tree-optimization/87126
2628         * tree-ssa-sccvn.c (vn_reference_insert): Remove assert.
2630 2018-08-28  Jim Wilson  <jimw@sifive.com>
2632         * config/riscv/pic.md: Rewrite.
2633         * config/riscv/riscv.c (riscv_address_insns): Return cost of 3 for
2634         invalid address.
2635         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Delete.
2636         (SOFTF, default_load, softload, softstore): New.
2638 2018-08-28  Jeff Law  <law@redhat.com>
2640         * fold-const.c (fold_binary_loc): Remove recently added assert.
2642 2018-08-28  Joern Rennecke  <joern.rennecke@riscy-ip.com>
2644         * genpreds.c (write_predicate_subfunction): Also add ATTRIBUTE_UNUSED
2645         to OP parmeter of generated function.
2647 2018-08-28  MCC CS  <deswurstes@users.noreply.github.com>
2649         PR tree-optimization/87009
2650         * match.pd: Add boolean optimizations.
2652 2018-08-28  Martin Sebor  <msebor@redhat.com>
2654         PR middle-end/86631
2655         * calls.c (alloc_max_size): Treat HOST_WIDE_INT special.
2656         * gimple-ssa-warn-alloca.c (adjusted_warn_limit): New function.
2657         (pass_walloca::gate): Use it.
2658         (alloca_call_type): Same.
2659         (pass_walloca::execute): Same.
2660         * stor-layout.c (layout_decl): Treat HOST_WIDE_INT special.
2662 2018-08-28  David Malcolm  <dmalcolm@redhat.com>
2664         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): Change version check on
2665         GCC_VERSION for usage of "__gcc_dump_printf__" format from
2666         >= 3005 to >= 9000.
2668 2018-08-28  Richard Biener  <rguenther@suse.de>
2670         PR tree-optimization/87124
2671         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Guard against
2672         constants before looking up avail.
2674 2018-08-28  Jakub Jelinek  <jakub@redhat.com>
2676         PR middle-end/87099
2677         * calls.c (maybe_warn_nonstring_arg): Punt early if
2678         warn_stringop_overflow is zero.  Don't call get_range_strlen
2679         on 3rd argument, keep iterating until lenrng[1] is INTEGER_CST.
2680         Swap comparison operands to have constants on rhs.  Only use
2681         lenrng[1] if non-NULL and INTEGER_CST.  Don't uselessly
2682         increment lenrng[0].
2684 2018-08-28  Richard Sandiford  <richard.sandiford@arm.com>
2686         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Fix unguarded
2687         use of tree_to_shwi.  Remove duplicated test for the size being
2688         a whole number of bytes.
2690 2018-08-28  Richard Biener  <rguenther@suse.de>
2692         PR tree-optimization/87117
2693         * tree-ssa-sccvn.c (eliminate_dom_walker::eliminate_cleanup):
2694         Handle removed stmt without LHS (GIMPLE_NOP).
2696 2018-08-28  Richard Biener  <rguenther@suse.de>
2698         PR tree-optimization/87117
2699         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Exclude
2700         void which is is_gimple_reg_type by checking for COMPLETE_TYPE_P.
2702 2018-08-28  Richard Biener  <rguenther@suse.de>
2704         PR tree-optimization/87117
2705         * tree-ssa-pre.c (compute_avail): Do not make expressions
2706         with predicated values available.
2707         (get_expr_value_id): Assert we do not run into predicated value
2708         expressions.
2710 2018-08-28  Richard Biener  <rguenther@suse.de>
2712         PR tree-optimization/87117
2713         * tree-ssa-operands.c (add_stmt_operand): STRING_CST may
2714         get virtual operands.
2715         (get_expr_operands): Handle STRING_CST like other decls.
2717 2018-08-28  Martin Liska  <mliska@suse.cz>
2719         * tree.h: Update documentation of fndecl_built_in_p
2720         functions.
2723 2018-08-27  Jeff Law  <law@redhat.com>
2724         PR tree-optimization/87110
2725         * tree-ssa-dse.c (compute_trims): Handle non-constant
2726         TYPE_SIZE_UNIT.
2728 2018-08-27  Martin Sebor  <msebor@redhat.com>
2730         PR tree-optimization/86914
2731         * tree-ssa-strlen.c (maybe_set_strlen_range): Avoid MEM_REF.
2733 2018-08-27  Martin Sebor  <msebor@redhat.com>
2735         PR tree-optimization/87112
2736         * builtins.c (expand_builtin_strnlen): Convert c_strlen result to
2737         the type of the bound argument.
2739 2018-08-27  Jeff Law  <law@redhat.com>
2741         * tree-ssa-dse.c (compute_trims): Handle case where the reference's
2742         type does not have a TYPE_SIZE_UNIT.
2744 2018-08-27  Steve Ellcey  <sellcey@cavium.com>
2746         * config/aarch64/aarch64-speculation.cc: Replace include of cfg.h
2747         with include of backend.h.
2749 2018-08-27  Richard Biener  <rguenther@suse.de>
2751         PR tree-optimization/86927
2752         * tree-vect-loop.c (vect_create_epilog_for_reduction): Properly
2753         use const cond reduction code.
2755 2018-08-27  Alexander Monakov  <amonakov@ispras.ru>
2757         PR tree-optimization/85758
2758         * match.pd ((X & Y) ^ Y): Add :s qualifier to inner expression.
2760 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
2762         PR c++/87091
2763         * diagnostic-show-locus.c (class layout_range): Update for
2764         conversion of show_caret_p to a tri-state.
2765         (layout_range::layout_range): Likewise.
2766         (make_range): Likewise.
2767         (layout::maybe_add_location_range): Likewise.
2768         (layout::should_print_annotation_line_p): Don't show annotation
2769         lines for ranges that are SHOW_LINES_WITHOUT_RANGE.
2770         (layout::get_state_at_point): Update for conversion of
2771         show_caret_p to a tri-state.  Bail out early for
2772         SHOW_LINES_WITHOUT_RANGE, so that such ranges don't affect
2773         underlining or source colorization.
2774         (gcc_rich_location::add_location_if_nearby): Update for conversion
2775         of show_caret_p to a tri-state.
2776         (selftest::test_one_liner_multiple_carets_and_ranges): Likewise.
2777         (selftest::test_one_liner_fixit_replace_equal_secondary_range):
2778         Likewise.
2779         (selftest::test_one_liner_labels): Likewise.
2780         * gcc-rich-location.c (gcc_rich_location::add_expr): Update for
2781         conversion of show_caret_p to a tri-state.
2782         * pretty-print.c (text_info::set_location): Likewise.
2783         * pretty-print.h (text_info::set_location): Likewise.
2784         * substring-locations.c (format_warning_n_va): Likewise.
2785         * tree-diagnostic.c (default_tree_printer): Likewise.
2786         * tree-pretty-print.c (newline_and_indent): Likewise.
2788 2018-08-27  David Malcolm  <dmalcolm@redhat.com>
2790         PR c++/87091
2791         * diagnostic-show-locus.c (get_line_span_for_fixit_hint): Show the
2792         line above for line-insertion fix-it hints.
2793         (selftest::test_fixit_insert_containing_newline): Update the
2794         expected results, and add a test with line-numbering enabled.
2796 2018-08-27  Martin Liska  <mliska@suse.cz>
2798         PR sanitizer/86962
2799         * sanopt.c (sanitize_rewrite_addressable_params): Ignore
2800         params with DECL_HAS_VALUE_EXPR_P.
2802 2018-08-27  Martin Liska  <mliska@suse.cz>
2804         * config/i386/i386.c (ix86_expand_set_or_movmem): Dump
2805         selected expansion strategy.
2807 2018-08-27  Martin Liska  <mliska@suse.cz>
2809         * builtins.h (is_builtin_fn): Remove and fndecl_built_in_p.
2810         * builtins.c (is_builtin_fn): Likewise.
2811         * attribs.c (diag_attr_exclusions): Use new function
2812         fndecl_built_in_p and remove check for FUNCTION_DECL if
2813         possible.
2814         (builtin_mathfn_code): Likewise.
2815         (fold_builtin_expect): Likewise.
2816         (fold_call_expr): Likewise.
2817         (fold_builtin_call_array): Likewise.
2818         (fold_call_stmt): Likewise.
2819         (set_builtin_user_assembler_name): Likewise.
2820         (is_simple_builtin): Likewise.
2821         * calls.c (gimple_alloca_call_p): Likewise.
2822         (maybe_warn_nonstring_arg): Likewise.
2823         * cfgexpand.c (expand_call_stmt): Likewise.
2824         * cgraph.c (cgraph_update_edges_for_call_stmt_node): Likewise.
2825         (cgraph_edge::verify_corresponds_to_fndecl): Likewise.
2826         (cgraph_node::verify_node): Likewise.
2827         * cgraphclones.c (build_function_decl_skip_args): Likewise.
2828         (cgraph_node::create_clone): Likewise.
2829         * config/arm/arm.c (arm_insert_attributes): Likewise.
2830         * config/i386/i386.c (ix86_gimple_fold_builtin): Likewise.
2831         * dse.c (scan_insn): Likewise.
2832         * expr.c (expand_expr_real_1): Likewise.
2833         * fold-const.c (operand_equal_p): Likewise.
2834         (fold_binary_loc): Likewise.
2835         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
2836         * gimple-low.c (lower_stmt): Likewise.
2837         * gimple-pretty-print.c (dump_gimple_call): Likewise.
2838         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
2839         Likewise.
2840         * gimple.c (gimple_build_call_from_tree): Likewise.
2841         (gimple_call_builtin_p): Likewise.
2842         (gimple_call_combined_fn): Likewise.
2843         * gimplify.c (gimplify_call_expr): Likewise.
2844         (gimple_boolify): Likewise.
2845         (gimplify_modify_expr): Likewise.
2846         (gimplify_addr_expr): Likewise.
2847         * hsa-gen.c (gen_hsa_insns_for_call): Likewise.
2848         * ipa-cp.c (determine_versionability): Likewise.
2849         * ipa-fnsummary.c (compute_fn_summary): Likewise.
2850         * ipa-param-manipulation.c (ipa_modify_formal_parameters): Likewise.
2851         * ipa-split.c (visit_bb): Likewise.
2852         (split_function): Likewise.
2853         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
2854         * lto-cgraph.c (input_node): Likewise.
2855         * lto-streamer-out.c (write_symbol): Likewise.
2856         * omp-low.c (setjmp_or_longjmp_p): Likewise.
2857         (lower_omp_1): Likewise.
2858         * predict.c (strip_predict_hints): Likewise.
2859         * print-tree.c (print_node): Likewise.
2860         * symtab.c (symtab_node::output_to_lto_symbol_table_p): Likewise.
2861         * trans-mem.c (is_tm_irrevocable): Likewise.
2862         (is_tm_load): Likewise.
2863         (is_tm_simple_load): Likewise.
2864         (is_tm_store): Likewise.
2865         (is_tm_simple_store): Likewise.
2866         (is_tm_abort): Likewise.
2867         (tm_region_init_1): Likewise.
2868         * tree-call-cdce.c (gen_shrink_wrap_conditions): Likewise.
2869         * tree-cfg.c (verify_gimple_call): Likewise.
2870         (move_stmt_r): Likewise.
2871         (stmt_can_terminate_bb_p): Likewise.
2872         * tree-eh.c (lower_eh_constructs_2): Likewise.
2873         * tree-if-conv.c (if_convertible_stmt_p): Likewise.
2874         * tree-inline.c (remap_gimple_stmt): Likewise.
2875         (copy_bb): Likewise.
2876         (estimate_num_insns): Likewise.
2877         (fold_marked_statements): Likewise.
2878         * tree-sra.c (scan_function): Likewise.
2879         * tree-ssa-ccp.c (surely_varying_stmt_p): Likewise.
2880         (optimize_stack_restore): Likewise.
2881         (pass_fold_builtins::execute): Likewise.
2882         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
2883         (mark_all_reaching_defs_necessary_1): Likewise.
2884         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Likewise.
2885         * tree-ssa-forwprop.c (simplify_builtin_call): Likewise.
2886         (pass_forwprop::execute): Likewise.
2887         * tree-ssa-loop-im.c (stmt_cost): Likewise.
2888         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
2889         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Likewise.
2890         * tree-ssa-strlen.c (get_string_length): Likewise.
2891         * tree-ssa-structalias.c (handle_lhs_call): Likewise.
2892         (find_func_aliases_for_call): Likewise.
2893         * tree-ssa-ter.c (find_replaceable_in_bb): Likewise.
2894         * tree-stdarg.c (optimize_va_list_gpr_fpr_size): Likewise.
2895         * tree-tailcall.c (find_tail_calls): Likewise.
2896         * tree.c (need_assembler_name_p): Likewise.
2897         (free_lang_data_in_decl): Likewise.
2898         (get_call_combined_fn): Likewise.
2899         * ubsan.c (is_ubsan_builtin_p): Likewise.
2900         * varasm.c (incorporeal_function_p): Likewise.
2901         * tree.h (DECL_BUILT_IN): Remove and replace with
2902         fndecl_built_in_p.
2903         (DECL_BUILT_IN_P): Transfort to fndecl_built_in_p.
2904         (fndecl_built_in_p): New.
2906 2018-08-27  Martin Liska  <mliska@suse.cz>
2908         PR tree-optimization/86847
2909         * tree-switch-conversion.c (switch_decision_tree::dump_case_nodes):
2910         Dump also subtree probability.
2911         (switch_decision_tree::do_jump_if_equal): New function.
2912         (switch_decision_tree::emit_case_nodes): Handle special
2913         situations in balanced tree that can be emitted much simpler.
2914         Fix calculation of probabilities that happen in tree expansion.
2915         * tree-switch-conversion.h (struct cluster): Add
2916         is_single_value_p.
2917         (struct simple_cluster): Likewise.
2918         (struct case_tree_node): Add new function has_child.
2919         (do_jump_if_equal): New.
2921 2018-08-27  Martin Liska  <mliska@suse.cz>
2923         * tree-switch-conversion.c (bit_test_cluster::find_bit_tests):
2924         Add new argument to bit_test_cluster constructor.
2925         (bit_test_cluster::emit): Set bits really number of values
2926         handlel by a test.
2927         (bit_test_cluster::hoist_edge_and_branch_if_true): Add
2928         probability argument.
2929         * tree-switch-conversion.h (struct bit_test_cluster):
2930         Add m_handles_entire_switch.
2932 2018-08-27  Martin Liska  <mliska@suse.cz>
2934         PR tree-optimization/86702
2935         * tree-switch-conversion.c (jump_table_cluster::emit):
2936         Make probabilities even for values in jump table
2937         according to number of cases handled.
2938         (switch_decision_tree::compute_cases_per_edge): Pass
2939         argument to reset_out_edges_aux function.
2940         (switch_decision_tree::analyze_switch_statement): Likewise.
2941         * tree-switch-conversion.h (switch_decision_tree::reset_out_edges_aux):
2942         Make it static.
2944 2018-08-27  Martin Liska  <mliska@suse.cz>
2946         * cfgexpand.c (expand_asm_stmt): Use label_to_block and pass
2947         cfun argument explicitly.
2948         * gimple-pretty-print.c (dump_gimple_switch): Likewise.
2949         * hsa-gen.c (gen_hsa_insns_for_switch_stmt): Use new
2950         function gimple_switch_default_bb.
2951         (convert_switch_statements):
2952         (expand_builtins):
2953         * ipa-fnsummary.c (set_switch_stmt_execution_predicate):
2954         * stmt.c (label_to_block_fn): Use label_to_block and pass
2955         cfun argument explicitly and use gimple_switch_label_bb.
2956         (expand_case): Likewise.
2957         * tree-cfg.c (lower_phi_internal_fn): Use label_to_block and pass
2958         cfun argument explicitly. Likewise.
2959         (make_edges_bb): Likewise.
2960         (make_cond_expr_edges): Likewise.
2961         (get_cases_for_edge): Likewise.
2962         (make_gimple_switch_edges): Likewise.
2963         (label_to_block_fn): Likewise.
2964         (label_to_block): Likewise.
2965         (make_goto_expr_edges): Likewise.
2966         (make_gimple_asm_edges): Likewise.
2967         (main_block_label): Likewise.
2968         (group_case_labels_stmt): Likewise.
2969         (find_taken_edge_computed_goto): Likewise.
2970         (find_taken_edge_switch_expr): Likewise.
2971         (gimple_verify_flow_info): Likewise.
2972         (gimple_redirect_edge_and_branch): Likewise.
2973         (gimple_switch_label_bb): New function.
2974         (gimple_switch_default_bb): Likewise.
2975         (gimple_switch_edge): Likewise.
2976         (gimple_switch_default_edge): Likewise.
2977         * tree-cfg.h (label_to_block_fn): Remove and replace ...
2978         (label_to_block): ... with this.
2979         (gimple_switch_label_bb): New.
2980         (gimple_switch_default_bb): Likewise.
2981         (gimple_switch_edge): Likewise.
2982         (gimple_switch_default_edge): Likewise.
2983         * tree-cfgcleanup.c (convert_single_case_switch): Use
2984         new gimple functions and pass new argument to label_to_block.
2985         (cleanup_control_flow_bb):
2986         * tree-eh.c (make_eh_dispatch_edges): Use label_to_block and pass
2987         cfun argument explicitly.
2988         (make_eh_edges): Likewise.
2989         (redirect_eh_dispatch_edge): Likewise.
2990         (lower_resx): Likewise.
2991         (lower_eh_dispatch): Likewise.
2992         (maybe_remove_unreachable_handlers): Likewise.
2993         (unsplit_eh): Likewise.
2994         (cleanup_empty_eh): Likewise.
2995         (verify_eh_edges): Likewise.
2996         (verify_eh_dispatch_edge): Likewise.
2997         * tree-ssa-dom.c (record_edge_info): Likewise.
2998         * tree-ssa-forwprop.c (simplify_gimple_switch_label_vec): Likewise.
2999         * tree-ssa-threadedge.c (thread_around_empty_blocks): Likewise.
3000         (thread_through_normal_block): Likewise.
3001         * tree-ssa-uncprop.c (associate_equivalences_with_edges): Likewise.
3002         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds):
3003         * tree-switch-conversion.c (switch_conversion::collect): Use new
3004         gimple functions.
3005         (switch_conversion::check_final_bb): Likewise.
3006         (switch_conversion::gather_default_values): Pass new argument
3007         to label_to_block.
3008         (switch_conversion::build_constructors): Likewise.
3009         (switch_decision_tree::compute_cases_per_edge): Use new
3010         gimple_switch_edge function.
3011         (switch_decision_tree::analyze_switch_statement): Pass new argument
3012         to label_to_block.
3013         (switch_decision_tree::try_switch_expansion): Use
3014         gimple_switch_default_edge.
3015         * tree-vrp.c (find_switch_asserts): Pass new argument
3016         to label_to_block.
3017         * vr-values.c (vr_values::vrp_visit_switch_stmt): Likewise.
3018         (vr_values::simplify_switch_using_ranges): Likewise.
3020 2018-08-27  Richard Biener  <rguenther@suse.de>
3022         * cfganal.h (rev_post_order_and_mark_dfs_back_seme): Declare.
3023         * cfganal.c (rev_post_order_and_mark_dfs_back_seme): New function.
3025         * tree-ssa-sccvn.h (struct vn_pval): New structure.
3026         (struct vn_nary_op_s): Add unwind_to member.  Add
3027         predicated_values flag and put result into a union together
3028         with a linked list of vn_pval.
3029         (struct vn_ssa_aux): Add name member to make maintaining
3030         a map of SSA name to vn_ssa_aux possible.  Remove no longer
3031         needed info, dfsnum, low, visited, on_sccstack, use_processed
3032         and range_info_anti_range_p members.
3033         (run_scc_vn, vn_eliminate, free_scc_vn, vn_valueize): Remove.
3034         (do_rpo_vn, run_rpo_vn, eliminate_with_rpo_vn, free_rpo_vn):
3035         New functions.
3036         (vn_valueize): New global.
3037         (vn_context_bb): Likewise.
3038         (VN_INFO_RANGE_INFO, VN_INFO_ANTI_RANGE_P, VN_INFO_RANGE_TYPE,
3039         VN_INFO_PTR_INFO): Remove.
3040         * tree-ssa-sccvn.c: ... (rewrite)
3041         (pass_fre::execute): For -O2+ initialize loops and run
3042         RPO VN in optimistic mode (iterating).  For -O1 and -Og
3043         run RPO VN in non-optimistic mode.
3044         * params.def (PARAM_SCCVN_MAX_SCC_SIZE): Remove.
3045         (PARAM_RPO_VN_MAX_LOOP_DEPTH): Add.
3046         * doc/invoke.texi (sccvn-max-scc-size): Remove.
3047         (rpo-vn-max-loop-depth): Document.
3048         * tree-ssa-alias.c (walk_non_aliased_vuses): Stop walking
3049         when valuezing the VUSE signals we walked out of the region.
3050         * tree-ssa-pre.c (phi_translate_1): Ignore predicated values.
3051         (phi_translate): Set VN context block to use for availability
3052         lookup.
3053         (compute_avail): Likewise.
3054         (pre_valueize): New function.
3055         (pass_pre::execute): Adjust to the RPO VN API.
3057         * tree-ssa-loop-ivcanon.c: Include tree-ssa-sccvn.h.
3058         (propagate_constants_for_unrolling): Remove.
3059         (tree_unroll_loops_completely): Perform value-numbering
3060         on the unrolled bodies loop parent.
3062 2018-08-27  Richard Biener  <rguenther@suse.de>
3064         * tree-ssa-pre.c (compute_antic): Re-use inverted postorder
3065         for partial antic compute.
3067 2018-08-27  Jakub Jelinek  <jakub@redhat.com>
3069         PR rtl-optimization/87065
3070         * combine.c (simplify_if_then_else): Formatting fix.
3071         (if_then_else_cond): Guard MULT optimization with SCALAR_INT_MODE_P
3072         check.
3073         (known_cond): Don't return const_true_rtx for vector modes.  Use
3074         CONST0_RTX instead of const0_rtx.  Formatting fixes.
3076 2018-08-27  Martin Liska  <mliska@suse.cz>
3078         PR gcov-profile/87069
3079         * gcov.c (process_file): Record files already processed
3080         and warn about a file being processed multiple times.
3082 2018-08-27  Martin Liska  <mliska@suse.cz>
3084         PR driver/83193
3085         * config/aarch64/aarch64.c (aarch64_override_options_internal):
3086         Set default values for x_aarch64_*_string strings.
3087         * config/aarch64/aarch64.opt: Remove --{march,mcpu,mtune}==
3088         prefix.  For -mabi do not print '=ABI' in help and use
3089         <option_value> format for -msve-vector-bits and -moverride
3090         options.
3092 2018-08-26  Jeff Law  <law@redhat.com>
3094         * config/mips/frame-header-opt.c: Include "backend.h" rather than
3095         "cfg.h"
3097 2018-08-26  Marek Polacek  <polacek@redhat.com>
3099         PR c++/87029, Implement -Wredundant-move.
3100         * doc/invoke.texi: Document -Wredundant-move.
3102 2018-08-25  Martin Sebor  <msebor@redhat.com>
3104         PR tree-optimization/87059
3105         * builtins.c (expand_builtin_strncmp): Convert MIN_EXPR operand
3106         to the same type as the other.
3107         * fold-const.c (fold_binary_loc): Assert expectation.
3109 2018-08-25  Iain Sandoe  <iain@sandoe.co.uk>
3111         * config/darwin.c (machopic_legitimize_pic_address): Clean up
3112         extraneous parentheses, dead code section and formatting.
3114 2018-08-24  David Malcolm  <dmalcolm@redhat.com>
3116         PR c++/87091
3117         * diagnostic-show-locus.c (layout::layout): Ensure the margin is
3118         wide enough for jumps in the line-numbering to be visible.
3119         (layout::print_gap_in_line_numbering): New member function.
3120         (layout::calculate_line_spans): When using line numbering, merge
3121         line spans that are only 1 line apart.
3122         (diagnostic_show_locus): When printing line numbers, show gaps in
3123         line numbering directly, rather than printing headers.
3124         (selftest::test_diagnostic_show_locus_fixit_lines): Add test of
3125         line-numbering with multiple line spans.
3126         (selftest::test_fixit_insert_containing_newline_2): Add test of
3127         line-numbering, in which the spans are close enough to be merged.
3129 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
3131         * gimple-ssa-evrp-analyze.c (set_ssa_range_info): Pass value_range
3132         to range_includes_zero_p.  Do not special case VR_ANTI_RANGE.
3133         * tree-vrp.c (range_is_nonnull): Remove.
3134         (range_includes_zero_p): Accept value_range instead of min/max.
3135         (extract_range_from_binary_expr_1): Do not early bail on
3136         POINTER_PLUS_EXPR.
3137         Use range_includes_zero_p instead of range_is_nonnull.
3138         (extract_range_from_unary_expr): Use range_includes_zero_p instead
3139         of range_is_nonnull.
3140         (vrp_meet_1): Pass value_range to range_includes_zero_p.  Do not
3141         special case VR_ANTI_RANGE.
3142         (vrp_finalize): Same.
3143         * tree-vrp.h (range_includes_zero_p): Pass value_range as argument
3144         instead of min/max.
3145         (range_is_nonnull): Remove.
3146         * vr-values.c (vrp_stmt_computes_nonzero): Use
3147         range_includes_zero_p instead of range_is_nonnull.
3148         (extract_range_basic): Pass value_range to range_includes_zero_p
3149         instead of range_is_nonnull.
3151 2018-08-24  Uros Bizjak  <ubizjak@gmail.com>
3153         * emit-rtl.c (init_emit_once): Do not emit MODE_POINTER_BOUNDS RTXes.
3154         * emit-rtl.h (rtl_data): Remove return_bnd.
3155         * explow.c (trunc_int_for_mode): Do not handle POINTER_BOUNDS_MODE_P.
3156         * function.c (diddle_return_value): Do not handle crtl->return_bnd.
3157         * genmodes.c (complete_mode): Do not handle MODE_POINTER_BOUNDS.
3158         (POINTER_BOUNDS_MODE): Remove definition.
3159         (make_pointer_bounds_mode): Remove.
3160         (get_mode_class): Do not handle MODE_POINTER_BOUNDS.
3161         * machmode.h (POINTER_BOUNDS_MODE_P): Remove definition.
3162         (scalare_mode::includes_p): Do not handle MODE_POINTER_BOUNDS.
3163         * mode-classes.def: Do not define MODE_POINTER_BOUNDS.
3164         * stor-layout.c (int_mode_for_mode): Do not handle MODE_POINTER_BOUNDS.
3165         * tree-core.h (enum tree_index): Remove TI_POINTER_BOUNDS_TYPE.
3166         * varasm.c (output_constant_pool_2): Do not handle MODE_POINTER_BOUNDS.
3168         * config/i386/i386-modes.def (BND32, BND64): Remove.
3169         * config/i386/i386.c (dbx_register_map): Remove bound registers.
3170         (dbx64_register_map): Ditto.
3171         (svr4_dbx_register_map): Ditto.
3172         (indirect_thunk_bnd_needed): Remove.
3173         (indirect_thunks_bnd_used): Ditto.
3174         (indirect_return_bnd_needed): Ditto.
3175         (indirect_return_via_cx_bnd): Ditto.
3176         (enum indirect_thunk_prefix): Remove indirect_thunk_prefix_bnd.
3177         (indirect_thunk_name): Remove handling of indirect_thunk_prefix_bnd.
3178         (output_indirect_thunk): Ditto.  Remove need_prefix argument.
3179         (output_indirect_thunk_function): Remove handling of
3180         indirect_return_bnd_needed, indirect_return_via_cx_bnd,
3181         indirect_thunk_bnd_needed and indirect_thunks_bnd_used variables.
3182         (ix86_save_reg): Remove handling of crtl->return_bnd.
3183         (ix86_legitimate_constant_p): Remove handling of POINTER_BOUNDS_MODE_P.
3184         (ix86_print_operand_address_as): Remove handling of UNSPEC_BNDMK_ADDR
3185         and UNSPEC_BNDLX_ADDR.
3186         (ix86_output_indirect_branch_via_reg): Remove handling of
3187         indirect_thunk_prefix_bnd.
3188         (ix86_output_indirect_branch_via_push): Ditto.
3189         (ix86_output_function_return): Ditto.
3190         (ix86_output_indirect_function_return): Ditto.
3191         (avoid_func_arg_motion): Do not handle UNSPEC_BNDSTX.
3192         * config/i386/i386.h (FIXED_REGISTERS): Remove bound registers.
3193         (CALL_USED_REGISTERS): Ditto.
3194         (REG_ALLOC_ORDER): Update for removal of bound registers.
3195         (HI_REGISTER_NAMES): Ditto.
3196         * config/i386/i386.md (UNSPEC_BNDMK, UNSPEC_BNDMK_ADDR, UNSPEC_BNDSTX)
3197         (UNSPEC_BNDLDX, UNSPEC_BNDLDX_ADDR, UNSPEC_BNDCL, UNSPEC_BNDCU)
3198         (UNSPEC_BNDCN, UNSPEC_MPX_FENCE): Remove.
3199         (BND0_REG, BND1_REG, BND2_REG, BND3_REG): Remove
3200         (FIRST_PSEUDO_REG): Update.
3201         (BND): Remove mode iterator.
3202         * config/i386/predicates.md (bnd_mem_operator): Remove.
3204 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
3206         * tree-vect-stmts.c (vectorizable_bswap): Handle variable-length
3207         vectors.
3209 2018-08-24  Richard Sandiford  <richard.sandiford@arm.com>
3211         * tree-vect-slp.c (vect_transform_slp_perm_load): Separate out
3212         the case in which the permute needs only a single element and
3213         repeats for every vector of the result.  Extend that case to
3214         handle variable-length vectors.
3215         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
3217 2018-08-24  H.J. Lu  <hongjiu.lu@intel.com>
3219         PR debug/79342
3220         * dwarf2out.c (save_macinfo_strings): Call set_indirect_string
3221         on DW_MACINFO_start_file for -gsplit-dwarf -g3.
3223 2018-08-24  Richard Biener  <rguenther@suse.de>
3225         * cfg.h (struct control_flow_graph): Add edge_flags_allocated and
3226         bb_flags_allocated members.
3227         (auto_flag): New RAII class for allocating flags.
3228         (auto_edge_flag): New RAII class for allocating edge flags.
3229         (auto_bb_flag): New RAII class for allocating bb flags.
3230         * cfgloop.c (verify_loop_structure): Allocate temporary edge
3231         flag dynamically.
3232         * cfganal.c (dfs_enumerate_from): Remove use of visited sbitmap
3233         in favor of temporarily allocated BB flag.
3234         * hsa-brig.c: Re-order includes.
3235         * hsa-dump.c: Likewise.
3236         * hsa-regalloc.c: Likewise.
3237         * print-rtl.c: Likewise.
3238         * profile-count.c: Likewise.
3240 2018-08-24  Segher Boessenkool  <segher@kernel.crashing.org>
3242         PR target/86989
3243         * config/rs6000/rs6000.c (toc_relative_expr_p): Check that the base is
3244         the TOC register.
3246 2018-08-24  Aldy Hernandez  <aldyh@redhat.com>
3248         PR 87073/bootstrap
3249         * wide-int-range.cc (wide_int_range_div): Do not ignore result
3250         from wide_int_range_multiplicative_op.
3252 2018-08-23  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
3254         * tree-vect-data-refs.c (vect_grouped_store_supported): Fix typo
3255         "permutaion".
3257 2018-08-23  Giuliano Belinassi  <giuliano.belinassi@usp.br>
3259         * genmatch.c (parser::parse_operation): Fix typo 'exapnded'
3260         to 'expanded'.
3262 2018-08-23  Alexander Monakov  <amonakov@ispras.ru>
3264         * tree-scalar-evolution.c (final_value_replacement_loop): Dump
3265         full GENERIC expression used for replacement.
3267 2018-08-23  Aldy Hernandez  <aldyh@redhat.com>
3269         * tree-vrp.c (abs_extent_range): Remove.
3270         (extract_range_into_wide_ints): Pass wide ints by reference.
3271         (extract_range_from_binary_expr_1): Rewrite the *DIV_EXPR code.
3272         Pass wide ints by reference in all calls to
3273         extract_range_into_wide_ints.
3274         * wide-int-range.cc (wide_int_range_div): New.
3275         * wide-int-range.h (wide_int_range_div): New.
3276         (wide_int_range_includes_zero_p): New.
3277         (wide_int_range_zero_p): New.
3279 2018-08-23  Matthew Malcomson  <matthew.malcomson@arm.com>
3281         * config/aarch64/aarch64.md (arches): New enum.
3282         (arch): New enum attr.
3283         (arch_enabled): New attr.
3284         (enabled): Now uses arch_enabled only.
3285         (simd, sve, fp16): Removed attribute.
3286         (fp): Attr now defined in terms of 'arch'.
3287         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64, *movti_aarch64,
3288         *movhf_aarch64, <optab><fcvt_target><GPF:mode>2,
3289         <FCVT_F2FIXED:fcvt_fixed_insn><GPF:mode>3,
3290         <FCVT_FIXED2F:fcvt_fixed_insn><GPI:mode>3): Merge 'fp' and 'simd'
3291         attributes into 'arch'.
3292         (*movsf_aarch64, *movdf_aarch64, *movtf_aarch64, *add<mode>3_aarch64,
3293         subdi3, neg<mode>2, <optab><mode>3, one_cmpl<mode>2,
3294         *<NLOGICAL:optab>_one_cmpl<mode>3, *xor_one_cmpl<mode>3,
3295         *aarch64_ashl_sisd_or_int_<mode>3, *aarch64_lshr_sisd_or_int_<mode>3,
3296         *aarch64_ashr_sisd_or_int_<mode>3, *aarch64_sisd_ushl): Convert use of
3297         'simd' attribute into 'arch'.
3298         (load_pair_sw_<SX:mode><SX2:mode>, load_pair_dw_<DX:mode><DX2:mode>,
3299         store_pair_sw_<SX:mode><SX2:mode>, store_pair_dw_<DX:mode><DX2:mode>):
3300         Convert use of 'fp' attribute to 'arch'.
3301         * config/aarch64/aarch64-simd.md (move_lo_quad_internal_<mode>,
3302         move_lo_quad_internal_<mode>): (different modes) Merge 'fp' and 'simd'
3303         into 'arch'.
3304         (move_lo_quad_internal_be_<mode>, move_lo_quad_internal_be_<mode>):
3305         (different modes) Merge 'fp' and 'simd' into 'arch'.
3306         (*aarch64_combinez<mode>, *aarch64_combinez_be<mode>): Merge 'fp' and
3307         'simd' into 'arch'.
3309 2018-08-23  Segher Boessenkool  <segher@kernel.crashing.org>
3311         PR rtl-optimization/87026
3312         * expmed.c (canonicalize_comparison): If we can no longer create
3313         pseudoregisters, don't.
3315 2018-08-23  Richard Earnshaw  <rearnsha@arm.com>
3317         PR target/86951
3318         * config/arm/arm-protos.h (arm_emit_speculation_barrier): New
3319         prototype.
3320         * config/arm/arm.c (speculation_barrier_libfunc): New static
3321         variable.
3322         (arm_init_libfuncs): Initialize it.
3323         (arm_emit_speculation_barrier): New function.
3324         * config/arm/arm.md (speculation_barrier): Call
3325         arm_emit_speculation_barrier for architectures that do not have
3326         DSB or ISB.
3327         (speculation_barrier_insn): Only match on Armv7 or later.
3329 2018-08-23  Richard Biener  <rguenther@suse.de>
3331         PR middle-end/87024
3332         * tree-inline.c (copy_bb): Drop unused __builtin_va_arg_pack_len
3333         calls.
3335 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
3337         * config/aarch64/aarch64.c (aarch64_evpc_sve_tbl): Fix handling
3338         of single-vector TBLs.
3339         (aarch64_vectorize_vec_perm_const): Set one_vector_p when only
3340         one input is given.
3342 2018-08-23  Richard Sandiford  <richard.sandiford@arm.com>
3344         PR target/85910
3345         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const_1): Fix
3346         aarch64_evpc_tbl guard.
3348 2018-08-22  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3350         * tree-ssa-dse.c (compute_trims): Avoid folding away undefined
3351         behaviour.
3353 2018-08-22  Martin Sebor  <msebor@redhat.com>
3355         PR middle-end/87052
3356         * tree-pretty-print.c (pretty_print_string): Add argument.
3357         (dump_generic_node): Call to pretty_print_string with string size.
3359 2018-08-22  Segher Boessenkool  <segher@kernel.crashing.org>
3361         PR rtl-optimization/86771
3362         * combine.c (try_combine): Do not allow splitting a resulting PARALLEL
3363         of two SETs into those two SETs, one to be placed at i2, if that SETs
3364         destination is modified between i2 and i3.
3366 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
3368         PR tree-optimization/86725
3369         * tree-vect-loop.c (vect_inner_phi_in_double_reduction_p): New
3370         function.
3371         (vect_analyze_scalar_cycles_1): Check it.
3373 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
3375         PR tree-optimization/86725
3376         * tree-vect-loop.c (vect_is_simple_reduction): When treating
3377         an outer loop phi as a double reduction, make sure that the
3378         single user of the phi result is an inner loop phi.
3380 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
3382         * tree-vect-data-refs.c (vect_analyze_group_access_1): Convert
3383         grouped stores with gaps to a strided group.
3385 2018-08-22  Richard Sandiford  <richard.sandiford@arm.com>
3387         * tree-vect-stmts.c (get_group_load_store_type)
3388         (get_load_store_type): Only test STMT_VINFO_STRIDED_P for the
3389         first statement in a group.
3391 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
3393         * config/darwin.h (LINK_COMMAND_SPEC_A): Sync LTO options with
3394         the sequence used in gcc/gcc.c.
3396 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
3398         PR other/704
3399         * gcc-ar.c (main): Don’t try to invoke the plug-in if we’re not
3400         building it.
3402 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
3404         * config/darwin10.h (LINK_GCC_C_SEQUENCE_SPEC): Adjust to use the
3405         Darwin10-specific unwinder-shim.
3406         * config/darwin12.h (LINK_GCC_C_SEQUENCE_SPEC): Remove.
3407         * config/rs6000/darwin.h (DARWIN_CRT1_SPEC, DARWIN_DYLIB1_SPEC):
3408         New to cater for Darwin10 Rosetta.
3410 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
3412         * config/i386/i386.c (ix86_output_addr_diff_elt): Move the MACH-O
3413         specific test before the one for HAVE_AS_GOTOFF_IN_DATA.
3415 2018-08-22  Iain Sandoe  <iain@sandoe.co.uk>
3417         PR bootstrap/81033
3418         PR target/81733
3419         PR target/52795
3420         * gcc/dwarf2out.c (FUNC_SECOND_SECT_LABEL): New.
3421         (dwarf2out_switch_text_section): Generate a local label for the
3422         second function sub-section and apply it as the second FDE start
3423         label.
3424         * gcc/final.c (final_scan_insn_1): Emit second FDE label after the
3425         second sub-section start.
3427 2018-08-22  Richard Biener  <rguenther@suse.de>
3429         PR tree-optimization/86988
3430         * tree-vrp.c (vrp_prop::check_mem_ref): Bail out on VLAs.
3432 2018-08-22  Richard Biener  <rguenther@suse.de>
3434         PR tree-optimization/86945
3435         * tree-cfg.c (generate_range_test): Use unsigned arithmetic.
3437 2018-08-22  Alexandre Oliva <oliva@adacore.com>
3439         * config/rs6000/rs6000.c (SMALL_DATA_RELOC, SMALL_DATA_REG): Add
3440         a comment about how uses of r2 for .sdata2 come about.
3442 2018-08-22  Alexandre Oliva <aoliva@redhat.com>
3444         * tree-ssa-reassoc.c (is_reassociable_op): Fix cut&pasto.
3446 2018-08-21  Marek Polacek  <polacek@redhat.com>
3448         PR c++/86981, Implement -Wpessimizing-move.
3449         * doc/invoke.texi: Document -Wpessimizing-move.
3451 2018-08-21  Jan Hubicka  <jh@suse.cz>
3453         * tree.c (find_decls_types_r): Do not check for redundant typedefs.
3454         * tree.h (is_redundant_typedef): Remove.
3455         * dwarf2out.c (is_redundant_typedef): Turn into static function.
3457 2018-08-21  Jan Hubicka  <jh@suse.cz>
3459         * tree.c (free_lang_data_in_decl): Remove types from DECL_CONTEXT
3460         when possible.
3462 2018-08-21  Tamar Christina  <tamar.christina@arm.com>
3464         * expmed.c (extract_low_bits): Reject invalid subregs early.
3466 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3468         PR middle-end/86121
3469         * tree-ssa-strlen.c (adjust_last_stmt): Avoid folding away undefined
3470         behaviour.
3472 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
3474         * config/vxworks.h: Guard vxworks_asm_out_constructor and
3475         vxworks_asm_out_destructor by !HAVE_INITFINI_ARRAY_SUPPORT
3476         * config/vxworks.c: Likewise.
3478 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
3480         * config/vxworks.c: Set targetm.have_ctors_dtors
3481         if HAVE_INITFINI_ARRAY_SUPPORT.
3482         * config/vxworks.h: Set SUPPORTS_INIT_PRIORITY
3483         if HAVE_INITFINI_ARRAY_SUPPORT.
3485 2018-08-21  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
3487         * config/vxworks.h: Add $(WIND_BASE)/target/h/wrn/coreip to
3488         default search path for VxWorks < 7.
3490 2018-08-21  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3492         * gimple-ssa-sprintf.c (decl_constant_value): Remove.
3493         (get_format_string): Refer to c_getstr.
3495 2018-08-21  Tom de Vries  <tdevries@suse.de>
3497         * cgraph.h (debuginfo_early_init, debuginfo_init, debuginfo_fini)
3498         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
3499         (debuginfo_early_stop): Declare.
3500         * cgraphunit.c (debuginfo_early_init, debuginfo_init, debuginfo_fini)
3501         (debuginfo_start, debuginfo_stop, debuginfo_early_start)
3502         (debuginfo_early_stop): New function.
3503         (symbol_table::finalize_compilation_unit): Call debuginfo_early_start
3504         and debuginfo_early_stop.
3505         * dwarf2out.c (dwarf2out_finish, dwarf2out_early_finish): Dump dwarf.
3506         * toplev.c (compile_file): Call debuginfo_start and debuginfo_stop.
3507         (general_init): Call debuginfo_early_init.
3508         (finalize): Call debuginfo_fini.
3509         (do_compile): Call debuginfo_init.
3510         * doc/invoke.texi (@gccoptlist): Add -fdump-debug and
3511         -fdump-early-debug.
3512         (@item -fdump-debug, @item -fdump-earlydebug): Add.
3514 2018-08-21  Tom de Vries  <tdevries@suse.de>
3516         * dwarf2out.c (print_dw_val, print_loc_descr, print_die): Handle
3517         flag_dump_noaddr and flag_dump_unnumbered.
3519 2018-08-21  Aldy Hernandez  <aldyh@redhat.com>
3521         * wide-int-range.cc (wide_int_range_abs): New.
3522         (wide_int_range_order_set): Rename from wide_int_range_min_max.
3523         * wide-int-range.h (wide_int_range_abs): New.
3524         (wide_int_range_min_max): New.
3525         * tree-vrp.c (extract_range_from_unary_expr): Rewrite ABS_EXPR
3526         case to call wide_int_range_abs.
3527         Rewrite MIN/MAX_EXPR to call wide_int_range_min_max.
3528         (extract_range_from_abs_expr): Delete.
3530 2018-08-20  Michael Meissner  <meissner@linux.ibm.com>
3532         PR target/87033
3533         * config/rs6000/rs6000.md (extendsi<mode>2): Change constraints
3534         from 'Y' to 'YZ' to enable the LWAX instruction to be generated
3535         for indexed loads.
3537 2018-08-20  Nathan Sidwell  <nathan@acm.org>
3538             Jeff Law <law@redhat.com>
3540         * config/s390/s390-c (s390_macro_to_expand): Use cpp_macro_p.
3541         * config/spu/spu-c.c (spu_macro_to_expand): Likewise.
3543 2018-08-20  David Malcolm  <dmalcolm@redhat.com>
3545         PR other/84889
3546         * attribs.c (diag_attr_exclusions): Add auto_diagnostic_group instance.
3547         (decl_attributes): Likewise.
3548         * calls.c (maybe_warn_nonstring_arg): Add auto_diagnostic_group
3549         instance.
3550         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3551         * diagnostic-core.h (class auto_diagnostic_group): New class.
3552         * diagnostic.c (diagnostic_initialize): Initialize the new fields.
3553         (diagnostic_report_diagnostic): Handle the first diagnostics within
3554         a group.
3555         (emit_diagnostic): Add auto_diagnostic_group instance.
3556         (inform): Likewise.
3557         (inform_n): Likewise.
3558         (warning): Likewise.
3559         (warning_at): Likewise.
3560         (warning_n): Likewise.
3561         (pedwarn): Likewise.
3562         (permerror): Likewise.
3563         (error): Likewise.
3564         (error_n): Likewise.
3565         (error_at): Likewise.
3566         (sorry): Likewise.
3567         (fatal_error): Likewise.
3568         (internal_error): Likewise.
3569         (internal_error_no_backtrace): Likewise.
3570         (auto_diagnostic_group::auto_diagnostic_group): New ctor.
3571         (auto_diagnostic_group::~auto_diagnostic_group): New dtor.
3572         * diagnostic.h (struct diagnostic_context): Add fields
3573         "diagnostic_group_nesting_depth",
3574         "diagnostic_group_emission_count", "begin_group_cb",
3575         "end_group_cb".
3576         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior):
3577         Add auto_diagnostic_group instance(s).
3578         (find_explicit_erroneous_behavior): Likewise.
3579         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Likewise.
3580         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Likewise.
3581         * gimplify.c (warn_implicit_fallthrough_r): Likewise.
3582         (gimplify_va_arg_expr): Likewise.
3583         * hsa-gen.c (HSA_SORRY_ATV): Likewise.
3584         (HSA_SORRY_AT): Likewise.
3585         * ipa-devirt.c (compare_virtual_tables): Likewise.
3586         (warn_odr): Likewise.
3587         * multiple_target.c (expand_target_clones): Likewise.
3588         * opts-common.c (cmdline_handle_error): Likewise.
3589         * reginfo.c (globalize_reg): Likewise.
3590         * substring-locations.c (format_warning_n_va): Likewise.
3591         * tree-inline.c (expand_call_inline): Likewise.
3592         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Likewise.
3593         * tree-ssa-loop-niter.c
3594         (do_warn_aggressive_loop_optimizations): Likewise.
3595         * tree-ssa-uninit.c (warn_uninit): Likewise.
3596         * tree.c (warn_deprecated_use): Likewise.
3598 2018-08-20  H.J. Lu  <hongjiu.lu@intel.com>
3600         PR target/87014
3601         * config/i386/i386.md (eh_return): Always update EH return
3602         address in word_mode.
3604 2018-08-20  Chung-Lin Tang  <cltang@codesourcery.com>
3606         * targhooks.c (std_gimplify_va_arg_expr): Properly handle case of when
3607         TARGET_SPLIT_COMPLEX_ARG is defined.
3609 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3611         * expr.c (store_field): Change gcc_assert to gcc_checking_assert.
3613 2018-08-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3615         PR target/86984
3616         * expr.c (expand_assignment): Assert that bitpos is positive.
3617         (store_field): Likewise
3618         (expand_expr_real_1): Make sure that bitpos is positive.
3619         * config/alpha/alpha.h (CONSTANT_ADDRESS_P): Avoid signed
3620         integer overflow.
3622 2018-08-20  Nathan Sidwell  <nathan@acm.org>
3624         * Makefile.in (CPP_ID_DATA_H): Delete.
3625         (CPP_INTERNAL_H): Don't add it.
3626         (GTFILES): Replace CPP_ID_DATA_H with CPPLIB_H.
3627         * gengtype.c (open_base_files): Replace cpp-id-data.h with cpplib.h
3629 2018-08-20  Richard Biener  <rguenther@suse.de>
3631         PR tree-optimization/78655
3632         * tree-vrp.c (extract_range_from_binary_expr_1): Make
3633         pointer + offset nonnull if either operand is nonnull work.
3635 2018-08-20  Tom de Vries  <tdevries@suse.de>
3637         * dwarf2out.c (add_scalar_info): Don't add reference to existing die
3638         unless the referenced die describes the added property using
3639         DW_AT_location or DW_AT_const_value.  Fall back to exprloc case.
3640         Otherwise, add a DW_AT_location to the referenced die.
3642 2018-08-19  Uros Bizjak  <ubizjak@gmail.com>
3644         PR target/86994
3645         * config/i386/i386.c (ix86_rtx_costs) [case SET]: Check source for
3646         register_operand when calling ix86_set_reg_reg_cost.
3647         [case CONST_INT, case CONST, case LABEL_REF, case SYMBOL_REF]:
3648         Set *total to 0 for operands that satisfy x86_64_immediate_operand
3649         predicate and to 1 otherwise.
3651 2018-08-18  Iain Sandoe  <iain@sandoe.co.uk>
3653         * config/darwin.c (darwin_override_options): If -gsplit-dwarf is set,
3654         emit a diagnostic that it is not supported and reset the option.
3655         * config/darwin.h (DRIVER_SELF_SPECS): Note that gsplit-dwarf is not
3656         supported and consume the option.  (ASM_FINAL_SPEC): New.
3658 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
3660         * doc/md.texi (Patterns): Use @ref instead of @xref in the middle of
3661         a sentence.
3663 2018-08-17  Sandra Loosemore  <sandra@codesourcery.com>
3665         C-SKY port: Documentation
3667         * doc/extend.texi (C-SKY Function Attributes): New section.
3668         * doc/invoke.texi (Option Summary): Add C-SKY options.
3669         (C-SKY Options): New section.
3670         * doc/md.texi (Machine Constraints): Document C-SKY constraints.
3672 2018-08-17  Jojo  <jijie_rong@c-sky.com>
3673             Huibin Wang  <huibin_wang@c-sky.com>
3674             Sandra Loosemore  <sandra@codesourcery.com>
3675             Chung-Lin Tang  <cltang@codesourcery.com>
3677         C-SKY port: Backend implementation
3679         * config/csky/*: New.
3680         * common/config/csky/*: New.
3682 2018-08-17  Jojo  <jijie_rong@c-sky.com>
3683             Huibin Wang  <huibin_wang@c-sky.com>
3684             Sandra Loosemore  <sandra@codesourcery.com>
3685             Chung-Lin Tang  <cltang@codesourcery.com>
3686             Andrew Jenner  <andrew@codesourcery.com>
3688         C-SKY port: Configury
3690         * config.gcc (csky-*-*): New.
3691         * configure.ac: Add csky to targets for dwarf2 debug_line support.
3692         * configure: Regenerated.
3694 2018-08-17  David Malcolm  <dmalcolm@redhat.com>
3696         * dump-context.h: Include "dumpfile.h".
3697         (dump_context::dump_printf_va): Convert final param from va_list
3698         to va_list *.  Convert from ATTRIBUTE_PRINTF to
3699         ATTRIBUTE_GCC_DUMP_PRINTF.
3700         (dump_context::dump_printf_loc_va): Likewise.
3701         * dumpfile.c: Include "stringpool.h".
3702         (make_item_for_dump_printf_va): Delete.
3703         (make_item_for_dump_printf): Delete.
3704         (class dump_pretty_printer): New class.
3705         (dump_pretty_printer::dump_pretty_printer): New ctor.
3706         (dump_pretty_printer::emit_items): New member function.
3707         (dump_pretty_printer::emit_any_pending_textual_chunks): New member
3708         function.
3709         (dump_pretty_printer::emit_item): New member function.
3710         (dump_pretty_printer::stash_item): New member function.
3711         (dump_pretty_printer::format_decoder_cb): New member function.
3712         (dump_pretty_printer::decode_format): New member function.
3713         (dump_context::dump_printf_va): Reimplement in terms of
3714         dump_pretty_printer.
3715         (dump_context::dump_printf_loc_va): Convert final param from va_list
3716         to va_list *.
3717         (dump_context::begin_scope): Reimplement call to
3718         make_item_for_dump_printf.
3719         (dump_printf): Update for change to dump_printf_va.
3720         (dump_printf_loc): Likewise.
3721         (selftest::test_capture_of_dump_calls): Convert "stmt" from
3722         greturn * to gimple *.  Add a test_decl.  Add tests of dump_printf
3723         with %T, %E, and %G.
3724         * dumpfile.h (ATTRIBUTE_GCC_DUMP_PRINTF): New macro.
3725         (dump_printf): Replace ATTRIBUTE_PRINTF_2 with
3726         ATTRIBUTE_GCC_DUMP_PRINTF (2, 3).
3727         (dump_printf_loc): Replace ATTRIBUTE_PRINTF_3 with
3728         ATTRIBUTE_GCC_DUMP_PRINTF (3, 0).
3729         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Convert
3730         use of HOST_WIDE_INT_PRINT_DEC on unsigned HOST_WIDE_INT "count"
3731         within a dump_printf_loc call to "%wu".
3732         (vector_alignment_reachable_p): Merge two dump_printf[_loc] calls,
3733         converting a use of HOST_WIDE_INT_PRINT_DEC to "%wd".  Add a
3734         missing space after "=".
3735         * tree-vect-loop.c (vect_analyze_loop_2) Within a dump_printf
3736         call, convert use of HOST_WIDE_INT_PRINT_DEC to "%wd".
3737         * tree-vect-slp.c (vect_slp_bb): Within a dump_printf_loc call,
3738         convert use of HOST_WIDE_INT_PRINT_UNSIGNED to "%wu".
3739         * tree-vectorizer.c (try_vectorize_loop_1): Likewise.  Remove
3740         duplicate "vectorized" from message.
3742 2018-08-17  Szabolcs Nagy  <szabolcs.nagy@arm.com>
3744         * config/arm/arm-builtins.c (arm_init_simd_builtin_types): Clear
3745         polyNxK_t element's TYPE_STRING_FLAG.
3747 2018-08-17  Segher Boessenkool  <segher@kernel.crashing.org>
3749         * config/rs6000/rs6000.md (*cbranch, *creturn): Name these patterns
3750         (they were unnamed before).  Fix comments.
3752 2018-08-17  Nathan Sidwell  <nathan@acm.org>
3754         * cppbuiltin.c: Include "cpplib.h", not "cpp-id-data.h".
3756 2018-08-17  Richard Biener  <rguenther@suse.de>
3758         PR tree-optimization/86841
3759         * wide-int-range.cc (wide_int_range_lshift): Use to_uhwi.
3761 2018-08-17  Martin Liska  <mliska@suse.cz>
3763         * common.opt: Remove Warn, Init and Report for options with
3764         Ignore/Deprecated flag. Warning is done automatically for
3765         Deprecated flags.
3766         * config/i386/i386.opt: Likewise.
3767         * config/ia64/ia64.opt: Likewise.
3768         * config/rs6000/rs6000.opt: Likewise.
3769         * cppbuiltin.c (define_builtin_macros_for_compilation_flags):
3770         Remove usage of flag_check_pointer_bounds.
3771         * lto-wrapper.c (merge_and_complain): Do not handle
3772         OPT_fcheck_pointer_bounds.
3773         (append_compiler_options): Likewise.
3774         * opt-functions.awk: Do not handle Deprecated.
3775         * optc-gen.awk: Check that Var, Report and Init are not
3776         used for an option with Ignore/Deprecated flag.
3777         * opts-common.c (decode_cmdline_option): Do not report
3778         CL_ERR_DEPRECATED.
3779         (read_cmdline_option): Report warning for OPT_SPECIAL_deprecated
3780         options.
3781         * opts.h (struct cl_option): Remove cl_deprecated flag.
3782         (CL_ERR_DEPRECATED): Remove error enum value.
3784 2018-08-17  Richard Biener  <rguenther@suse.de>
3786         PR middle-end/86505
3787         * tree-inline.c (copy_bb): When inlining __builtin_va_arg_pack_len ()
3788         across a va-arg-pack using call adjust its return value accordingly.
3790 2018-08-16  Martin Sebor  <msebor@redhat.com>
3792         PR tree-optimization/86853
3793         * gimple-ssa-sprintf.c (struct format_result): Rename member.
3794         (struct fmtresult): Add member and initialize it in ctors.
3795         (format_character): Handle %C.  Extend range to NUL.  Set MAYFAIL.
3796         (format_string): Handle %S the same as %ls.  Set MAYFAIL.
3797         (format_directive): Set POSUNDER4K when MAYFAIL is set.
3798         (parse_directive): Handle %C same as %c.
3799         (sprintf_dom_walker::compute_format_length): Adjust.
3800         (is_call_safe): Adjust.
3802 2018-08-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3804         * builtins.c (c_strlen): Add new parameter eltsize.  Use it
3805         for determining how to count the elements.
3806         * builtins.h (c_strlen): Adjust prototype.
3807         * expr.c (string_constant): Add new parameter mem_size.
3808         Set *mem_size appropriately.
3809         * expr.h (string_constant): Adjust protoype.
3810         * gimple-fold.c (get_range_strlen): Add new parameter eltsize.
3811         * gimple-fold.h (get_range_strlen): Adjust prototype.
3812         * gimple-ssa-sprintf.c (get_string_length): Add new parameter eltsize.
3813         (format_string): Call get_string_length with eltsize.
3815 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
3817         * diagnostic.c (default_diagnostic_start_span_fn): Call pp_string
3818         to emit the span, rather than setting it as the prefix.
3820 2018-08-16  David Malcolm  <dmalcolm@redhat.com>
3822         * diagnostic-show-locus.c (layout::start_annotation_line): Add
3823         "margin_char" parameter, defaulting to space.  Use it in place
3824         of pp_space for the initial part of the margin.
3825         (layout::print_leading_fixits): Use '+' when filling the margin
3826         of line-insertion fix-it hints.
3828 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
3830         * config/rs6000/rs6000.md (two unnamed define_insn and define_split):
3831         Delete.
3833 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
3835         * config/rs6000/altivec.md: Don't set length attribute to the default
3836         value.
3837         * config/rs6000/darwin.md: Ditto.
3838         * config/rs6000/dfp.md: Ditto.
3839         * config/rs6000/htm.md: Ditto.
3840         * config/rs6000/rs6000.md: Ditto.
3841         * config/rs6000/sync.md: Ditto.
3842         * config/rs6000/vsx.md: Ditto.
3844 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
3846         * config/rs6000/altivec.md: Don't set length attribute to the default
3847         value, for branch instructions.
3848         * config/rs6000/darwin.md: Ditto.
3849         * config/rs6000/rs6000.md: Ditto.
3851 2018-08-16  Segher Boessenkool  <segher@kernel.crashing.org>
3853         * config/rs6000/rs6000.md (length): Always define as const_int 4.
3854         (unnamed conditional branch define_insn): Set length to 4 or 8
3855         depending on offset.
3856         (<bd>_<mode>): Similar, for alternative 0.
3857         (<bd>tf_<mode>): Ditto.
3859 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
3861         * expr.c (copy_blkmode_to_reg): Perform larger copies when safe.
3863 2018-08-16  Matthew Malcomson  <matthew.malcomson@arm.com>
3865         * doc/rtl.texi: Replace old RTX class names with new names.
3868 2018-08-16  Vlad Lazar  <vlad.lazar@arm.com>
3870         * expmed.h (canonicalize_comparison): New declaration.
3871         * expmed.c (canonicalize_comparison, equivalent_cmp_code): New function.
3872         * expmed.c (emit_store_flag_1): Add call to canonicalize_comparison.
3873         * optabs.c (prepare_cmp_insn): Likewise.
3874         * rtl.h (unsigned_condition_p): New function which checks if a
3875         comparison operator is unsigned.
3877 2018-08-16  Nathan Sidwell  <nathan@acm.org>
3879         * config/rs6000/rs6000-c.c (rs6000_macro_to_expend): Use cpp_macro_p.
3880         * config/powerpcspc/powerpcspe-c.c (rs6000_macro_to_expend): Likewise.
3882 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
3884         PR target/84711
3885         * config/arm/arm.c (arm_can_change_mode_class): Disallow subreg.
3886         * config/arm/neon.md (movv4hf, movv8hf): Refactored to..
3887         (mov<mov>): ..this and enable unconditionally.
3889 2018-08-16  Tamar Christina  <tamar.christina@arm.com>
3891         * config/arm/neon.md (*neon_mov<mode>): Remove reg-to-reg alternative.
3893 2018-08-16  Sam Tebbs  <sam.tebbs@arm.com>
3895         * config/aarch64/aarch64.opt (mlow-precision-recip-sqrt)
3896         (mlow-precision-sqrt, mlow-precision-div, mverbose-cost-dump): Replace
3897         "Common" with "Target".
3899 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
3901         * config/i386/i386.opt (mmitigate-rop): Mark as deprecated.
3902         * doc/invoke.texi (mmitigate-rop): Remove.
3903         * config/i386/i386.c: Do not include "regrename.h".
3904         (ix86_rop_should_change_byte_p, reg_encoded_number)
3905         (ix86_get_modrm_for_rop, set_rop_modrm_reg_bits, ix86_mitigate_rop):
3906         Remove.
3907         (ix86_reorg): Remove call to ix86_mitigate_rop.
3908         * config/i386/i386.md (attr "modrm_class"): Remove.
3909         (cmp<mode>_ccno_1, mov<mode>_xor, movstrict<mode>_xor)
3910         (x86_mov<mode>cc_0_m1. x86_mov<mode>cc_0_m1_se)
3911         (x86_mov<mode>cc_0_m1_neg): Remove modrm_class attribute override.
3913 2018-08-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
3915         * config/rs6000/rs600.c (rs6000_gimple_fold_builtin): Add entries to
3916         allow folding of mergeh() and mergel() for the float and double types.
3917         (fold_mergehl_helper): Rework to handle building a permute tree
3918         for float vectors.
3920 2018-08-15  Uros Bizjak  <ubizjak@gmail.com>
3922         * config/i386/i386.c (expand_vec_perm_movs): Enable V4SFmode
3923         for TARGET_SSE.
3925 2018-08-15  David Malcolm  <dmalcolm@redhat.com>
3927         * common.opt (fdiagnostics-show-labels): New option.
3928         * diagnostic-show-locus.c (class layout_range): Add field
3929         "m_label".
3930         (class layout): Add field "m_show_labels_p".
3931         (layout_range::layout_range): Add param "label" and use it to
3932         initialize m_label.
3933         (make_range): Pass in NULL for new "label" param of layout_range's
3934         ctor.
3935         (layout::layout): Initialize m_show_labels_p.
3936         (layout::maybe_add_location_range): Pass in loc_range->m_label
3937         when constructing layout_range instances.
3938         (struct line_label): New struct.
3939         (layout::print_any_labels): New member function.
3940         (layout::print_line): Call it if label-printing is enabled.
3941         (selftest::test_one_liner_labels): New test.
3942         (selftest::test_diagnostic_show_locus_one_liner): Call it.
3943         * diagnostic.c (diagnostic_initialize): Initialize
3944         context->show_labels_p.
3945         * diagnostic.h (struct diagnostic_context): Add field
3946         "show_labels_p".
3947         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
3948         -fno-diagnostics-show-labels.
3949         * dwarf2out.c (gen_producer_string): Add
3950         OPT_fdiagnostics_show_labels to the ignored options.
3951         * gcc-rich-location.c (gcc_rich_location::add_expr): Add "label"
3952         param.
3953         (gcc_rich_location::maybe_add_expr): Likewise.
3954         * gcc-rich-location.h (gcc_rich_location::gcc_rich_location): Add
3955         label" param, defaulting to NULL.
3956         (gcc_rich_location::add_expr): Add "label" param.
3957         (gcc_rich_location::maybe_add_expr): Likewise.
3958         (class text_range_label): New class.
3959         (class range_label_for_type_mismatch): New class.
3960         * gimple-ssa-sprintf.c (fmtwarn): Pass NULL for new label params
3961         of format_warning_va.
3962         (fmtwarn_n): Likewise for new params of format_warning_n_va.
3963         * lto-wrapper.c (merge_and_complain): Add
3964         OPT_fdiagnostics_show_labels to the "pick one setting" options.
3965         (append_compiler_options): Likewise to the dropped options.
3966         (append_diag_options): Likewise to the passed-on options.
3967         * opts.c (common_handle_option): Handle the new option.
3968         * selftest-diagnostic.c
3969         (test_diagnostic_context::test_diagnostic_context): Enable
3970         show_labels_p.
3971         * substring-locations.c: Include "gcc-rich-location.h".
3972         (format_warning_n_va): Add "fmt_label" and "param_label" params
3973         and use them as appropriate.
3974         (format_warning_va): Add "fmt_label" and "param_label" params,
3975         passing them on to format_warning_n_va.
3976         (format_warning_at_substring): Likewise.
3977         (format_warning_at_substring_n): Likewise.
3978         * substring-locations.h (format_warning_va): Add "fmt_label" and
3979         "param_label" params.
3980         (format_warning_n_va): Likewise.
3981         (format_warning_at_substring): Likewise.
3982         (format_warning_at_substring_n): Likewise.
3983         * toplev.c (general_init): Initialize global_dc->show_labels_p.
3985 2018-08-15  Qing Zhao  <qing.zhao@oracle.com>
3987         PR testsuite/86519
3988         * builtins.c (expand_builtin_memcmp): Do not expand the call
3989         when overflow is detected.
3991 2018-08-15  Martin Sebor  <msebor@redhat.com>
3993         PR tree-optimization/71625
3994         * config/aarch64/aarch64-builtins.c
3995         (aarch64_init_simd_builtin_types): Clear Poly8_t's TYPE_STRING_FLAG.
3997 2018-08-15  Ilya Leoshkevich  <iii@linux.ibm.com>
3999         * config/s390/s390.c (s390_reorg): Remove loop.
4001 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
4003         * config/darwin.c
4004          (darwin_function_switched_text_sections): Delete.
4005         * gcc/config/darwin.h
4006          (TARGET_ASM_FUNCTION_SWITCHED_TEXT_SECTIONS): Likewise.
4008 2018-08-15  Iain Sandoe  <iain@sandoe.co.uk>
4010         PR target/81685
4011         * config/darwin.h: (DEBUG_STR_OFFSETS_SECTION, DEBUG_LOCLISTS_SECTION,
4012         DEBUG_RNGLISTS_SECTION) new macros.  (DEBUG_PUBNAMES_SECTION,
4013         DEBUG_PUBTYPES_SECTION) update to include GNU variant.
4015 2018-08-15  Martin Liska  <mliska@suse.cz>
4017         PR tree-optimization/86925
4018         * predict.c (expr_expected_value_1): When taking
4019         later predictor, assign also probability.
4020         Use fold_build2_initializer_loc in order to fold
4021         the expression in -frounding-math.
4023 2018-08-14  Allan Sandfeld Jensen <allan.jensen@qt.io>
4025         * config/i386/i386.c (expand_vec_perm_movs): New method matching movs
4026         patterns.
4027         (expand_vec_perm_1): Try the new method.
4029 2018-08-14  Ilya Leoshkevich  <iii@linux.ibm.com>
4031         PR target/86547
4032         * lra-lives.c (remove_some_program_points_and_update_live_ranges):
4033         Check whether lra_live_max_point is 0 before dividing.
4035 2018-08-14  Martin Sebor  <msebor@redhat.com>
4037         PR tree-optimization/86650
4038         * tree-vrp.c (vrp_prop::check_array_ref): Print an inform message.
4039         (vrp_prop::check_mem_ref): Same.
4041 2018-08-13  Liu Hao <lh_mouse@126.com>
4043         * pretty-print.c (eat_esc_sequence): Swap the foreground and
4044         background colors if the COMMON_LVB_REVERSE_VIDEO flag is set,
4045         and clear it thereafter, as it only works for DBCS.
4047 2018-08-13  Liu Hao <lh_mouse@126.com>
4049         * pretty-print.c (mingw_ansi_fputs): Do not call _close() on the
4050         handle returned by _get_osf_handle().
4052 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
4054         * gcc/config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support
4055         for folding vec_perm.
4057 2018-08-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
4059         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin):
4060         Add support for gimple-folding of vec_pack() and vec_unpack()
4061         intrinsics.
4063 2018-08-13  Will Schmidt <will_schmidt@vnet.ibm.com>
4065         * config/rs6000/rs6000.c (rs6000_builtin_valid_without_lhs): Add
4066         vec_xst variants to the list.
4067         (rs6000_gimple_fold_builtin): Add support for folding unaligned
4068         vector loads and stores.
4070 2018-08-13  David Edelsohn  <dje.gcc@gmail.com>
4072         * config.gcc (rs6000-ibm-aix4.x): Delete.
4073         (rs6000-ibm-aix5.1): Delete.
4074         (rs6000-ibm-aix5.2): Delete.
4075         (rs6000-ibm-aix5.3): Delete.
4076         * config/rs6000/aix43.h: Delete.
4077         * config/rs6000/aix51.h: Delete.
4078         * config/rs6000/aix52.h: Delete.
4079         * config/rs6000/t-aix43: Delete.
4081 2018-08-13  Ilya Leoshkevich  <iii@linux.ibm.com>
4083         * config/s390/s390.c (s390_decompose_constant_pool_ref):
4084         New function.
4085         (s390_decompose_address): Factor out constant pool ref
4086         decomposition.
4088 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
4090         * config/nds32/nds32-predicates.c
4091         (nds32_can_use_bclr_p): Change return type as bool.
4092         (nds32_can_use_bset_p): Ditto.
4093         (nds32_can_use_btgl_p): Ditto.
4094         (nds32_can_use_bitci_p): Ditto.
4095         * config/nds32/nds32-protos.h
4096         (nds32_can_use_bclr_p): Change declaration.
4097         (nds32_can_use_bset_p): Ditto.
4098         (nds32_can_use_btgl_p): Ditto.
4099         (nds32_can_use_bitci_p): Ditto.
4101 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
4103         * config/nds32/nds32.c (nds32_expand_prologue, nds32_expand_epilogue):
4104         Support -msched-prolog-epilog option.
4105         * config/nds32/nds32.opt (msched-prolog-epilog): New option.
4107 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
4109         * common/config/nds32/nds32-common.c
4110         (nds32_option_optimization_table): Enalbe -malways-align.
4112 2018-08-12  Chung-Ju Wu  <jasonwucj@gmail.com>
4114         * config.gcc (nds32*): Add nds32_isr.h and nds32_init.inc in
4115         extra_headers.
4116         * common/config/nds32/nds32-common.c (nds32_handle_option): Handle
4117         OPT_misr_secure_ case.
4118         * config/nds32/nds32-isr.c: Implementation of backward compatibility.
4119         * config/nds32/nds32-protos.h (nds32_isr_function_critical_p): New.
4120         * config/nds32/nds32.c (nds32_attribute_table): Add critical and
4121         secure attribute.
4122         * config/nds32/nds32.h (nds32_isr_nested_type): Add NDS32_CRITICAL.
4123         (nds32_isr_info): New field security_level.
4124         (TARGET_ISR_VECTOR_SIZE_4_BYTE): New macro.
4125         * config/nds32/nds32.md (return_internal): Consider critical attribute.
4126         * config/nds32/nds32.opt (misr-secure): New option.
4127         * config/nds32/nds32_init.inc: New file.
4128         * config/nds32/nds32_isr.h: New file.
4130 2018-08-11  John David Anglin  <danglin@gcc.gnu.org>
4132         * config/pa/pa.md (UNSPEC_MEMORY_BARRIER): New unspec enum.
4133         Update comment for atomic instructions.
4134         (atomic_storeqi, atomic_storehi, atomic_storesi, atomic_storesf,
4135         atomic_loaddf, atomic_loaddf_1, atomic_storedf, atomic_storedf_1):
4136         Remove.
4137         (atomic_loaddi): Revise fence expansion to only emit fence prior to
4138         load for __ATOMIC_SEQ_CST model.
4139         (atomic_loaddi_1): Remove float register target.
4140         (atomic_storedi): Handle CONST_INT values.
4141         (atomic_storedi_1): Remove float register source.  Add special case
4142         for zero value.
4143         (memory_barrier): New expander and insn.
4145 2018-08-11  Jakub Jelinek  <jakub@redhat.com>
4147         PR tree-optimization/86835
4148         * tree-ssa-math-opts.c (insert_reciprocals): Even when inserting
4149         new_stmt after def_gsi, make sure to insert new_square_stmt after
4150         that stmt, not 2 stmts before it.
4152 2018-08-10  Alexander Monakov  <amonakov@ispras.ru>
4154         PR target/82418
4155         * config/i386/i386.md (<s>mul<mode>3_highpart): Use DWIH mode iterator
4156         instead of SWI48.
4158 2018-08-10  Martin Liska  <mliska@suse.cz>
4160         PR target/83610
4161         * builtin-types.def (BT_FN_LONG_LONG_LONG_DOUBLE): Add new
4162         function type.
4163         * builtins.c (expand_builtin_expect_with_probability):
4164         New function.
4165         (expand_builtin_expect_with_probability): New function.
4166         (build_builtin_expect_predicate): Add new argumnet probability
4167         for BUILT_IN_EXPECT_WITH_PROBABILITY.
4168         (fold_builtin_expect):
4169         (fold_builtin_2):
4170         (fold_builtin_3):
4171         * builtins.def (BUILT_IN_EXPECT_WITH_PROBABILITY):
4172         * builtins.h (fold_builtin_expect): Set new argument.
4173         * doc/extend.texi: Document __builtin_expect_with_probability.
4174         * doc/invoke.texi: Likewise.
4175         * gimple-fold.c (gimple_fold_call): Pass new argument.
4176         * ipa-fnsummary.c (find_foldable_builtin_expect): Handle
4177         also BUILT_IN_EXPECT_WITH_PROBABILITY.
4178         * predict.c (get_predictor_value): New function.
4179         (expr_expected_value): Add new argument probability. Assume
4180         that predictor and probability are always non-null.
4181         (expr_expected_value_1): Likewise.  For __builtin_expect and
4182         __builtin_expect_with_probability set probability.  Handle
4183         combination in binary expressions.
4184         (tree_predict_by_opcode): Simplify code by simply calling
4185         get_predictor_value.
4186         (pass_strip_predict_hints::execute): Add handling of
4187         BUILT_IN_EXPECT_WITH_PROBABILITY.
4188         * predict.def (PRED_BUILTIN_EXPECT_WITH_PROBABILITY): Add
4189         new predictor.
4190         * tree.h (DECL_BUILT_IN_P): New function.
4192 2018-08-10  Martin Liska  <mliska@suse.cz>
4194         PR tree-optimization/85799
4195         * passes.def: Add argument for pass_strip_predict_hints.
4196         * predict.c (class pass_strip_predict_hints): Add new argument
4197         early_p.
4198         (strip_predictor_early): New function.
4199         (pass_strip_predict_hints::execute): Call the function to
4200         strip predictors.
4201         (strip_predict_hints): New function.
4202         * predict.def: Fix comment.
4204 2018-08-10  Thomas Preud'homme  <thomas.preudhomme@linaro.org>
4206         * Makefile.in: Clarify which tm.texi to copy over to assert the
4207         right to grant a GFDL license for all.
4209 2018-08-09  Jeff Law  <law@redhat.com>
4211         * config/m68k/m68k.c (m68k_adjust_decorated_operand): Remove
4212         unused variable.
4214 2018-08-09  Andreas Schwab  <schwab@linux-m68k.org>
4216         * config/m68k/m68k-protos.h (m68k_final_prescan_insn): Remove
4217         prototype.
4219 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
4221         * tree-vect-loop.c (vectorizable_reduction): Allow inner-loop
4222         reductions for variable-length vectors.
4224 2018-08-09  David Malcolm  <dmalcolm@redhat.com>
4226         PR other/84889
4227         * common.opt (fdiagnostics-show-line-numbers): New option.
4228         * diagnostic-show-locus.c (class layout): Add fields
4229         "m_show_line_numbers_p" and "m_linenum_width";
4230         (num_digits): New function.
4231         (test_num_digits): New function.
4232         (layout::layout): Initialize new fields.  Update m_x_offset
4233         logic to handle any left margin.
4234         (layout::print_source_line): Print line number when requested.
4235         (layout::start_annotation_line): New member function.
4236         (layout::print_annotation_line): Call it.
4237         (layout::print_leading_fixits): Likewise.
4238         (layout::print_trailing_fixits): Likewise.  Update calls to
4239         move_to_column for new parameter.
4240         (layout::get_x_bound_for_row): Add "add_left_margin" param and use
4241         it to potentially call start_annotation_line.
4242         (layout::show_ruler): Call start_annotation_line.
4243         (selftest::test_line_numbers_multiline_range): New selftest.
4244         (selftest::diagnostic_show_locus_c_tests): Call test_num_digits
4245         and selftest::test_line_numbers_multiline_range.
4246         * diagnostic.c (diagnostic_initialize): Initialize
4247         show_line_numbers_p.
4248         * diagnostic.h (struct diagnostic_context): Add field
4249         "show_line_numbers_p".
4250         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
4251         -fno-diagnostics-show-line-numbers.
4252         * dwarf2out.c (gen_producer_string): Add
4253         OPT_fdiagnostics_show_line_numbers to the ignored options.
4254         * lto-wrapper.c (merge_and_complain): Likewise to the "pick
4255         one setting" options.
4256         (append_compiler_options): Likewise to the dropped options.
4257         (append_diag_options): Likewise to the passed-on options.
4258         * opts.c (common_handle_option): Handle the new option.
4259         * toplev.c (general_init): Set up global_dc->show_line_numbers_p.
4261 2018-08-09  Kelvin Nilsen  <kelvin@gcc.gnu.org>
4263         * doc/extend.texi (PowerPC AltiVec Built-in Functions Available on
4264         ISA 2.07): Correct spelling of bcdsub to be __builtin_bcdsub.  Add
4265         third argument of type "const signed char" to descriptions of
4266         __builtin_bcdadd, __builtin_bcdadd_lt, __builtin_bcdadd_eq,
4267         __builtin_bcdadd_gt, __builtin_bcdadd_ov, __builtin_bcdsub,
4268         __builtin_bcdsub_lt, __builtin_bcdsub_eq, __builtin_bcdsub_gt,
4269         __builtin_bcdsub_ov functions.
4271 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
4273         PR tree-optimization/86858
4274         * tree-vect-loop.c (vect_is_simple_reduction): Restore
4275         flow_bb_inside_loop_p calls.
4277 2018-08-09  Richard Sandiford  <richard.sandiford@arm.com>
4279         PR tree-optimization/86871
4280         * tree-vect-stmts.c (vect_transform_stmt): Use gimple_get_lhs
4281         instead of gimple_assign_lhs.
4283 2018-08-09  Richard Earnshaw  <rearnsha@arm.com>
4285         PR target/86887
4286         * config/aarch64/aarch64.md (add<mode>3_carryinC_zero): Add missing
4287         register constraint to operand 0.
4288         (add<mode>3_carryinC): Likewise.
4289         (add<mode>3_carryinV_zero, add<mode>3_carryinV): Likewise.
4291 2018-08-09  Martin Liska  <mliska@suse.cz>
4293         PR c/86895
4294         * common.opt: Remove extra line.
4296 2018-08-09  Martin Liska  <mliska@suse.cz>
4298         * params.def (PARAM_ALIGN_LOOP_ITERATIONS): Remove double dots
4299         at the end of a line, make first letter capital and end up
4300         a sentence with a dot.
4301         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): Likewise.
4302         (PARAM_LOOP_BLOCK_TILE_SIZE): Likewise.
4303         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Likewise.
4304         (PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP): Likewise.
4305         (PARAM_MAX_ISL_OPERATIONS): Likewise.
4306         (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): Likewise.
4307         (PARAM_PROFILE_FUNC_INTERNAL_ID): Likewise.
4308         (PARAM_INDIR_CALL_TOPN_PROFILE): Likewise.
4309         (PARAM_SLP_MAX_INSNS_IN_BB): Likewise.
4310         (PARAM_IPA_CP_EVAL_THRESHOLD): Likewise.
4311         (PARAM_IPA_CP_RECURSION_PENALTY): Likewise.
4312         (PARAM_IPA_CP_SINGLE_CALL_PENALTY): Likewise.
4313         (PARAM_IPA_CP_LOOP_HINT_BONUS): Likewise.
4314         (PARAM_IPA_CP_ARRAY_INDEX_HINT_BONUS): Likewise.
4315         (PARAM_TREE_REASSOC_WIDTH): Likewise.
4316         (PARAM_HSA_GEN_DEBUG_STORES): Likewise.
4317         (PARAM_MAX_SPECULATIVE_DEVIRT_MAYDEFS): Likewise.
4318         (PARAM_MAX_VRP_SWITCH_ASSERTIONS): Likewise.
4320 2018-08-09  Andreas Krebbel  <krebbel@linux.ibm.com>
4322         PR target/84332
4323         * config/s390/s390.c (s390_option_override_internal): Reduce the
4324         stack-clash-protection-probe-interval param if it would be too big
4325         for z900.
4327 2018-08-08  Andreas Schwab  <schwab@linux-m68k.org>
4329         PR target/46179
4330         * config/m68k/m68k.h (FINAL_PRESCAN_INSN): Don't define.
4331         * config/m68k/m68k.c (handle_move_double): Don't call
4332         m68k_final_prescan_insn.
4333         (m68k_adjust_decorated_operand): Renamed from
4334         m68k_final_prescan_insn, remove first and third operand and
4335         simplify.
4336         (print_operand): Call it.
4337         (print_operand_address): Call it.
4339 2018-08-08  Nathan Sidwell  <nathan@acm.org>
4341         * diagnostic.c (diagnostic_report_current_module): Use
4342         linemap_included_from & linemap_included_from_linemap.
4344 2018-08-08  Hongbo Zhang  <hongbo.zhang@linaro.org>
4346         * config/aarch64/aarch64-cores.def: Add phecda core.
4347         * config/aarch64/aarch64-tune.md: Regenerate.
4348         * doc/invoke.texi: Add phecda core.
4350 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
4352         PR target/85295
4353         * config/s390/constraints.md ("NxHD0", "NxSD0"): New constraint
4354         definitions.
4355         * config/s390/s390.md ("movti"): Add more alternatives for
4356         constant to GPR copies.
4358 2018-08-08  Andreas Krebbel  <krebbel@linux.ibm.com>
4360         * config/s390/s390.c: Fix whitespace damage throughout the file.
4361         * config/s390/s390.h: Likewise.
4362         * config/s390/tpf.h: Likewise.
4364 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
4366         * config/s390/s390.c (s390_loadrelative_operand_p):
4367         Remove TARGET_CPU_ZARCH usages.
4368         (s390_rtx_costs): Likewise.
4369         (s390_legitimate_constant_p): Likewise.
4370         (s390_cannot_force_const_mem): Likewise.
4371         (legitimate_reload_constant_p): Likewise.
4372         (s390_preferred_reload_class): Likewise.
4373         (legitimize_pic_address): Likewise.
4374         (legitimize_tls_address): Likewise.
4375         (s390_split_branches): Removed.
4376         (s390_add_execute): Removed.
4377         (s390_dump_pool): Remove TARGET_CPU_ZARCH usages.
4378         (s390_mainpool_start): Likewise.
4379         (s390_mainpool_finish): Likewise.
4380         (s390_mainpool_cancel): Removed.
4381         (s390_chunkify_start): Remove TARGET_CPU_ZARCH usages.
4382         (s390_chunkify_cancel): Likewise.
4383         (s390_return_addr_rtx): Likewise.
4384         (s390_register_info): Remove split_branches_pending_p uages.
4385         (s390_optimize_register_info): Likewise.
4386         (s390_init_frame_layout): Remove TARGET_CPU_ZARCH and
4387         split_branches_pending_p usages.
4388         (s390_can_eliminate): Remove TARGET_CPU_ZARCH usages.
4389         (s390_load_got): Likewise.
4390         (s390_expand_split_stack_prologue): Likewise.
4391         (output_asm_nops): Likewise.
4392         (s390_function_profiler): Likewise.
4393         (s390_emit_call): Likewise.
4394         (s390_conditional_register_usage): Likewise.
4395         (s390_optimize_prologue): Likewise.
4396         (s390_reorg): Remove TARGET_CPU_ZARCH and
4397         split_branches_pending_p usages.
4398         (s390_option_override_internal): Remove TARGET_CPU_ZARCH
4399         usages.
4400         (s390_output_indirect_thunk_function): Likewise.
4401         * config/s390/s390.h (TARGET_CPU_ZARCH): Removed.
4402         (TARGET_CPU_ZARCH_P): Removed.
4403         (struct machine_function): Remove split_branches_pending_p.
4404         * config/s390/s390.md: Remove TARGET_CPU_ZARCH usages.
4406 2018-08-08  Ilya Leoshkevich  <iii@linux.ibm.com>
4408         * common/config/s390/s390-common.c (processor_flags_table):
4409         Remove flags.
4410         * config.gcc: Remove with_arch/with_tune support.
4411         * config/s390/2064.md: Remove cpu attribute comparisons.
4412         * config/s390/driver-native.c (s390_host_detect_local_cpu):
4413         Remove MTN.
4414         * config/s390/linux.h (ASM_SPEC):
4415         Remove -march support.
4416         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal):
4417         Use a table to get an arch level.
4418         * config/s390/s390-opts.h (enum processor_type):
4419         Remove enum values.
4420         * config/s390/s390.c
4421         (processor_table): Remove entries, add arch_level values.
4422         (s390_issue_rate): Remove cases.
4423         (s390_option_override): Adjust
4424         s390_option_override_internal() call.
4425         (s390_option_override_internal): Remove deprecation warning.
4426         (s390_valid_target_attribute_tree): Adjust
4427         s390_option_override_internal() call.
4428         * config/s390/s390.h (struct s390_processor):
4429         Share with s390-c.c, add arch_level field.
4430         * config/s390/s390.md:
4431         Remove occurrences in cpu attribute.
4432         * config/s390/s390.opt: Remove -march/-mtune support.
4433         * config/s390/tpf.h (ASM_SPEC): Remove -march support.
4434         * doc/invoke.texi: Remove deprecation warning.
4436 2018-08-08  Luis Machado  <luis.machado@linaro.org>
4438         * config/aarch64/aarch64.c (qdf24xx_vector_cost): New static
4439         global.
4440         (qdf24xx_tunings): Set vector cost structure to
4441         qdf24xx_vector_cost.
4443         * config/aarch64/aarch64.c (qdf24xx_addrcost_table)
4444         <register_sextend>: Set to 3.
4446 2018-08-07  Richard Sandiford  <richard.sandiford@arm.com>
4448         PR target/86838
4449         * config/aarch64/iterators.md (FRECP, frecp_suffix): Delete.
4450         * config/aarch64/aarch64-simd.md
4451         (aarch64_frecp<FRECP:frecp_suffix><mode>): Fold FRECPE into...
4452         (@aarch64_frecpe<mode>): ...here and the move FRECPX to...
4453         (aarch64_frecpx<mode>): ...this new pattern.
4454         * config/aarch64/aarch64-simd-builtins.def: Remove comment
4455         about aarch64_frecp<FRECP:frecp_suffix><mode>.
4457 2018-08-07  Martin Liska  <mliska@suse.cz>
4459         PR middle-end/83023
4460         * predict.c (expr_expected_value_1): Handle DECL_IS_MALLOC,
4461         BUILT_IN_REALLOC and DECL_IS_OPERATOR_NEW.
4462         * predict.def (PRED_MALLOC_NONNULL): New predictor.
4463         * doc/extend.texi: Document that malloc attribute adds
4464         hit to compiler.
4466 2018-08-06  John David Anglin  <danglin@gcc.gnu.org>
4468         PR target/86785
4469         * config/pa/pa.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4470         Define to speculation_safe_value_not_needed.
4472 2018-08-06  Jeff Law  <law@redhat.com>
4474         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Pass down
4475         the vr_values instance to cprop_into_stmt.
4476         (cprop_into_stmt): Pass vr_values instance down to cprop_operand.
4477         (cprop_operand): Also query EVRP to determine if OP is a constant.
4479 2018-08-06  Nathan Sidwell  <nathan@acm.org>
4481         * diagnostic.c (diagnostic_report_current_module): Reroll
4482         included-at loop.  Translate text.
4484 2018-08-06  David Malcolm  <dmalcolm@redhat.com>
4486         * function-tests.c (selftest::test_expansion_to_rtl): Call
4487         free_after_compilation.
4489 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
4491         * config/aarch64/aarch64.md: Add clobber highs to tls_desc.
4493 2018-08-06  Andreas Krebbel  <krebbel@linux.ibm.com>
4495         * config/s390/s390.c (s390_loop_unroll_adjust): Prevent small
4496         loops with memory block operations from getting unrolled.
4498 2018-08-06  Ulrich Weigand  <uweigand@de.ibm.com>
4500         PR target/86807
4501         * config/spu/spu.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4502         Define to speculation_safe_value_not_needed.
4504 2018-08-06  Jeff Law  <law@redhat.com>
4506         * reload1.c (forget_old_reloads_1): Adjust CLOBBER_HIGH
4507         assert.
4509 2018-08-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
4511         PR target/86662
4512         * gcc/tree.c (build_common_tree_nodes): Initialize integer_types array
4513         with all enabled __intN types.
4515         * gcc/testsuite/gcc.target/msp430/pr86662.c: New test.
4517 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
4519         * alias.c (record_set): Check for clobber high.
4520         * cfgexpand.c (expand_gimple_stmt): Likewise.
4521         * combine-stack-adj.c (single_set_for_csa): Likewise.
4522         * combine.c (find_single_use_1): Likewise.
4523         (set_nonzero_bits_and_sign_copies): Likewise.
4524         (get_combine_src_dest): Likewise.
4525         (is_parallel_of_n_reg_sets): Likewise.
4526         (try_combine): Likewise.
4527         (record_dead_and_set_regs_1): Likewise.
4528         (reg_dead_at_p_1): Likewise.
4529         (reg_dead_at_p): Likewise.
4530         * dce.c (deletable_insn_p): Likewise.
4531         (mark_nonreg_stores_1): Likewise.
4532         (mark_nonreg_stores_2): Likewise.
4533         * df-scan.c (df_find_hard_reg_defs): Likewise.
4534         (df_uses_record): Likewise.
4535         (df_get_call_refs): Likewise.
4536         * dwarf2out.c (mem_loc_descriptor): Likewise.
4537         * haifa-sched.c (haifa_classify_rtx): Likewise.
4538         * ira-build.c (create_insn_allocnos): Likewise.
4539         * ira-costs.c (scan_one_insn): Likewise.
4540         * ira.c (equiv_init_movable_p): Likewise.
4541         (rtx_moveable_p): Likewise.
4542         (interesting_dest_for_shprep): Likewise.
4543         * jump.c (mark_jump_label_1): Likewise.
4544         * postreload-gcse.c (record_opr_changes): Likewise.
4545         * postreload.c (reload_cse_simplify): Likewise.
4546         (struct reg_use): Add source expr.
4547         (reload_combine): Check for clobber high.
4548         (reload_combine_note_use): Likewise.
4549         (reload_cse_move2add): Likewise.
4550         (move2add_note_store): Likewise.
4551         * print-rtl.c (print_pattern): Likewise.
4552         * recog.c (decode_asm_operands): Likewise.
4553         (store_data_bypass_p): Likewise.
4554         (if_test_bypass_p): Likewise.
4555         * regcprop.c (kill_clobbered_value): Likewise.
4556         (kill_set_value): Likewise.
4557         * reginfo.c (reg_scan_mark_refs): Likewise.
4558         * reload1.c (maybe_fix_stack_asms): Likewise.
4559         (eliminate_regs_1): Likewise.
4560         (elimination_effects): Likewise.
4561         (mark_not_eliminable): Likewise.
4562         (scan_paradoxical_subregs): Likewise.
4563         (forget_old_reloads_1): Likewise.
4564         * reorg.c (find_end_label): Likewise.
4565         (try_merge_delay_insns): Likewise.
4566         (redundant_insn): Likewise.
4567         (own_thread_p): Likewise.
4568         (fill_simple_delay_slots): Likewise.
4569         (fill_slots_from_thread): Likewise.
4570         (dbr_schedule): Likewise.
4571         * resource.c (update_live_status): Likewise.
4572         (mark_referenced_resources): Likewise.
4573         (mark_set_resources): Likewise.
4574         * rtl.c (copy_rtx): Likewise.
4575         * rtlanal.c (reg_referenced_p): Likewise.
4576         (single_set_2): Likewise.
4577         (noop_move_p): Likewise.
4578         (note_stores): Likewise.
4579         * sched-deps.c (sched_analyze_reg): Likewise.
4580         (sched_analyze_insn): Likewise.
4582 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
4584         * cse.c (invalidate_reg): New function extracted from...
4585         (invalidate): ...here.
4586         (canonicalize_insn): Check for clobber high.
4587         (invalidate_from_clobbers): invalidate clobber highs.
4588         (invalidate_from_sets_and_clobbers): Likewise.
4589         (count_reg_usage): Check for clobber high.
4590         (insn_live_p): Likewise.
4591         * cselib.c (cselib_expand_value_rtx_1):Likewise.
4592         (cselib_invalidate_regno): Check for clobber in setter.
4593         (cselib_invalidate_rtx): Pass through setter.
4594         (cselib_invalidate_rtx_note_stores):
4595         (cselib_process_insn): Check for clobber high.
4596         * cselib.h (cselib_invalidate_rtx): Add operand.
4598 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
4600         * lra-eliminations.c (lra_eliminate_regs_1): Check for clobber high.
4601         (mark_not_eliminable): Likewise.
4602         * lra-int.h (struct lra_insn_reg): Add clobber high marker.
4603         * lra-lives.c (process_bb_lives): Check for clobber high.
4604         * lra.c (new_insn_reg): Remember clobber highs.
4605         (collect_non_operand_hard_regs): Check for clobber high.
4606         (lra_set_insn_recog_data): Likewise.
4607         (add_regs_to_insn_regno_info): Likewise.
4608         (lra_update_insn_regno_info): Likewise.
4610 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
4612         * rtl.h (reg_is_clobbered_by_clobber_high): Add declarations.
4613         * rtlanal.c (reg_is_clobbered_by_clobber_high): Add function.
4615 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
4617         * emit-rtl.c (verify_rtx_sharing): Check for CLOBBER_HIGH.
4618         (copy_insn_1): Likewise.
4619         (gen_hard_reg_clobber_high): New gen function.
4620         * genconfig.c (walk_insn_part): Check for CLOBBER_HIGH.
4621         * genemit.c (gen_exp): Likewise.
4622         (gen_emit_seq): Pass through info.
4623         (gen_insn): Check for CLOBBER_HIGH.
4624         (gen_expand): Pass through info.
4625         (gen_split): Likewise.
4626         (output_add_clobbers): Likewise.
4627         * genrecog.c (validate_pattern): Check for CLOBBER_HIGH.
4628         (remove_clobbers): Likewise.
4629         * rtl.h (gen_hard_reg_clobber_high): New declaration.
4631 2018-08-06  Alan Hayward  <alan.hayward@arm.com>
4633         * doc/rtl.texi (clobber_high): Add.
4634         (parallel): Add in clobber high
4635         * rtl.c (rtl_check_failed_code3): Add function.
4636         * rtl.def (CLOBBER_HIGH): Add expression.
4637         * rtl.h (RTL_CHECKC3): Add macro.
4638         (rtl_check_failed_code3): Add declaration.
4639         (XC3EXP): Add macro.
4641 2018-08-05  H.J. Lu  <hongjiu.lu@intel.com>
4643         PR target/86386
4644         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Set
4645         cfun->machine->max_used_stack_alignment if needed.
4647 2018-08-04  Martin Sebor  <msebor@redhat.com>
4649         PR tree-optimization/86571
4650         * gimple-ssa-sprintf.c (format_floating): Extend upper bound of
4651         NaN output to 4.
4653 2018-08-03  Sandra Loosemore  <sandra@codesourcery.com>
4655         PR target/86799
4656         * config/nios2/nios2.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4657         Define.
4659 2018-08-03  Jeff Law  <law@redhat.com>
4661         PR target/86795
4662         * config/mn10300/mn10300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4663         Define to speculation_safe_value_not_needed.
4665 2018-08-03  David Malcolm  <dmalcolm@redhat.com>
4667         * doc/gcov.texi (-x): Remove duplicate "to".
4668         * doc/invoke.texi (-Wnoexcept-type): Remove duplicate "calls".
4669         (-Wif-not-aligned): Remove duplicate "is".
4670         (-flto): Remove duplicate "the".
4671         (MicroBlaze Options): In examples of "-mcpu=cpu-type", remove
4672         duplicate "v5.00.b".
4673         (MSP430 Options): Remove duplicate "and" from the description
4674         of "-mgprel-sec=regexp".
4675         (x86 Options): Remove duplicate copies of "vmldLog102" and
4676         vmlsLog104 from description of "-mveclibabi=type".
4678 2018-08-03  Richard Sandiford  <richard.sandiford@arm.com>
4680         * internal-fn.h (first_commutative_argument): Declare.
4681         * internal-fn.c (first_commutative_argument): New function.
4682         * tree-vect-slp.c (vect_get_and_check_slp_defs): Remove extra
4683         restrictions for pattern statements.  Use first_commutative_argument
4684         to look for commutative operands in calls to internal functions.
4686 2018-08-03  Aldy Hernandez  <aldyh@redhat.com>
4688         * Makefile.in (wide-int-range.o): New.
4689         * tree-vrp.c: Move all the wide_int_* functions to...
4690         * wide-int-range.cc: ...here.
4691         * tree-vrp.h: Move all the wide_int_* prototypes to...
4692         * wide-int-range.h: ...here.
4694 2018-08-03  Tom de Vries  <tdevries@suse.de>
4696         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): Return
4697         UI_NONE.
4698         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Remove define.
4699         * except.c (output_function_exception_table): Do early exit if
4700         targetm_common.except_unwind_info (&global_options) == UI_NONE.
4702 2018-08-03  Martin Liska  <mliska@suse.cz>
4704         * predict.c (dump_prediction): Change to 2 digits
4705         in fraction part.
4707 2018-08-03  Siddhesh Poyarekar  <siddhesh@sourceware.org>
4709         * config/aarch64/falkor.md (falkor_am_1_vxvy_vxvy): Move
4710         neon_dup_q to...
4711         (falkor_am_1_gtov_gtov): ... a new insn reservation.
4713 2018-07-19  Ilya Leoshkevich  <iii@linux.ibm.com>
4715         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Replace > with >=.
4716         * df-problems.c (df_remove_dead_eq_notes): Replace > with >=.
4717         * dwarf2out.c (mem_loc_descriptor): Replace > with >=.
4718         * lra-constraints.c (spill_hard_reg_in_range): Replace <= with <.
4719         * lra-remat.c (call_used_input_regno_present_p): Replace <= with <.
4721 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
4723         * diagnostic-show-locus.c (diagnostic_show_locus): Use
4724         pp_take_prefix when saving the existing prefix.
4725         * diagnostic.c (diagnostic_append_note): Likewise.
4726         * langhooks.c (lhd_print_error_function): Likewise.
4727         * pretty-print.c (pp_set_prefix): Drop the "const" from "prefix"
4728         param's type.  Free the existing prefix.
4729         (pp_take_prefix): New function.
4730         (pretty_printer::pretty_printer): Drop the prefix parameter.
4731         Rename the length parameter to match the comment.
4732         (pretty_printer::~pretty_printer): Free the prefix.
4733         * pretty-print.h (pretty_printer::pretty_printer): Drop the prefix
4734         parameter.
4735         (struct pretty_printer): Drop the "const" from "prefix" field's
4736         type and clarify memory management.
4737         (pp_set_prefix): Drop the "const" from the 2nd param.
4738         (pp_take_prefix): New decl.
4740 2018-08-02  Aaron Sawdey  <acsawdey@linux.ibm.com>
4742         * config/rs6000/rs6000-string.c (select_block_compare_mode): Move test
4743         for word_mode_ok here instead of passing as argument.
4744         (expand_block_compare): Change select_block_compare_mode() call.
4745         (expand_strncmp_gpr_sequence): New function.
4746         (expand_strn_compare): Make use of expand_strncmp_gpr_sequence.
4748 2018-08-02  Jeff Law  <law@redhat.com>
4750         PR target/86790
4751         * config/m68k/m68k.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4752         Define to speculation_safe_value_not_needed.
4754         PR target/86784
4755         * config/h8300/h8300.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4756         Define to speculation_safe_value_not_needed.
4758 2018-08-02  Tom de Vries  <tdevries@suse.de>
4760         PR target/86660
4761         * common/config/nvptx/nvptx-common.c (nvptx_except_unwind_info): New
4762         function.  Return UI_TARGET unconditionally.
4763         (TARGET_EXCEPT_UNWIND_INFO): Redefine to nvptx_except_unwind_info.
4764         * config/nvptx/nvptx.c (TARGET_ASM_BYTE_OP): Emit commented out '.byte'.
4766 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
4768         * genemit.c (print_overload_test): Fix typo.
4770 2018-08-02  Richard Biener  <rguenther@suse.de>
4772         PR tree-optimization/86816
4773         * tree-ssa-tail-merge.c (tail_merge_valueize): New function
4774         which checks for value availability before querying it.
4775         (gvn_uses_equal): Use it.
4776         (same_succ_hash): Likewise.
4777         (gimple_equal_p): Likewise.
4779 2018-08-02  Nick Clifton  <nickc@redhat.com>
4781         PR target/86813
4782         * config/stormy16/stormy16.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4783         Define to speculation_safe_value_not_needed.
4785         PR target/86810
4786         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4787         Define to speculation_safe_value_not_needed.
4789         PR target/86810
4790         * config/v850/v850.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4791         Define to speculation_safe_value_not_needed.
4793         PR target/86803
4794         * config/rx/rx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4795         Define to speculation_safe_value_not_needed.
4797         PR target/86797
4798         * config/msp430/msp430.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4799         Define to speculation_safe_value_not_needed.
4801         PR target/86791
4802         * config/mcore/mcore.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4803         Define to speculation_safe_value_not_needed.
4805         PR target/86789
4806         * config/m32r/m32r.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4807         Define to speculation_safe_value_not_needed.
4809         PR target/86787
4810         * config/iq2000/iq2000.c (TARGET_HAVE_SPECULATION_SAFE_VALUE):
4811         Define to speculation_safe_value_not_needed.
4813         PR target/86782
4814         * config/frv/frv.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
4815         speculation_safe_value_not_needed.
4817         PR target/86781
4818         * config/fr30/fr30.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define
4819         to speculation_safe_value_not_needed.
4821 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
4823         * doc/md.texi: Expand the documentation of instruction names
4824         to mention port-local uses.  Document '@' in pattern names.
4825         * read-md.h (overloaded_instance, overloaded_name): New structs.
4826         (mapping): Declare.
4827         (md_reader::handle_overloaded_name): New member function.
4828         (md_reader::get_overloads): Likewise.
4829         (md_reader::m_first_overload): New member variable.
4830         (md_reader::m_next_overload_ptr): Likewise.
4831         (md_reader::m_overloads_htab): Likewise.
4832         * read-md.c (md_reader::md_reader): Initialize m_first_overload,
4833         m_next_overload_ptr and m_overloads_htab.
4834         * read-rtl.c (iterator_group): Add "type" and "get_c_token" fields.
4835         (get_mode_token, get_code_token, get_int_token): New functions.
4836         (map_attr_string): Add an optional argument that passes back
4837         the associated iterator.
4838         (overloaded_name_hash, overloaded_name_eq_p, named_rtx_p):
4839         (md_reader::handle_overloaded_name, add_overload_instance): New
4840         functions.
4841         (apply_iterators): Handle '@' names.  Report an error if '@'
4842         is used without iterators.
4843         (initialize_iterators): Initialize the new iterator_group fields.
4844         * genopinit.c (handle_overloaded_code_for)
4845         (handle_overloaded_gen): New functions.
4846         (main): Use them to print declarations of maybe_code_for_* and
4847         maybe_gen_* functions, and inline definitions of code_for_* and gen_*.
4848         * genemit.c (print_overload_arguments, print_overload_test)
4849         (handle_overloaded_code_for, handle_overloaded_gen): New functions.
4850         (main): Use it to print definitions of maybe_code_for_* and
4851         maybe_gen_* functions.
4852         * config/aarch64/aarch64.c (aarch64_split_128bit_move): Use
4853         gen_aarch64_mov{low,high}_di and gen_aarch64_movdi_{low,high}
4854         instead of explicit mode checks.
4855         (aarch64_split_simd_combine): Likewise gen_aarch64_simd_combine.
4856         (aarch64_split_simd_move): Likewise gen_aarch64_split_simd_mov.
4857         (aarch64_emit_load_exclusive): Likewise gen_aarch64_load_exclusive.
4858         (aarch64_emit_store_exclusive): Likewise gen_aarch64_store_exclusive.
4859         (aarch64_expand_compare_and_swap): Likewise
4860         gen_aarch64_compare_and_swap and gen_aarch64_compare_and_swap_lse
4861         (aarch64_gen_atomic_cas): Likewise gen_aarch64_atomic_cas.
4862         (aarch64_emit_atomic_swap): Likewise gen_aarch64_atomic_swp.
4863         (aarch64_constant_pool_reload_icode): Delete.
4864         (aarch64_secondary_reload): Use code_for_aarch64_reload_movcp
4865         instead of aarch64_constant_pool_reload_icode.  Use
4866         code_for_aarch64_reload_mov instead of explicit mode checks.
4867         (rsqrte_type, get_rsqrte_type, rsqrts_type, get_rsqrts_type): Delete.
4868         (aarch64_emit_approx_sqrt): Use gen_aarch64_rsqrte instead of
4869         get_rsqrte_type and gen_aarch64_rsqrts instead of gen_rqrts_type.
4870         (recpe_type, get_recpe_type, recps_type, get_recps_type): Delete.
4871         (aarch64_emit_approx_div): Use gen_aarch64_frecpe instead of
4872         get_recpe_type and gen_aarch64_frecps instead of get_recps_type.
4873         (aarch64_atomic_load_op_code): Delete.
4874         (aarch64_emit_atomic_load_op): Likewise.
4875         (aarch64_gen_atomic_ldop): Use UNSPECV_ATOMIC_* instead of
4876         aarch64_atomic_load_op_code.  Use gen_aarch64_atomic_load
4877         instead of aarch64_emit_atomic_load_op.
4878         * config/aarch64/aarch64.md (aarch64_reload_movcp<GPF_TF:mode><P:mode>)
4879         (aarch64_reload_movcp<VALL:mode><P:mode>, aarch64_reload_mov<mode>)
4880         (aarch64_movdi_<mode>low, aarch64_movdi_<mode>high)
4881         (aarch64_mov<mode>high_di, aarch64_mov<mode>low_di): Add a '@'
4882         character before the pattern name.
4883         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
4884         (aarch64_rsqrte<mode>, aarch64_rsqrts<mode>)
4885         (aarch64_simd_combine<mode>, aarch64_frecpe<mode>)
4886         (aarch64_frecps<mode>): Likewise.
4887         * config/aarch64/atomics.md (atomic_compare_and_swap<mode>)
4888         (aarch64_compare_and_swap<mode>, aarch64_compare_and_swap<mode>_lse)
4889         (aarch64_load_exclusive<mode>, aarch64_store_exclusive<mode>)
4890         (aarch64_atomic_swp<mode>, aarch64_atomic_cas<mode>)
4891         (aarch64_atomic_load<atomic_ldop><mode>): Likewise.
4893 2018-08-02  Richard Sandiford  <richard.sandiford@arm.com>
4895         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
4896         Allow HFmode constants if TARGET_FP_F16INST.
4898 2018-08-02  Jackson Woodruff  <jackson.woodruff@arm.com>
4900         PR target/86014
4901         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp):
4902         No longer check last store for clobber of address register.
4904 2018-08-02  Martin Liska  <mliska@suse.cz>
4906         PR gcov-profile/86817
4907         * gcov.c (process_all_functions): New function.
4908         (main): Call it.
4909         (process_file): Move functions processing to
4910         process_all_functions.
4912 2018-08-02  David Malcolm  <dmalcolm@redhat.com>
4914         * dumpfile.c (dump_user_location_t::dump_user_location_t): Add
4915         "const" to the "gimple *" and "rtx_insn *" parameters.
4916         * dumpfile.h (dump_user_location_t::dump_user_location_t):
4917         Likewise.
4918         (dump_location_t::dump_location_t): Likewise.
4920 2018-08-01  Martin Sebor  <msebor@redhat.com>
4922         PR tree-optimization/86650
4923         * gimple-pretty-print.c (percent_G_format): Accept a "gimple *"
4924         rather than a "gcall *".  Directly pass the data of interest
4925         to percent_K_format, rather than building a temporary CALL_EXPR
4926         to hold it.
4927         * gimple-fold.c (gimple_fold_builtin_strncpy): Adjust.
4928         (gimple_fold_builtin_strncat): Adjust.
4929         * gimple-ssa-warn-restrict.h (check_bounds_or_overlap): Replace
4930         gcall* argument with gimple*.
4931         * gimple-ssa-warn-restrict.c (check_call): Same.
4932         (wrestrict_dom_walker::before_dom_children): Same.
4933         (builtin_access::builtin_access): Same.
4934         (check_bounds_or_overlap): Same
4935         (maybe_diag_overlap): Same.
4936         (maybe_diag_offset_bounds): Same.
4937         * tree-diagnostic.c (default_tree_printer): Move usage of
4938         EXPR_LOCATION (t) and TREE_BLOCK (t) from within percent_K_format
4939         to this callsite.
4940         * tree-pretty-print.c (percent_K_format): Add argument.
4941         * tree-pretty-print.h: Add argument.
4942         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Adjust.
4943         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Adjust.
4944         (maybe_diag_stxncpy_trunc): Same.
4945         (handle_builtin_stxncpy): Same.
4946         (handle_builtin_strcat): Same.
4948 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4950         * match.pd: Optimise pointer range checks.
4952 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4954         PR tree-optimization/86758
4955         * tree-vect-stmts.c (vectorizable_simd_clone_call): Don't try
4956         to remove pattern statements.
4958 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4960         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Use the
4961         result of dfs_enumerate_from when constructing stmt_vec_infos,
4962         instead of additionally calling get_loop_body.
4964 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4966         * tree-vectorizer.h (vect_create_data_ref_ptr): Remove inv_p
4967         parameter.
4968         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Likewise.
4969         When creating an iv, assert that the step is not known to be zero.
4970         (vect_setup_realignment): Update call accordingly.
4971         * tree-vect-stmts.c (vectorizable_store): Likewise.
4972         (vectorizable_load): Likewise.  Handle VMAT_INVARIANT separately.
4974 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
4976         * tree-vectorizer.h (vect_stmt_to_vectorize): New function.
4977         * tree-vect-loop.c (vect_update_vf_for_slp): Use it.
4978         (vectorizable_reduction): Likewise.
4979         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4980         (vect_detect_hybrid_slp_stmts): Likewise.
4981         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
4983 2018-08-01  Aldy Hernandez  <aldyh@redhat.com>
4985         * tree-vrp (zero_nonzero_bits_from_bounds): Rename to...
4986         (wide_int_set_zero_nonzero_bits): ...this.
4987         (zero_nonzero_bits_from_vr): Rename to...
4988         (vrp_set_zero_nonzero_bits): ...this.
4989         (extract_range_from_multiplicative_op_1): Abstract wide int
4990         code...
4991         (wide_int_range_multiplicative_op): ...here.
4992         (extract_range_from_binary_expr_1): Extract wide int binary
4993         operations into their own functions.
4994         (wide_int_range_lshift): New.
4995         (wide_int_range_can_optimize_bit_op): New.
4996         (wide_int_range_shift_undefined_p): New.
4997         (wide_int_range_bit_xor): New.
4998         (wide_int_range_bit_ior): New.
4999         (wide_int_range_bit_and): New.
5000         (wide_int_range_trunc_mod): New.
5001         (extract_range_into_wide_ints): New.
5002         (vrp_shift_undefined_p): New.
5003         (extract_range_from_multiplicative_op): New.
5004         (vrp_can_optimize_bit_op): New.
5005         * tree-vrp.h (value_range::dump): New.
5006         (wide_int_range_multiplicative_op): New.
5007         (wide_int_range_lshift):New.
5008         (wide_int_range_shift_undefined_p): New.
5009         (wide_int_range_bit_xor): New.
5010         (wide_int_range_bit_ior): New.
5011         (wide_int_range_bit_and): New.
5012         (wide_int_range_trunc_mod): New.
5013         (zero_nonzero_bits_from_bounds): Rename to...
5014         (wide_int_set_zero_nonzero_bits): ...this.
5015         (zero_nonzero_bits_from_vr): Rename to...
5016         (vrp_set_zero_nonzero_bits): ...this.
5017         (range_easy_mask_min_max): Rename to...
5018         (wide_int_range_can_optimize_bit_op): this.
5019         * vr-values.c (simplify_bit_ops_using_ranges): Rename
5020         zero_nonzero_bits_from_vr into vrp_set_zero_nonzero_bits.
5022 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5024         * tree-vectorizer.h (vect_orig_stmt): New function.
5025         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Use it.
5026         * tree-vect-loop.c (vect_model_reduction_cost): Likewise.
5027         (vect_create_epilog_for_reduction): Likewise.
5028         (vectorizable_live_operation): Likewise.
5029         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
5030         (vect_detect_hybrid_slp_stmts, vect_schedule_slp): Likewise.
5031         * tree-vect-stmts.c (vectorizable_call): Likewise.
5032         (vectorizable_simd_clone_call, vect_remove_stores): Likewise.
5034 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5036         * tree-vectorizer.h (vect_transform_stmt): Remove grouped_store
5037         argument.
5038         * tree-vect-stmts.c (vect_transform_stmt): Likewise.
5039         * tree-vect-loop.c (vect_transform_loop_stmt): Update call accordingly.
5040         (vect_transform_loop): Likewise.
5041         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5043 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5045         * tree-vectorizer.h (vect_schedule_slp): Return void.
5046         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
5047         (vect_schedule_slp): Likewise.
5049 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5051         * tree-vect-loop.c (vect_transform_loop_stmt): Remove slp_scheduled
5052         argument.
5053         (vect_transform_loop): Update calls accordingly.  Schedule SLP
5054         instances before the main loop, if any exist.
5056 2018-08-01  Richard Sandiford  <richard.sandiford@arm.com>
5058         PR tree-optimization/86749
5059         * tree-vect-patterns.c (vect_determine_min_output_precision_1):
5060         If the lhs is used in a COND_EXPR, check that it is being used
5061         as the "then" or "else" value.
5063 2018-08-01  Tom de Vries  <tdevries@suse.de>
5065         PR target/86800
5066         * config/nvptx/nvptx.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Define to
5067         speculation_safe_value_not_needed.
5069 2018-08-01  Richard Biener  <rguenther@suse.de>
5071         * tree-ssa-sccvn.c (visit_phi): Compare invariant addresses
5072         as base and offset.
5074 2018-08-01  Martin Liska  <mliska@suse.cz>
5076         * value-prof.c (gimple_divmod_fixed_value_transform): Unify
5077         format how successful transformation is dumped.
5078         (gimple_mod_pow2_value_transform): Likewise.
5079         (gimple_mod_subtract_transform): Likewise.
5080         (gimple_stringops_transform): Likewise.
5082 2018-08-01  Martin Liska  <mliska@suse.cz>
5084         PR value-prof/35543
5085         * value-prof.c (interesting_stringop_to_profile_p):
5086         Simplify the code and add BUILT_IN_MEMMOVE.
5087         (gimple_stringops_transform): Likewise.
5089 2018-08-01  Sam Tebbs  <sam.tebbs@arm.com>
5091         * config/aarch64/aarch64-simd.md
5092         (*aarch64_get_lane_zero_extendsi<mode>): Rename to...
5093         (*aarch64_get_lane_zero_extend<GPI:mode><VDQQH:mode>): ... This and
5094         use GPI iterator instead of SI mode.
5096 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
5098         * config/rs6000/rs6000.md (speculation_barrier): Renamed from
5099         rs6000_speculation_barrier.
5100         * config/rs6000/rs6000.c (rs6000_expand_builtin): Adjust for
5101         new barrier pattern name.
5103 2018-08-01  Richard Earnshaw  <rearnsha@arm.com>
5105         * config/i386/i386.md (unspecv): Add UNSPECV_SPECULATION_BARRIER.
5106         (speculation_barrier): New insn.
5108 2018-08-01  Richard Biener  <rguenther@suse.de>
5110         PR bootstrap/86724
5111         * graphite.h: Include isl/id.h and isl/space.h to allow build
5112         with ISL 0.20.
5114 2018-08-01  Jan Willem Jagersma  <jwjagersma@gmail.com>
5116         PR target/86651
5117         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
5118         mode for COFF targets.
5119         * defaults.h (TARGET_COFF): Define.
5120         * config/i386/djgpp.h (TARGET_ASM_LTO_START, TARGET_ASM_LTO_END,
5121         TARGET_COFF): Define.
5122         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): Declare.
5123         * config/i386/djgpp.c (saved_debug_info_level): New static variable.
5124         (i386_djgpp_asm_lto_start, i386_djgpp_asm_lto_end): New functions.
5126 2018-07-31  Alexandre Oliva <oliva@adacore.com>
5128         * gimple-streamer-in.c (input_bb): Restore BB discriminator.
5129         * gimple-streamer-out.c (output_bb): Save it.
5130         * lto-streamer-in.c (input_struct_function_base): Restore
5131         instance discriminator if available.  Create map on demand.
5132         * lto-streamer-out.c (output_struct_function_base): Save it if
5133         available.
5134         * final.c (decl_to_instance_map): Document LTO strategy.
5136 2018-07-31  Alexandre Oliva  <oliva@adacore.com>
5137             Olivier Hainque  <hainque@adacore.com>
5139         * debug.h (decl_to_instance_map_t): New type.
5140         (decl_to_instance_map): Declare.
5141         (maybe_create_decl_to_instance_map): New inline function.
5142         * final.c (bb_discriminator, last_bb_discriminator): New statics,
5143         to track basic block discriminators.
5144         (final_start_function_1): Initialize them.
5145         (final_scan_insn_1): On NOTE_INSN_BASIC_BLOCK, track
5146         bb_discriminator.
5147         (decl_to_instance_map): New variable.
5148         (map_decl_to_instance, maybe_set_discriminator): New functions.
5149         (notice_source_line): Set discriminator.
5151 2018-07-31  Ian Lance Taylor  <iant@golang.org>
5153         * targhooks.c (default_have_speculation_safe_value): Add
5154         ATTRIBUTE_UNUSED.
5156 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
5158         * dump-context.h: Include "pretty-print.h".
5159         (dump_context::refresh_dumps_are_enabled): New decl.
5160         (dump_context::emit_item): New decl.
5161         (class dump_context): Add fields "m_test_pp" and
5162         "m_test_pp_flags".
5163         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
5164         (temp_dump_context::get_dumped_text): New decl.
5165         (class temp_dump_context): Add field "m_pp".
5166         * dumpfile.c (refresh_dumps_are_enabled): Convert to...
5167         (dump_context::refresh_dumps_are_enabled): ...and add a test for
5168         m_test_pp.
5169         (set_dump_file): Update for above change.
5170         (set_alt_dump_file): Likewise.
5171         (dump_loc): New overload, taking a pretty_printer *.
5172         (dump_context::dump_loc): Call end_any_optinfo.  Dump the location
5173         to any test pretty-printer.
5174         (make_item_for_dump_gimple_stmt): New function, adapted from
5175         optinfo::add_gimple_stmt.
5176         (dump_context::dump_gimple_stmt): Call it, and use the result,
5177         eliminating the direct usage of dump_file and alt_dump_file in
5178         favor of indirectly using them via emit_item.
5179         (make_item_for_dump_gimple_expr): New function, adapted from
5180         optinfo::add_gimple_expr.
5181         (dump_context::dump_gimple_expr): Call it, and use the result,
5182         eliminating the direct usage of dump_file and alt_dump_file in
5183         favor of indirectly using them via emit_item.
5184         (make_item_for_dump_generic_expr): New function, adapted from
5185         optinfo::add_tree.
5186         (dump_context::dump_generic_expr): Call it, and use the result,
5187         eliminating the direct usage of dump_file and alt_dump_file in
5188         favor of indirectly using them via emit_item.
5189         (make_item_for_dump_printf_va): New function, adapted from
5190         optinfo::add_printf_va.
5191         (make_item_for_dump_printf): New function.
5192         (dump_context::dump_printf_va): Call make_item_for_dump_printf_va,
5193         and use the result, eliminating the direct usage of dump_file and
5194         alt_dump_file in favor of indirectly using them via emit_item.
5195         (make_item_for_dump_dec): New function.
5196         (dump_context::dump_dec): Call it, and use the result,
5197         eliminating the direct usage of dump_file and alt_dump_file in
5198         favor of indirectly using them via emit_item.
5199         (make_item_for_dump_symtab_node): New function, adapted from
5200         optinfo::add_symtab_node.
5201         (dump_context::dump_symtab_node): Call it, and use the result,
5202         eliminating the direct usage of dump_file and alt_dump_file in
5203         favor of indirectly using them via emit_item.
5204         (dump_context::begin_scope): Reimplement, avoiding direct usage
5205         of dump_file and alt_dump_file in favor of indirectly using them
5206         via emit_item.
5207         (dump_context::emit_item): New member function.
5208         (temp_dump_context::temp_dump_context): Add param "test_pp_flags".
5209         Set up test pretty-printer on the underlying context.  Call
5210         refresh_dumps_are_enabled.
5211         (temp_dump_context::~temp_dump_context): Call
5212         refresh_dumps_are_enabled.
5213         (temp_dump_context::get_dumped_text): New member function.
5214         (selftest::verify_dumped_text): New function.
5215         (ASSERT_DUMPED_TEXT_EQ): New macro.
5216         (selftest::test_capture_of_dump_calls): Run all tests twice, with
5217         and then without optinfo enabled.  Add uses of
5218         ASSERT_DUMPED_TEXT_EQ to all tests.  Add test of nested scopes.
5219         * dumpfile.h: Update comment for the dump_* API.
5220         * optinfo-emit-json.cc
5221         (selftest::test_building_json_from_dump_calls): Update for new
5222         param for temp_dump_context ctor.
5223         * optinfo.cc (optinfo_item::optinfo_item): Remove "owned" param
5224         and "m_owned" field.
5225         (optinfo_item::~optinfo_item): Likewise.
5226         (optinfo::add_item): New member function.
5227         (optinfo::emit): Update comment.
5228         (optinfo::add_string): Delete.
5229         (optinfo::add_printf): Delete.
5230         (optinfo::add_printf_va): Delete.
5231         (optinfo::add_gimple_stmt): Delete.
5232         (optinfo::add_gimple_expr): Delete.
5233         (optinfo::add_tree): Delete.
5234         (optinfo::add_symtab_node): Delete.
5235         (optinfo::add_dec): Delete.
5236         * optinfo.h (class dump_context): New forward decl.
5237         (optinfo::add_item): New decl.
5238         (optinfo::add_string): Delete.
5239         (optinfo::add_printf): Delete.
5240         (optinfo::add_printf_va): Delete.
5241         (optinfo::add_gimple_stmt): Delete.
5242         (optinfo::add_gimple_expr): Delete.
5243         (optinfo::add_tree): Delete.
5244         (optinfo::add_symtab_node): Delete.
5245         (optinfo::add_dec): Delete.
5246         (optinfo::add_poly_int): Delete.
5247         (optinfo_item::optinfo_item): Remove "owned" param.
5248         (class optinfo_item): Remove field "m_owned".
5250 2018-07-31  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
5252         PR middle-end/86705
5253         * gcc/cfgexpand.c (set_parm_rtl): Use the alignment of Pmode when
5254         MAX_SUPPORTED_STACK_ALIGNMENT would otherwise be exceeded by the
5255         requested variable alignment.
5256         (expand_one_ssa_partition): Likewise.
5257         (expand_one_var): Likewise.
5259 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
5261         * config/pdp11/pdp11.c (TARGET_HAVE_SPECULATION_SAFE_VALUE): Redefine
5262         to speculation_safe_value_not_needed.
5264 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
5266         * targhooks.h (speculation_safe_value_not_needed): New prototype.
5267         * targhooks.c (speculation_safe_value_not_needed): New function.
5268         * target.def (have_speculation_safe_value): Update documentation.
5269         * doc/tm.texi: Regenerated.
5271 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
5273         * config/aarch64/iterators.md (ALLI_TI): New iterator.
5274         * config/aarch64/aarch64.md (despeculate_copy<ALLI_TI:mode>): New
5275         expand.
5276         (despeculate_copy<ALLI:mode>_insn): New insn.
5277         (despeculate_copyti_insn): New insn.
5278         (despeculate_simple<ALLI:mode>): New insn
5279         (despeculate_simpleti): New insn.
5280         * config/aarch64/aarch64.c (aarch64_speculation_safe_value): New
5281         function.
5282         (TARGET_SPECULATION_SAFE_VALUE): Redefine to
5283         aarch64_speculation_safe_value.
5284         (aarch64_print_operand): Handle const0_rtx in modifier 'H'.
5286 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
5288         * config/aarch64/aarch64-speculation.cc: New file.
5289         * config/aarch64/aarch64-passes.def (pass_track_speculation): Add
5290         before pass_reorder_blocks.
5291         * config/aarch64/aarch64-protos.h (make_pass_track_speculation): Add
5292         prototype.
5293         * config/aarch64/aarch64.c (aarch64_conditional_register_usage): Fix
5294         X14 and X15 when tracking speculation.
5295         * config/aarch64/aarch64.md (register name constants): Add
5296         SPECULATION_TRACKER_REGNUM and SPECULATION_SCRATCH_REGNUM.
5297         (unspec): Add UNSPEC_SPECULATION_TRACKER.
5298         (speculation_barrier): New insn attribute.
5299         (cmp<mode>): Allow SP in comparisons.
5300         (speculation_tracker): New insn.
5301         (speculation_barrier): Add speculation_barrier attribute.
5302         * config/aarch64/t-aarch64: Add make rule for aarch64-speculation.o.
5303         * config.gcc (aarch64*-*-*): Add aarch64-speculation.o to extra_objs.
5304         * doc/invoke.texi (AArch64 Options): Document -mtrack-speculation.
5306 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
5308         * config/aarch64/aarch64.md (cb<optab><mode>1): Disable when
5309         aarch64_track_speculation is true.
5310         (tb<optab><mode>1): Likewise.
5311         * config/aarch64/aarch64.c (aarch64_split_compare_regs): Do not
5312         generate CB[N]Z when tracking speculation.
5313         (aarch64_split_compare_and_swap): Likewise.
5314         (aarch64_split_atomic_op): Likewise.
5316 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
5318         * config/aarch64/aarch64.opt (mtrack-speculation): New target option.
5320 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
5322         * config/aarch64.md (unspecv): Add UNSPECV_SPECULAION_BARRIER.
5323         (speculation_barrier): New insn.
5325 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
5327         * config/arm/unspecs.md (unspecv): Add VUNSPEC_SPECULATION_BARRIER.
5328         * config/arm/arm.md (speculation_barrier): New expand.
5329         (speculation_barrier_insn): New pattern.
5331 2018-07-31  Richard Earnshaw  <rearnsha@arm.com>
5333         * builtin-types.def (BT_FN_PTR_PTR_VAR): New function type.
5334         (BT_FN_I1_I1_VAR, BT_FN_I2_I2_VAR, BT_FN_I4_I4_VAR): Likewise.
5335         (BT_FN_I8_I8_VAR, BT_FN_I16_I16_VAR): Likewise.
5336         * builtin-attrs.def (ATTR_NOVOPS_NOTHROW_LEAF_LIST): New attribute
5337         list.
5338         * builtins.def (BUILT_IN_SPECULATION_SAFE_VALUE_N): New builtin.
5339         (BUILT_IN_SPECULATION_SAFE_VALUE_PTR): New internal builtin.
5340         (BUILT_IN_SPECULATION_SAFE_VALUE_1): Likewise.
5341         (BUILT_IN_SPECULATION_SAFE_VALUE_2): Likewise.
5342         (BUILT_IN_SPECULATION_SAFE_VALUE_4): Likewise.
5343         (BUILT_IN_SPECULATION_SAFE_VALUE_8): Likewise.
5344         (BUILT_IN_SPECULATION_SAFE_VALUE_16): Likewise.
5345         * builtins.c (expand_speculation_safe_value): New function.
5346         (expand_builtin): Call it.
5347         * doc/cpp.texi: Document predefine __HAVE_SPECULATION_SAFE_VALUE.
5348         * doc/extend.texi: Document __builtin_speculation_safe_value.
5349         * doc/md.texi: Document "speculation_barrier" pattern.
5350         * doc/tm.texi.in: Pull in TARGET_SPECULATION_SAFE_VALUE and
5351         TARGET_HAVE_SPECULATION_SAFE_VALUE.
5352         * doc/tm.texi: Regenerated.
5353         * target.def (have_speculation_safe_value, speculation_safe_value): New
5354         hooks.
5355         * targhooks.c (default_have_speculation_safe_value): New function.
5356         (default_speculation_safe_value): New function.
5357         * targhooks.h (default_have_speculation_safe_value): Add prototype.
5358         (default_speculation_safe_value): Add prototype.
5360 2018-07-31  David Malcolm  <dmalcolm@redhat.com>
5362         * dump-context.h (dump_context::dump_loc): New decl.
5363         * dumpfile.c (dump_context::dump_loc): New member function.
5364         (dump_context::dump_gimple_stmt_loc): Reimplement using dump_loc
5365         and dump_gimple_stmt.
5366         (dump_context::dump_gimple_expr_loc): Likewise, using
5367         dump_gimple_expr.
5368         (dump_context::dump_generic_expr_loc): Likewise, using
5369         dump_generic_expr.
5370         (dump_context::dump_printf_loc_va): Likewise, using
5371         dump_printf_va.
5372         (dump_context::begin_scope): Explicitly using the global function
5373         "dump_loc", rather than the member function.
5375 2018-07-31  Martin Sebor  <msebor@redhat.com>
5377         PR tree-optimization/86741
5378         * tree-vrp.c (vrp_prop::check_mem_ref): Avoid incomplete types.
5380 2018-07-31  Andreas Krebbel  <krebbel@linux.ibm.com>
5382         * config/s390/s390.c (s390_expand_setmem): Make the unrolling to
5383         depend on whether prefetch instructions will be emitted or not.
5384         Use TARGET_SETMEM_PFD for checking whether prefetch instructions
5385         will be emitted or not.
5386         * config/s390/s390.h (TARGET_SETMEM_PREFETCH_DISTANCE)
5387         (TARGET_SETMEM_PFD): New macros.
5389 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5391         * tree-vectorizer.h (stmt_vec_info): Turn back into a typedef.
5392         (NULL_STMT_VEC_INFO): Delete.
5393         (stmt_vec_info::operator*): Likewise.
5394         (stmt_vec_info::operator gimple *): Likewise.
5395         * tree-vect-loop.c (vectorizable_reduction): Use NULL instead
5396         of NULL_STMT_VEC_INFO.
5397         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
5398         (vect_reassociating_reduction_p): Likewise.
5399         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
5400         (vectorizable_store): Likewise.
5401         * tree-vectorizer.c (vec_info::set_vinfo_for_stmt): Likewise.
5402         (vec_info::free_stmt_vec_infos): Likewise.
5404 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5406         * tree-vectorizer.h (vect_stmt_in_region_p): Delete.
5407         * tree-vectorizer.c (vect_stmt_in_region_p): Likewise.
5409 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5411         * tree-vectorizer.h (vec_info::new_vinfo_for_stmt)
5412         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
5413         (vec_info::free_stmt_vec_info): New private member functions.
5414         (set_stmt_vec_info_vec, free_stmt_vec_infos, vinfo_for_stmt)
5415         (set_vinfo_for_stmt, new_stmt_vec_info, free_stmt_vec_info): Delete.
5416         * tree-parloops.c (gather_scalar_reductions): Remove calls to
5417         set_stmt_vec_info_vec and free_stmt_vec_infos.
5418         * tree-vect-loop.c (_loop_vec_info): Remove call to
5419         set_stmt_vec_info_vec.
5420         * tree-vect-stmts.c (new_stmt_vec_info, set_stmt_vec_info_vec)
5421         (free_stmt_vec_infos, free_stmt_vec_info): Delete in favor of...
5422         * tree-vectorizer.c (vec_info::new_stmt_vec_info)
5423         (vec_info::set_vinfo_for_stmt, vec_info::free_stmt_vec_infos)
5424         (vec_info::free_stmt_vec_info): ...these new functions.  Remove
5425         assignments in {vec_info::,}new_stmt_vec_info that are redundant
5426         with the clearing in the xcalloc.
5427         (stmt_vec_info_vec): Delete.
5428         (vec_info::vec_info): Don't call set_stmt_vec_info_vec.
5429         (vectorize_loops): Likewise.
5430         (vec_info::~vec_info): Remove argument from call to
5431         free_stmt_vec_infos.
5432         (vec_info::add_stmt): Remove vinfo argument from call to
5433         new_stmt_vec_info.
5435 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5437         * tree-vectorizer.h (free_stmt_vec_info): Take a stmt_vec_info
5438         rather than a gimple stmt.
5439         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.  Don't free
5440         information for pattern statements when passed the original
5441         statement; instead wait to be passed the pattern statement itself.
5442         Don't call set_vinfo_for_stmt here.
5443         (free_stmt_vec_infos): Update call to free_stmt_vec_info.
5444         * tree-vect-loop.c (_loop_vec_info::~loop_vec_info): Don't free
5445         stmt_vec_infos here.
5446         * tree-vect-slp.c (_bb_vec_info::~bb_vec_info): Likewise.
5447         * tree-vectorizer.c (vec_info::remove_stmt): Nullify the statement's
5448         stmt_vec_infos entry.
5450 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5452         * tree-vectorizer.h (vec_info::replace_stmt): Declare.
5453         * tree-vectorizer.c (vec_info::replace_stmt): New function.
5454         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Use it.
5455         * tree-vect-stmts.c (vectorizable_call): Likewise.
5456         (vectorizable_simd_clone_call): Likewise.
5458 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5460         * tree-vectorizer.h (vec_info::remove_stmt): Declare.
5461         * tree-vectorizer.c (vec_info::remove_stmt): New function.
5462         * tree-vect-loop-manip.c (vect_set_loop_condition): Use it.
5463         * tree-vect-loop.c (vect_transform_loop): Likewise.
5464         * tree-vect-slp.c (vect_schedule_slp): Likewise.
5465         * tree-vect-stmts.c (vect_remove_stores): Likewise.
5467 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5469         * tree-vectorizer.h (vec_info::lookup_dr): New member function.
5470         (vect_dr_stmt): Delete.
5471         * tree-vectorizer.c (vec_info::lookup_dr): New function.
5472         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Use it instead
5473         of DR_VECT_AUX.
5474         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
5475         (vect_analyze_data_ref_dependence, vect_record_base_alignments)
5476         (vect_verify_datarefs_alignment, vect_peeling_supportable)
5477         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
5478         (vect_analyze_data_refs): Likewise.
5479         (vect_slp_analyze_data_ref_dependence): Likewise.  Take a vec_info
5480         argument.
5481         (vect_find_same_alignment_drs): Likewise.
5482         (vect_slp_analyze_node_dependences): Update calls accordingly.
5483         (vect_analyze_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
5484         instead of DR_VECT_AUX.
5485         (vect_get_peeling_costs_all_drs): Take a loop_vec_info instead
5486         of a vector data references.  Use vec_info::lookup_dr instead of
5487         DR_VECT_AUX.
5488         (vect_peeling_hash_get_lowest_cost): Update calls accordingly.
5489         (vect_enhance_data_refs_alignment): Likewise.  Use vec_info::lookup_dr
5490         instead of DR_VECT_AUX.
5492 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5494         * tree-vectorizer.h (_loop_vec_info::unaligned_dr): Change to
5495         dr_vec_info.
5496         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
5497         accordingly.
5498         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5499         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5500         (vect_gen_prolog_loop_niters): Likewise.
5502 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5504         * tree-vectorizer.h (set_dr_misalignment, dr_misalignment)
5505         (DR_TARGET_ALIGNMENT, aligned_access_p, known_alignment_for_access_p)
5506         (vect_known_alignment_in_bytes, vect_dr_behavior)
5507         (vect_get_scalar_dr_size): Take references as dr_vec_infos
5508         instead of data_references.  Update calls to other routines for
5509         which the same change has been made.
5510         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Take
5511         dr_vec_infos instead of stmt_vec_infos.
5512         (vect_analyze_data_ref_dependence): Update call accordingly.
5513         (vect_slp_analyze_data_ref_dependence)
5514         (vect_record_base_alignments): Use DR_VECT_AUX.
5515         (vect_calculate_target_alignment, vect_compute_data_ref_alignment)
5516         (vect_update_misalignment_for_peel, verify_data_ref_alignment)
5517         (vector_alignment_reachable_p, vect_get_data_access_cost)
5518         (vect_peeling_supportable, vect_analyze_group_access_1)
5519         (vect_analyze_group_access, vect_analyze_data_ref_access)
5520         (vect_vfa_segment_size, vect_vfa_access_size, vect_vfa_align)
5521         (vect_compile_time_alias, vect_small_gap_p)
5522         (vectorizable_with_step_bound_p, vect_duplicate_ssa_name_ptr_info):
5523         (vect_supportable_dr_alignment): Take references as dr_vec_infos
5524         instead of data_references.  Update calls to other routines for
5525         which the same change has been made.
5526         (vect_verify_datarefs_alignment, vect_get_peeling_costs_all_drs)
5527         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
5528         (vect_slp_analyze_and_verify_node_alignment)
5529         (vect_analyze_data_ref_accesses, vect_prune_runtime_alias_test_list)
5530         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
5531         (vect_setup_realignment): Use dr_vec_infos.  Update calls after
5532         above changes.
5533         (_vect_peel_info::dr): Replace with...
5534         (_vect_peel_info::dr_info): ...this new field.
5535         (vect_peeling_hash_get_most_frequent)
5536         (vect_peeling_hash_choose_best_peeling): Update accordingly.
5537         (vect_peeling_hash_get_lowest_cost):
5538         (vect_enhance_data_refs_alignment): Likewise.  Update calls to other
5539         routines for which the same change has been made.
5540         (vect_peeling_hash_insert): Likewise.  Take a dr_vec_info instead of a
5541         data_reference.
5542         * tree-vect-loop-manip.c (get_misalign_in_elems)
5543         (vect_gen_prolog_loop_niters): Use dr_vec_infos.  Update calls after
5544         above changes.
5545         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5546         * tree-vect-stmts.c (vect_get_store_cost, vect_get_load_cost)
5547         (vect_truncate_gather_scatter_offset, compare_step_with_zero)
5548         (get_group_load_store_type, get_negative_load_store_type)
5549         (vect_get_data_ptr_increment, vectorizable_store)
5550         (vectorizable_load): Likewise.
5551         (ensure_base_align): Take a dr_vec_info instead of a data_reference.
5552         Update calls to other routines for which the same change has been made.
5554 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5556         * tree-vectorizer.h (vec_info::move_dr): New member function.
5557         (dataref_aux): Rename to...
5558         (dr_vec_info): ...this and add "dr" and "stmt" fields.
5559         (_stmt_vec_info::dr_aux): Update accordingly.
5560         (_stmt_vec_info::data_ref_info): Delete.
5561         (STMT_VINFO_GROUPED_ACCESS, DR_GROUP_FIRST_ELEMENT)
5562         (DR_GROUP_NEXT_ELEMENT, DR_GROUP_SIZE, DR_GROUP_STORE_COUNT)
5563         (DR_GROUP_GAP, DR_GROUP_SAME_DR_STMT, REDUC_GROUP_FIRST_ELEMENT):
5564         (REDUC_GROUP_NEXT_ELEMENT, REDUC_GROUP_SIZE): Use dr_aux.dr instead
5565         of data_ref.
5566         (STMT_VINFO_DATA_REF): Likewise.  Turn into an lvalue.
5567         (STMT_VINFO_DR_INFO): New macro.
5568         (DR_VECT_AUX): Use STMT_VINFO_DR_INKFO and vect_dr_stmt.
5569         (set_dr_misalignment): Update after rename of dataref_aux.
5570         (vect_dr_stmt): Move earlier in file.  Return dr_aux.stmt.
5571         * tree-vect-stmts.c (new_stmt_vec_info): Remove redundant
5572         initialization of STMT_VINFO_DATA_REF.
5573         * tree-vectorizer.c (vec_info::move_dr): New function.
5574         * tree-vect-patterns.c (vect_recog_bool_pattern)
5575         (vect_recog_mask_conversion_pattern)
5576         (vect_recog_gather_scatter_pattern): Use it.
5577         * tree-vect-data-refs.c (vect_analyze_data_refs): Initialize
5578         the "dr" and "stmt" fields of dr_vec_info instead of
5579         STMT_VINFO_DATA_REF.
5581 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5583         * tree-vectorizer.h (_stmt_vec_info::pattern_stmt_p): New field.
5584         (is_pattern_stmt_p): Use it.
5585         * tree-vect-patterns.c (vect_init_pattern_stmt): Set pattern_stmt_p
5586         on pattern statements.
5588 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5590         * tree-vect-patterns.c (vect_mark_pattern_stmts): Take the
5591         original stmt as a stmt_vec_info rather than a gimple stmt.
5592         (vect_pattern_recog_1): Take the statement directly as a
5593         stmt_vec_info, rather than via a gimple_stmt_iterator.
5594         Update call to vect_mark_pattern_stmts.
5595         (vect_pattern_recog): Update calls accordingly.
5597 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5599         * tree-vectorizer.h (vect_get_vec_defs_for_stmt_copy)
5600         (vect_get_vec_def_for_stmt_copy): Take a vec_info rather than
5601         a vect_def_type for the first argument.
5602         * tree-vect-stmts.c (vect_get_vec_defs_for_stmt_copy): Likewise.
5603         (vect_get_vec_def_for_stmt_copy): Likewise.  Return the original
5604         operand if it isn't defined by a vectorized statement.
5605         (vect_build_gather_load_calls): Remove the mask_dt argument and
5606         update calls to vect_get_vec_def_for_stmt_copy.
5607         (vectorizable_bswap): Likewise the dt argument.
5608         (vectorizable_call): Update calls to vectorizable_bswap and
5609         vect_get_vec_def_for_stmt_copy.
5610         (vectorizable_simd_clone_call, vectorizable_assignment)
5611         (vectorizable_shift, vectorizable_operation, vectorizable_condition)
5612         (vectorizable_comparison): Update calls to
5613         vect_get_vec_def_for_stmt_copy.
5614         (vectorizable_store): Likewise.  Remove now-unnecessary calls to
5615         vect_is_simple_use.
5616         (vect_get_loop_based_defs): Remove dt argument and update call
5617         to vect_get_vec_def_for_stmt_copy.
5618         (vectorizable_conversion): Update calls to vect_get_loop_based_defs
5619         and vect_get_vec_def_for_stmt_copy.
5620         (vectorizable_load): Update calls to vect_build_gather_load_calls
5621         and vect_get_vec_def_for_stmt_copy.
5622         * tree-vect-loop.c (vect_create_epilog_for_reduction)
5623         (vectorizable_reduction, vectorizable_live_operation): Update calls
5624         to vect_get_vec_def_for_stmt_copy.
5626 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5628         * tree-vect-data-refs.c (vect_record_base_alignment): Replace vec_info
5629         and gimple stmt arguments with a stmt_vec_info.
5630         (vect_record_base_alignments): Update calls accordingly.
5631         * tree-vect-slp.c (vect_record_max_nunits): Replace vec_info
5632         and gimple stmt arguments with a stmt_vec_info.
5633         (vect_build_slp_tree_1): Remove vinfo argument and update call
5634         to vect_record_max_nunits.
5635         (vect_build_slp_tree_2): Update calls to vect_build_slp_tree_1
5636         and vect_record_max_nunits.
5638 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5640         * tree-vectorizer.h (nested_in_vect_loop_p): Move further down
5641         file and take a stmt_vec_info instead of a gimple stmt.
5642         (supportable_widening_operation, vect_finish_replace_stmt)
5643         (vect_finish_stmt_generation, vect_get_store_rhs)
5644         (vect_get_vec_def_for_operand_1, vect_get_vec_def_for_operand)
5645         (vect_get_vec_defs, vect_init_vector, vect_transform_stmt)
5646         (vect_remove_stores, vect_analyze_stmt, vectorizable_condition)
5647         (vect_get_smallest_scalar_type, vect_check_gather_scatter)
5648         (vect_create_data_ref_ptr, bump_vector_ptr)
5649         (vect_permute_store_chain, vect_setup_realignment)
5650         (vect_transform_grouped_load, vect_record_grouped_load_vectors)
5651         (vect_create_addr_base_for_vector_ref, vectorizable_live_operation)
5652         (vectorizable_reduction, vectorizable_induction)
5653         (get_initial_def_for_reduction, is_simple_and_all_uses_invariant)
5654         (vect_get_place_in_interleaving_chain): Take stmt_vec_infos rather
5655         than gimple stmts as arguments.
5656         * tree-vect-data-refs.c (vect_get_smallest_scalar_type)
5657         (vect_preserves_scalar_order_p, vect_slp_analyze_node_dependences)
5658         (can_group_stmts_p, vect_check_gather_scatter)
5659         (vect_create_addr_base_for_vector_ref, vect_create_data_ref_ptr)
5660         (bump_vector_ptr, vect_permute_store_chain, vect_setup_realignment)
5661         (vect_permute_load_chain, vect_shift_permute_load_chain)
5662         (vect_transform_grouped_load)
5663         (vect_record_grouped_load_vectors): Likewise.
5664         * tree-vect-loop.c (vect_fixup_reduc_chain)
5665         (get_initial_def_for_reduction, vect_create_epilog_for_reduction)
5666         (vectorize_fold_left_reduction, is_nonwrapping_integer_induction)
5667         (vectorizable_reduction, vectorizable_induction)
5668         (vectorizable_live_operation, vect_loop_kill_debug_uses): Likewise.
5669         * tree-vect-patterns.c (type_conversion_p, adjust_bool_stmts)
5670         (vect_get_load_store_mask): Likewise.
5671         * tree-vect-slp.c (vect_get_place_in_interleaving_chain)
5672         (vect_analyze_slp_instance, vect_mask_constant_operand_p): Likewise.
5673         * tree-vect-stmts.c (vect_mark_relevant)
5674         (is_simple_and_all_uses_invariant)
5675         (exist_non_indexing_operands_for_use_p, process_use)
5676         (vect_init_vector_1, vect_init_vector, vect_get_vec_def_for_operand_1)
5677         (vect_get_vec_def_for_operand, vect_get_vec_defs)
5678         (vect_finish_stmt_generation_1, vect_finish_replace_stmt)
5679         (vect_finish_stmt_generation, vect_truncate_gather_scatter_offset)
5680         (compare_step_with_zero, vect_get_store_rhs, get_group_load_store_type)
5681         (get_negative_load_store_type, get_load_store_type)
5682         (vect_check_load_store_mask, vect_check_store_rhs)
5683         (vect_build_gather_load_calls, vect_get_strided_load_store_ops)
5684         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5685         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5686         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5687         (get_group_alias_ptr_type, vectorizable_store, hoist_defs_of_uses)
5688         (vectorizable_load, vectorizable_condition, vectorizable_comparison)
5689         (vect_analyze_stmt, vect_transform_stmt, vect_remove_stores)
5690         (supportable_widening_operation): Likewise.
5692 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5694         * tree-vect-data-refs.c (vect_describe_gather_scatter_call): Take
5695         a stmt_vec_info instead of a gcall.
5696         (vect_check_gather_scatter): Update call accordingly.
5697         * tree-vect-loop-manip.c (iv_phi_p): Take a stmt_vec_info instead
5698         of a gphi.
5699         (vect_can_advance_ivs_p, vect_update_ivs_after_vectorizer)
5700         (slpeel_update_phi_nodes_for_loops):): Update calls accordingly.
5701         * tree-vect-loop.c (vect_transform_loop_stmt): Take a stmt_vec_info
5702         instead of a gimple stmt.
5703         (vect_transform_loop): Update calls accordingly.
5704         * tree-vect-slp.c (vect_split_slp_store_group): Take and return
5705         stmt_vec_infos instead of gimple stmts.
5706         (vect_analyze_slp_instance): Update use accordingly.
5707         * tree-vect-stmts.c (read_vector_array, write_vector_array)
5708         (vect_clobber_variable, vect_stmt_relevant_p, permute_vec_elements)
5709         (vect_use_strided_gather_scatters_p, vect_build_all_ones_mask)
5710         (vect_build_zero_merge_argument, vect_get_gather_scatter_ops)
5711         (vect_gen_widened_results_half, vect_get_loop_based_defs)
5712         (vect_create_vectorized_promotion_stmts, can_vectorize_live_stmts):
5713         Take a stmt_vec_info instead of a gimple stmt and pass stmt_vec_infos
5714         down to subroutines.
5716 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5718         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Change the type
5719         of the worklist from a vector of gimple stmts to a vector of
5720         stmt_vec_infos.
5721         * tree-vect-stmts.c (vect_mark_relevant, process_use)
5722         (vect_mark_stmts_to_be_vectorized): Likewise
5724 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5726         * tree-vect-loop.c (vect_analyze_loop_operations): Look up the
5727         statement before passing it to vect_analyze_stmt.
5728         (vect_create_epilog_for_reduction): Use a stmt_vec_info to walk
5729         the chain of phi vector definitions.  Track the exit phi via its
5730         stmt_vec_info.
5731         (vectorizable_reduction): Set cond_stmt_vinfo directly from the
5732         STMT_VINFO_REDUC_DEF.
5733         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Use
5734         stmt_vec_infos to handle the statement chains.
5735         (vect_get_slp_defs): Record the first statement in the node
5736         using a stmt_vec_info.
5737         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Look up
5738         statements here and pass their stmt_vec_info down to subroutines.
5739         (vect_init_vector_1): Hoist call to vinfo_for_stmt and pass it
5740         down to vect_finish_stmt_generation.
5741         (vect_init_vector, vect_get_vec_defs, vect_finish_replace_stmt)
5742         (vect_finish_stmt_generation): Call vinfo_for_stmt and pass
5743         stmt_vec_infos to subroutines.
5744         (vect_remove_stores): Use stmt_vec_infos to handle the statement
5745         chains.
5747 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5749         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences):
5750         (vect_check_gather_scatter, vect_create_data_ref_ptr, bump_vector_ptr)
5751         (vect_permute_store_chain, vect_setup_realignment)
5752         (vect_permute_load_chain, vect_shift_permute_load_chain)
5753         (vect_transform_grouped_load): Use stmt_vec_info rather than gimple
5754         stmts internally, and when passing values to other vectorizer routines.
5755         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
5756         * tree-vect-loop.c (vect_analyze_scalar_cycles_1)
5757         (vect_analyze_loop_operations, get_initial_def_for_reduction)
5758         (vect_create_epilog_for_reduction, vectorize_fold_left_reduction)
5759         (vectorizable_reduction, vectorizable_induction)
5760         (vectorizable_live_operation, vect_transform_loop_stmt)
5761         (vect_transform_loop): Likewise.
5762         * tree-vect-patterns.c (vect_reassociating_reduction_p)
5763         (vect_recog_widen_op_pattern, vect_recog_mixed_size_cond_pattern)
5764         (vect_recog_bool_pattern, vect_recog_gather_scatter_pattern): Likewise.
5765         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5766         (vect_slp_analyze_node_operations_1): Likewise.
5767         * tree-vect-stmts.c (vect_mark_relevant, process_use)
5768         (exist_non_indexing_operands_for_use_p, vect_init_vector_1)
5769         (vect_mark_stmts_to_be_vectorized, vect_get_vec_def_for_operand)
5770         (vect_finish_stmt_generation_1, get_group_load_store_type)
5771         (get_load_store_type, vect_build_gather_load_calls)
5772         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
5773         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
5774         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
5775         (vectorizable_store, vectorizable_load, vectorizable_condition)
5776         (vectorizable_comparison, vect_analyze_stmt, vect_transform_stmt)
5777         (supportable_widening_operation): Likewise.
5778         (vect_get_vector_types_for_stmt): Likewise.
5779         * tree-vectorizer.h (vect_dr_behavior): Likewise.
5781 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5783         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5784         (vect_slp_analyze_node_dependences, vect_analyze_data_ref_accesses)
5785         (vect_permute_store_chain, vect_permute_load_chain)
5786         (vect_shift_permute_load_chain, vect_transform_grouped_load): Avoid
5787         repeated stmt_vec_info lookups.
5788         * tree-vect-loop-manip.c (vect_can_advance_ivs_p): Likewise.
5789         (vect_update_ivs_after_vectorizer): Likewise.
5790         * tree-vect-loop.c (vect_is_simple_reduction): Likewise.
5791         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
5792         * tree-vect-patterns.c (adjust_bool_stmts): Likewise.
5793         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5794         (vect_bb_slp_scalar_cost): Likewise.
5795         * tree-vect-stmts.c (get_group_alias_ptr_type): Likewise.
5797 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5799         * tree-vect-data-refs.c (vect_check_gather_scatter): Pass the
5800         gcall rather than the generic gimple stmt to gimple_call_internal_fn.
5801         (vect_get_smallest_scalar_type, can_group_stmts_p): Use dyn_cast
5802         to get gassigns and gcalls, rather than operating on generc gimple
5803         stmts.
5804         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p)
5805         (vect_mark_stmts_to_be_vectorized, vectorizable_store)
5806         (vectorizable_load, vect_analyze_stmt): Likewise.
5807         * tree-vect-loop.c (vectorizable_reduction): Likewise gphi.
5809 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5811         * tree-vectorizer.h (get_earlier_stmt, get_later_stmt): Take and
5812         return stmt_vec_infos rather than gimple stmts.  Do not accept
5813         null arguments.
5814         (vect_find_last_scalar_stmt_in_slp): Return a stmt_vec_info instead
5815         of a gimple stmt.
5816         * tree-vect-slp.c (vect_find_last_scalar_stmt_in_slp): Likewise.
5817         Update use of get_later_stmt.
5818         (vect_get_constant_vectors): Update call accordingly.
5819         (vect_schedule_slp_instance): Likewise
5820         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Likewise.
5821         (vect_slp_analyze_instance_dependence): Likewise.
5822         (vect_preserves_scalar_order_p): Update use of get_earlier_stmt.
5824 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5826         * tree-vectorizer.h (stmt_info_for_cost::stmt): Replace with...
5827         (stmt_info_for_cost::stmt_info): ...this new field.
5828         (add_stmt_costs): Update accordingly.
5829         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
5830         (vect_get_known_peeling_cost): Likewise.
5831         (vect_estimate_min_profitable_iters): Likewise.
5832         * tree-vect-stmts.c (record_stmt_cost): Likewise.
5834 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5836         * tree-vectorizer.h (_loop_vec_info::may_misalign_stmts): Change
5837         from an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5838         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Update
5839         accordingly.
5840         * tree-vect-loop-manip.c (vect_create_cond_for_align_checks): Likewise.
5842 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5844         * tree-vectorizer.h (_stmt_vec_info::same_dr_stmt): Change from
5845         a gimple stmt to a stmt_vec_info.
5846         * tree-vect-stmts.c (vectorizable_load): Update accordingly.
5848 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5850         * tree-vectorizer.h (vec_info::grouped_stores): Change from
5851         an auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5852         (_loop_vec_info::reduction_chains): Likewise.
5853         * tree-vect-loop.c (vect_fixup_scalar_cycles_with_patterns): Update
5854         accordingly.
5855         * tree-vect-slp.c (vect_analyze_slp): Likewise.
5857 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5859         * tree-vectorizer.h (_stmt_vec_info::first_element): Change from
5860         a gimple stmt to a stmt_vec_info.
5861         (_stmt_vec_info::next_element): Likewise.
5862         * tree-vect-data-refs.c (vect_update_misalignment_for_peel)
5863         (vect_slp_analyze_and_verify_node_alignment)
5864         (vect_analyze_group_access_1, vect_analyze_group_access)
5865         (vect_small_gap_p, vect_prune_runtime_alias_test_list)
5866         (vect_create_data_ref_ptr, vect_record_grouped_load_vectors)
5867         (vect_supportable_dr_alignment): Update accordingly.
5868         * tree-vect-loop.c (vect_fixup_reduc_chain): Likewise.
5869         (vect_fixup_scalar_cycles_with_patterns, vect_is_slp_reduction)
5870         (vect_is_simple_reduction, vectorizable_reduction): Likewise.
5871         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
5872         * tree-vect-slp.c (vect_build_slp_tree_1)
5873         (vect_attempt_slp_rearrange_stmts, vect_supported_load_permutation_p)
5874         (vect_split_slp_store_group, vect_analyze_slp_instance)
5875         (vect_analyze_slp, vect_transform_slp_perm_load): Likewise.
5876         * tree-vect-stmts.c (vect_model_store_cost, vect_model_load_cost)
5877         (get_group_load_store_type, get_load_store_type)
5878         (get_group_alias_ptr_type, vectorizable_store, vectorizable_load)
5879         (vect_transform_stmt, vect_remove_stores): Likewise.
5881 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5883         * tree-vectorizer.h (vect_dr_stmt): Return a stmt_vec_info rather
5884         than a gimple stmt.
5885         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
5886         (vect_slp_analyze_data_ref_dependence, vect_record_base_alignments)
5887         (vect_calculate_target_alignmentm, vect_compute_data_ref_alignment)
5888         (vect_update_misalignment_for_peel, vect_verify_datarefs_alignment)
5889         (vector_alignment_reachable_p, vect_get_data_access_cost)
5890         (vect_get_peeling_costs_all_drs, vect_peeling_hash_get_lowest_cost)
5891         (vect_peeling_supportable, vect_enhance_data_refs_alignment)
5892         (vect_find_same_alignment_drs, vect_analyze_data_refs_alignment)
5893         (vect_analyze_group_access_1, vect_analyze_group_access)
5894         (vect_analyze_data_ref_access, vect_analyze_data_ref_accesses)
5895         (vect_vfa_access_size, vect_small_gap_p, vect_analyze_data_refs)
5896         (vect_supportable_dr_alignment): Remove vinfo_for_stmt from the
5897         result of vect_dr_stmt and use the stmt_vec_info instead of
5898         the associated gimple stmt.
5899         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
5900         (vect_gen_prolog_loop_niters): Likewise.
5901         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
5903 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5905         * tree-vectorizer.h (_slp_tree::stmts): Change from a vec<gimple *>
5906         to a vec<stmt_vec_info>.
5907         * tree-vect-slp.c (vect_free_slp_tree): Update accordingly.
5908         (vect_create_new_slp_node): Take a vec<gimple *> instead of a
5909         vec<stmt_vec_info>.
5910         (_slp_oprnd_info::def_stmts): Change from a vec<gimple *>
5911         to a vec<stmt_vec_info>.
5912         (bst_traits::value_type, bst_traits::value_type): Likewise.
5913         (bst_traits::hash): Update accordingly.
5914         (vect_get_and_check_slp_defs): Change the stmts parameter from
5915         a vec<gimple *> to a vec<stmt_vec_info>.
5916         (vect_two_operations_perm_ok_p, vect_build_slp_tree_1): Likewise.
5917         (vect_build_slp_tree): Likewise.
5918         (vect_build_slp_tree_2): Likewise.  Update uses of
5919         SLP_TREE_SCALAR_STMTS.
5920         (vect_print_slp_tree): Update uses of SLP_TREE_SCALAR_STMTS.
5921         (vect_mark_slp_stmts, vect_mark_slp_stmts_relevant)
5922         (vect_slp_rearrange_stmts, vect_attempt_slp_rearrange_stmts)
5923         (vect_supported_load_permutation_p, vect_find_last_scalar_stmt_in_slp)
5924         (vect_detect_hybrid_slp_stmts, vect_slp_analyze_node_operations_1)
5925         (vect_slp_analyze_node_operations, vect_slp_analyze_operations)
5926         (vect_bb_slp_scalar_cost, vect_slp_analyze_bb_1)
5927         (vect_get_constant_vectors, vect_get_slp_defs)
5928         (vect_transform_slp_perm_load, vect_schedule_slp_instance)
5929         (vect_remove_slp_scalar_calls, vect_schedule_slp): Likewise.
5930         (vect_analyze_slp_instance): Build up a vec of stmt_vec_infos
5931         instead of gimple stmts.
5932         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Change
5933         the stores parameter for a vec<gimple *> to a vec<stmt_vec_info>.
5934         (vect_slp_analyze_instance_dependence): Update uses of
5935         SLP_TREE_SCALAR_STMTS.
5936         (vect_slp_analyze_and_verify_node_alignment): Likewise.
5937         (vect_slp_analyze_and_verify_instance_alignment): Likewise.
5938         * tree-vect-loop.c (neutral_op_for_slp_reduction): Likewise.
5939         (get_initial_defs_for_reduction): Likewise.
5940         (vect_create_epilog_for_reduction): Likewise.
5941         (vectorize_fold_left_reduction): Likewise.
5942         * tree-vect-stmts.c (vect_prologue_cost_for_slp_op): Likewise.
5943         (vect_model_simple_cost, vectorizable_shift, vectorizable_load)
5944         (can_vectorize_live_stmts): Likewise.
5946 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5948         * tree-vectorizer.h (_loop_vec_info::reductions): Change from an
5949         auto_vec<gimple *> to an auto_vec<stmt_vec_info>.
5950         (vect_force_simple_reduction): Take and return stmt_vec_infos rather
5951         than gimple stmts.
5952         * tree-parloops.c (valid_reduction_p): Take a stmt_vec_info instead
5953         of a gimple stmt.
5954         (gather_scalar_reductions): Update after above interface changes.
5955         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Likewise.
5956         (vect_is_simple_reduction): Take and return stmt_vec_infos rather
5957         than gimple stmts.
5958         (vect_force_simple_reduction): Likewise.
5959         * tree-vect-patterns.c (vect_pattern_recog_1): Update use of
5960         LOOP_VINFO_REDUCTIONS.
5961         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
5963 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5965         * tree-vectorizer.h (_stmt_vec_info::reduc_def): Change from
5966         a gimple stmt to a stmt_vec_info.
5967         * tree-vect-loop.c (vect_active_double_reduction_p)
5968         (vect_force_simple_reduction, vectorizable_reduction): Update
5969         accordingly.
5971 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5973         * tree-vectorizer.h (_slp_tree::vec_stmts): Change from a
5974         vec<gimple *> to a vec<stmt_vec_info>.
5975         * tree-vect-loop.c (vect_create_epilog_for_reduction): Change
5976         the reduction_phis argument from a vec<gimple *> to a
5977         vec<stmt_vec_info>.
5978         (vectorizable_reduction): Likewise the phis local variable that
5979         is passed to vect_create_epilog_for_reduction.  Update for new type
5980         of SLP_TREE_VEC_STMTS.
5981         (vectorizable_induction): Update for new type of SLP_TREE_VEC_STMTS.
5982         (vectorizable_live_operation): Likewise.
5983         * tree-vect-slp.c (vect_get_slp_vect_defs): Likewise.
5984         (vect_transform_slp_perm_load, vect_schedule_slp_instance): Likewise.
5986 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
5988         * tree-vectorizer.h (_stmt_vec_info::vectorized_stmt): Change from
5989         a gimple stmt to a stmt_vec_info.
5990         (vectorizable_condition, vectorizable_live_operation)
5991         (vectorizable_reduction, vectorizable_induction): Pass back the
5992         vectorized statement as a stmt_vec_info.
5993         * tree-vect-data-refs.c (vect_record_grouped_load_vectors): Update
5994         use of STMT_VINFO_VEC_STMT.
5995         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise,
5996         accumulating the inner phis that feed the STMT_VINFO_VEC_STMT
5997         as stmt_vec_infos rather than gimple stmts.
5998         (vectorize_fold_left_reduction): Change vec_stmt from a gimple stmt
5999         to a stmt_vec_info.
6000         (vectorizable_live_operation): Likewise.
6001         (vectorizable_reduction, vectorizable_induction): Likewise,
6002         updating use of STMT_VINFO_VEC_STMT.
6003         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Update use
6004         of STMT_VINFO_VEC_STMT.
6005         (vect_build_gather_load_calls, vectorizable_bswap, vectorizable_call)
6006         (vectorizable_simd_clone_call, vectorizable_conversion)
6007         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6008         (vectorizable_store, vectorizable_load, vectorizable_condition)
6009         (vectorizable_comparison, can_vectorize_live_stmts): Change vec_stmt
6010         from a gimple stmt to a stmt_vec_info.
6011         (vect_transform_stmt): Update use of STMT_VINFO_VEC_STMT.  Pass a
6012         pointer to a stmt_vec_info to the vectorizable_* routines.
6014 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6016         * tree-vectorizer.h (_stmt_vec_info::related_stmt): Change from
6017         a gimple stmt to a stmt_vec_info.
6018         (is_pattern_stmt_p): Update accordingly.
6019         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Likewise.
6020         (vect_record_grouped_load_vectors): Likewise.
6021         * tree-vect-loop.c (vect_determine_vf_for_stmt): Likewise.
6022         (vect_fixup_reduc_chain, vect_update_vf_for_slp): Likewise.
6023         (vect_model_reduction_cost): Likewise.
6024         (vect_create_epilog_for_reduction): Likewise.
6025         (vectorizable_reduction, vectorizable_induction): Likewise.
6026         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6027         Return the stmt_vec_info for the pattern statement.
6028         (vect_set_pattern_stmt): Update use of STMT_VINFO_RELATED_STMT.
6029         (vect_split_statement, vect_mark_pattern_stmts): Likewise.
6030         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6031         (vect_detect_hybrid_slp, vect_get_slp_defs): Likewise.
6032         * tree-vect-stmts.c (vect_mark_relevant): Likewise.
6033         (vect_get_vec_def_for_operand_1, vectorizable_call): Likewise.
6034         (vectorizable_simd_clone_call, vect_analyze_stmt, new_stmt_vec_info)
6035         (free_stmt_vec_info, vect_is_simple_use): Likewise.
6037 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6039         * tree-vectorizer.h (vect_finish_replace_stmt): Return a stmt_vec_info
6040         (vect_finish_stmt_generation): Likewise.
6041         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6042         (vect_finish_replace_stmt, vect_finish_stmt_generation): Likewise.
6043         (vect_build_gather_load_calls): Use the return value of the above
6044         functions instead of a separate call to vinfo_for_stmt.  Use narrow
6045         scopes for the input gimple stmt and wider scopes for the associated
6046         stmt_vec_info.  Use vec_info::lookup_def when setting these
6047         stmt_vec_infos from an SSA_NAME definition.
6048         (vectorizable_bswap, vectorizable_call, vectorizable_simd_clone_call)
6049         (vect_create_vectorized_demotion_stmts, vectorizable_conversion)
6050         (vectorizable_assignment, vectorizable_shift, vectorizable_operation)
6051         (vectorizable_store, vectorizable_load, vectorizable_condition)
6052         (vectorizable_comparison): Likewise.
6053         * tree-vect-loop.c (vectorize_fold_left_reduction): Likewise.
6054         (vectorizable_reduction): Likewise.
6056 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6058         * tree-vectorizer.h (vect_is_simple_use): Add an optional
6059         stmt_vec_info * parameter before the optional gimple **.
6060         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6061         (process_use, vect_get_vec_def_for_operand_1): Update callers.
6062         (vect_get_vec_def_for_operand, vectorizable_shift): Likewise.
6063         * tree-vect-loop.c (vectorizable_reduction): Likewise.
6064         (vectorizable_live_operation): Likewise.
6065         * tree-vect-patterns.c (type_conversion_p): Likewise.
6066         (vect_look_through_possible_promotion): Likewise.
6067         (vect_recog_rotate_pattern): Likewise.
6068         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
6070 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6072         * tree-vectorizer.h (stmt_vec_info): Temporarily change from
6073         a typedef to a wrapper class.
6074         (NULL_STMT_VEC_INFO): New macro.
6075         (vec_info::stmt_infos): Change to vec<stmt_vec_info>.
6076         (stmt_vec_info::operator*): New function.
6077         (stmt_vec_info::operator gimple *): Likewise.
6078         (set_vinfo_for_stmt): Use NULL_STMT_VEC_INFO.
6079         (add_stmt_costs): Likewise.
6080         * tree-vect-loop-manip.c (iv_phi_p): Likewise.
6081         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost)
6082         (vect_get_known_peeling_cost): Likewise.
6083         (vect_estimate_min_profitable_iters): Likewise.
6084         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6085         * tree-vect-slp.c (vect_remove_slp_scalar_calls): Likewise.
6086         * tree-vect-stmts.c (vect_build_gather_load_calls): Likewise.
6087         (vectorizable_store, free_stmt_vec_infos): Likewise.
6088         (new_stmt_vec_info): Change return type of xcalloc to
6089         _stmt_vec_info *.
6091 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6093         * tree-vectorizer.h (vec_info::lookup_single_use): Declare.
6094         * tree-vectorizer.c (vec_info::lookup_single_use): New function.
6095         * tree-vect-loop.c (vectorizable_reduction): Use it instead of
6096         a single_imm_use-based sequence.
6097         * tree-vect-stmts.c (supportable_widening_operation): Likewise.
6099 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6101         * tree-vectorizer.h (vec_info::lookup_def): Declare.
6102         * tree-vectorizer.c (vec_info::lookup_def): New function.
6103         * tree-vect-patterns.c (vect_get_internal_def): Use it.
6104         (vect_widened_op_tree): Likewise.
6105         * tree-vect-stmts.c (vect_is_simple_use): Likewise.
6106         * tree-vect-loop.c (vect_analyze_loop_operations): Likewise.
6107         (vectorizable_reduction): Likewise.
6108         (vect_valid_reduction_input_p): Take a stmt_vec_info instead
6109         of a gimple *.
6110         (vect_is_slp_reduction): Update calls accordingly.  Use
6111         vec_info::lookup_def.
6112         (vect_is_simple_reduction): Likewise
6113         * tree-vect-slp.c (vect_detect_hybrid_slp_1): Use vec_info::lookup_def.
6115 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6117         * tree-vectorizer.h (vec_info::lookup_stmt): Declare.
6118         * tree-vectorizer.c (vec_info::lookup_stmt): New function.
6119         * tree-vect-loop.c (vect_determine_vf_for_stmt): Use it instead
6120         of vinfo_for_stmt.
6121         (vect_determine_vectorization_factor, vect_analyze_scalar_cycles_1)
6122         (vect_compute_single_scalar_iteration_cost, vect_analyze_loop_form)
6123         (vect_update_vf_for_slp, vect_analyze_loop_operations)
6124         (vect_is_slp_reduction, vectorizable_induction)
6125         (vect_transform_loop_stmt, vect_transform_loop): Likewise.
6126         * tree-vect-patterns.c (vect_init_pattern_stmt):
6127         (vect_determine_min_output_precision_1, vect_determine_precisions)
6128         (vect_pattern_recog): Likewise.
6129         * tree-vect-stmts.c (vect_analyze_stmt, vect_transform_stmt): Likewise.
6130         * config/powerpcspe/powerpcspe.c (rs6000_density_test): Likewise.
6131         * config/rs6000/rs6000.c (rs6000_density_test): Likewise.
6132         * tree-vect-slp.c (vect_detect_hybrid_slp_stmts): Likewise.
6133         (vect_detect_hybrid_slp_1, vect_detect_hybrid_slp_2)
6134         (vect_detect_hybrid_slp): Likewise.  Change the walk_stmt_info
6135         info field from a loop to a loop_vec_info.
6137 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6139         * tree-vectorizer.h (stmt_vec_info): Move typedef earlier in file.
6140         (vec_info::add_stmt): Declare.
6141         * tree-vectorizer.c (vec_info::add_stmt): New function.
6142         * tree-vect-data-refs.c (vect_create_data_ref_ptr): Use it.
6143         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Likewise.
6144         (vect_create_epilog_for_reduction, vectorizable_reduction): Likewise.
6145         (vectorizable_induction): Likewise.
6146         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Likewise.
6147         * tree-vect-stmts.c (vect_finish_stmt_generation_1): Likewise.
6148         (vectorizable_simd_clone_call, vectorizable_store): Likewise.
6149         (vectorizable_load): Likewise.
6150         * tree-vect-patterns.c (vect_init_pattern_stmt): Likewise.
6151         (vect_recog_bool_pattern, vect_recog_mask_conversion_pattern)
6152         (vect_recog_gather_scatter_pattern): Likewise.
6153         (append_pattern_def_seq): Likewise.  Remove a check that is
6154         performed by add_stmt itself.
6156 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6158         * tree-vect-loop.c (vectorizable_reduction): Fix an instance in
6159         which make_ssa_name was called with new_stmt before new_stmt
6160         had been created.
6162 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6164         * tree-vect-loop.c (vect_valid_reduction_input_p): New function,
6165         split out from...
6166         (vect_is_slp_reduction): ...here...
6167         (vect_is_simple_reduction): ...and here.  Remove repetition of tests
6168         that are already known to be false.
6170 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6172         * tree-vectorizer.h (vect_free_slp_instance): Add a final_p parameter.
6173         * tree-vect-slp.c (vect_free_slp_tree): Likewise.  Don't update
6174         STMT_VINFO_NUM_SLP_USES when it's true.
6175         (vect_free_slp_instance): Add a final_p parameter and pass it to
6176         vect_free_slp_tree.
6177         (vect_build_slp_tree_2): Update call to vect_free_slp_instance.
6178         (vect_analyze_slp_instance): Likewise.
6179         (vect_slp_analyze_operations): Likewise.
6180         (vect_slp_analyze_bb_1): Likewise.
6181         * tree-vectorizer.c (vec_info): Likewise.
6182         * tree-vect-loop.c (vect_transform_loop): Likewise.
6184 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6186         * tree-vect-loop.c (vectorizable_reduction): Assert that the
6187         function is not called for second and subsequent members of
6188         a reduction group.
6190 2018-07-31  Richard Sandiford  <richard.sandiford@arm.com>
6192         * tree-vect-loop.c (get_initial_def_for_reduction): Move special
6193         cases for nested loops from here to ...
6194         (vect_create_epilog_for_reduction): ...here.  Only call
6195         vect_is_simple_use for inner-loop reductions.
6197 2018-07-31  Martin Liska  <mliska@suse.cz>
6199         PR gcov-profile/85338
6200         PR gcov-profile/85350
6201         PR gcov-profile/85372
6202         * profile.c (struct location_triplet): New.
6203         (struct location_triplet_hash): Likewise.
6204         (output_location): Do not output a BB that
6205         is already recorded for a line.
6206         (branch_prob): Use streamed_locations.
6208 2018-07-31  Martin Liska  <mliska@suse.cz>
6210         PR gcov-profile/85370
6211         * coverage.c (coverage_begin_function): Do not mark target
6212         clones as artificial functions.
6214 2018-07-31  Martin Liska  <mliska@suse.cz>
6216         PR gcov-profile/83813
6217         PR gcov-profile/84758
6218         PR gcov-profile/85217
6219         PR gcov-profile/85332
6220         * profile.c (branch_prob): Do not record GOTO expressions
6221         for GIMPLE statements which locations are already streamed.
6223 2018-07-31  Olivier Hainque  <hainque@adacore.com>
6225         * gcc.c (handle_spec_function): Accept a soft_matched_part
6226         argument, as do_spec_1.  Pass it down to ...
6227         (eval_spec_function): Accept a soft_matched_part argument,
6228         and pass it down to ...
6229         (do_spec_2): Accept a soft_matched_part argument, and pass
6230         it down to do_spec_1.
6231         (do_spec_1): Pass soft_matched_part to handle_spec_function.
6232         (handle_braces): Update call to handle_spec_function.
6233         (driver::set_up_specs): Update calls to do_spec_2.
6234         (compare_debug_dump_opt_spec_function): Likewise.
6235         (compare_debug_self_opt_spec_function): Likewise.
6237 2018-07-31  Olivier Hainque  <hainque@adacore.com>
6239         * common.opt (nolibc): New option.
6240         * doc/invoke.texi (Link Options): Document it.
6241         * gcc.c (LINK_GCC_C_SEQUENCE_SPEC): Honor nolibc.
6242         * config/alpha/linux.h: Likewise.
6243         * config/arc/elf.h: Likewise.
6244         * config/arm/uclinux-elf.h: Likewise.
6245         * config/arm/unknown-elf.h: Likewise.
6246         * config/avr/avrlibc.h: Likewise.
6247         * config/bfin/bfin.h: Likewise.
6248         * config/bfin/linux.h: Likewise.
6249         * config/bfin/uclinux.h: Likewise.
6250         * config/darwin.h: Likewise.
6251         * config/darwin10.h: Likewise.
6252         * config/darwin12.h: Likewise.
6253         * config/gnu-user.h: Likewise.
6254         * config/lm32/uclinux-elf.h: Likewise.
6255         * config/pa/pa-hpux11.h: Likewise.
6256         * config/pa/pa64-hpux.h: Likewise.
6257         * config/sparc/sparc.h: Likewise.
6259 2018-07-31  Olivier Hainque  <hainque@adacore.com>
6261         * gcc.c (getenv_spec_function): Prepend '/' to value for allowed
6262         undefined variables.
6264 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
6266         PR target/86640
6267         * config/arm/arm.c (arm_block_set_aligned_vect): Use gen_int_mode
6268         instead of GEN_INT.
6270 2018-07-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6272         * tree-ssa-forwprop.c (simplify_builtin_call): Don't create a not NUL
6273         terminated string literal.
6275 2018-07-30  Segher Boessenkool  <segher@kernel.crashing.org>
6277         PR rtl-optimization/85160
6278         * combine.c (is_just_move): New function.
6279         (try_combine): Allow combining two instructions into two if neither of
6280         the original instructions was a move.
6282 2018-07-30  Alexander Monakov  <amonakov@ispras.ru>
6284         PR target/86673
6285         * doc/extend.texi (Global Register Variables): Discourage use of type
6286         qualifiers.
6287         (Local Register Variables): Likewise.
6289 2018-07-30  Richard Sandiford  <richard.sandiford@arm.com>
6291         PR tree-optimization/86506
6292         * hwint.h (ceil_log2): Resync with hwint.c implementation.
6294 2018-07-30  Ilya Leoshkevich  <iii@linux.ibm.com>
6296         PR target/86547
6297         * lra-constraints.c (spill_hard_reg_in_range): When selecting the
6298         hard_regno, make sure no insn between `from` and `to` clobbers it.
6300 2018-07-30  Cesar Philippidis  <cesar@codesourcery.com>
6301             Tom de Vries  <tdevries@suse.de>
6303         * config/nvptx/nvptx.c (PTX_GANG_DEFAULT): Rename to ...
6304         (PTX_DEFAULT_RUNTIME_DIM): ... this.
6305         (nvptx_goacc_validate_dims): Set default worker and gang dims to
6306         PTX_DEFAULT_RUNTIME_DIM.
6307         (nvptx_dim_limit): Ignore GOMP_DIM_WORKER.
6309 2018-07-29  John David Anglin  <danglin@gcc.gnu.org>
6311         * config/pa/pa.c (pa_output_addr_vec): Align address table.
6312         * config/pa/pa.h (JUMP_TABLES_IN_TEXT_SECTION): Revise comment.
6313         * config/pa/pa32-linux.h (JUMP_TABLES_IN_TEXT_SECTION): Define.
6315 2018-07-27  Michael Meissner  <meissner@linux.ibm.com>
6317         * config/rs6000/constraints.md (wG constraint): Delete, no longer
6318         used.
6319         * config/rs6000/predicates.md (p9_fusion_reg_operand): Rename
6320         predicate to reflect toc fusion has been deleted.
6321         (toc_fusion_mem_raw): Delete, no longer used.
6322         (toc_fusion_mem_wrapped): Likewise.
6323         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete toc
6324         fusion mask bit.
6325         * config/rs6000/rs6000-protos.h (fusion_wrap_memory_address):
6326         Delete, no longer used.
6327         * config/rs6000/rs6000.c (struct rs6000_reg_addr): Delete fields
6328         meant to be used for toc fusion.
6329         (rs6000_debug_print_mode): Delete toc fusion debugging.
6330         (rs6000_debug_reg_global): Likewise.
6331         (rs6000_init_hard_regno_mode_ok): Delete setting up fields for toc
6332         fusion and secondary reload support that were never used.
6333         (rs6000_option_override_internal): Delete TOC fusion, that was only
6334         partially defined, and it did not work unless you also used the
6335         -mcmodel= switch.
6336         (rs6000_legitimate_address_p): Delete TOC fusion support.
6337         (rs6000_opt_masks): Likewise.
6338         (fusion_wrap_memory_address): Delete function, no longer used.
6339         (fusion_split_address); Delete TOC fusion support.
6340         * config/rs6000/rs6000.h (TARGET_TOC_FUSION_INT): Delete, no
6341         longer used with toc fusion being deleted.
6342         (TARGET_TOC_FUSION_FP): Likewise.
6343         * config/rs6000/rs6000.md (UNSPEC_FUSION_ADDIS): Delete TOC fusion
6344         UNSPEC.
6345         (toc fusion spliter): Delete TOC fusion support.
6346         (toc_fusionload_<mode>): Likewise.
6347         (toc_fusionload_di): Likewise.
6348         (fusion_gpr_load_<mode>): Delete generator function, this insn no
6349         longer needs to be named.  Rename predicate to delete TOC fusion.
6350         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
6351         (fusion_gpr_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
6352         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_load): Likewise.
6353         (fusion_vsx_<P:mode>_<GPR_FUSION:mode>_store): Likewise.
6354         (p9 fusion peephole2s): Rename predicate to delete TOC fusion.
6356 2018-07-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6358         * doc/extend.texi (Basic PowerPC Built-in Functions Available on
6359         ISA 2.05):  Replace __uint128_t with __uint128 and __int128_t with
6360         __int128 in built-in function prototypes.
6361         (PowerPC AltiVec Built-in Functions on ISA 2.07): Likewise.
6362         (PowerPC AltiVec Built-in Functions on ISA 3.0): Likewise.
6364 2018-07-27  Martin Sebor  <msebor@redhat.com>
6366         PR tree-optimization/86696
6367         * tree-ssa-strlen.c (get_min_string_length): Handle all integer
6368         types, including enums.
6369         (handle_char_store): Be prepared for the above function to fail.
6371 2018-07-26  Qing Zhao  <qing.zhao@oracle.com>
6373         * builtins.c (inline_expand_builtin_string_cmp): Disable inlining
6374         when optimization level is lower than 2 or optimize for size.
6376 2018-07-26  Martin Sebor  <msebor@redhat.com>
6378         PR tree-optimization/86043
6379         PR tree-optimization/86042
6380         * tree-ssa-strlen.c (handle_builtin_memcpy): Handle strict overlaps.
6381         (get_string_cst_length): Rename...
6382         (get_min_string_length): ...to this.  Add argument.
6383         (handle_char_store): Extend to handle multi-character stores by
6384         MEM_REF.
6385         * tree.c (initializer_zerop): Use new argument.  Handle MEM_REF.
6386         * tree.h (initializer_zerop): Add argument.
6388 2018-07-26  Jakub Jelinek  <jakub@redhat.com>
6390         PR middle-end/86660
6391         * omp-low.c (scan_sharing_clauses): Don't ignore map clauses for
6392         declare target to variables if they have always,{to,from,tofrom} map
6393         kinds.
6395 2018-07-26  Martin Liska  <mliska@suse.cz>
6397         PR lto/86548
6398         * lto-wrapper.c: Add linker_output as prefix
6399         for ltrans_output_file.
6401 2018-07-26  Segher Boessenkool  <segher@kernel.crashing.org>
6403         PR rtl-optimization/85805
6404         * combine.c (reg_nonzero_bits_for_combine): Only use the last set
6405         value for hard registers if that was written in the same mode.
6407 2018-07-26  Martin Liska  <mliska@suse.cz>
6409         PR gcov-profile/86536
6410         * gcov.c (format_gcov): Use printf format %.*f directly
6411         and do not handle special values.
6413 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
6415         * common/config/arc/arc-common.c (arc_option_optimization_table):
6416         Update default optimizations for size.
6418 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
6420         * config/arc/arc.md (movsf_insn): Add short instruction selection.
6421         * config/arc/constraints.md (CfZ): New constraint.
6422         * config/arc/fpu.md (addssf3_fpu): Use CfZ constraint.
6423         (subsf3_fpu): Likewise.
6424         (cmpsf_fpu): Likewise.
6425         (cmpsf_fpu_uneq): Likewise.
6427 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
6429         * config/arc/arc.c (compact_memory_operand_p): Check for uncached
6430         accesses as well.
6431         (arc_is_uncached_mem_p): uncached applies to both the variable and
6432         the pointer.
6434 2018-07-25  Claudiu Zissulescu  <claziss@synopsys.com>
6436         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Add additional
6437         register names.
6439 2018-07-25  David Malcolm  <dmalcolm@redhat.com>
6441         * optinfo-emit-json.cc (class optrecord_json_writer): Convert
6442         field "m_scopes" from vec to auto_vec.
6444 2018-07-25  Martin Liska  <mliska@suse.cz>
6446         * config/powerpcspe/powerpcspe-protos.h (rs6000_loop_align): Fix
6447         return type.
6449 2018-07-25  Richard Biener  <rguenther@suse.de>
6451         PR debug/86654
6452         * dwarf2out.c (dwarf2out_decl): Do not handle nested functions
6453         special wrt context_die late.
6454         (gen_subprogram_die): Re-use DIEs in local scope.
6456 2018-07-25  Richard Sandiford  <richard.sandiford@arm.com>
6458         PR tree-optimization/86644
6459         * hwint.c (ceil_log2): Fix comment.  Return 0 for 0.
6461 2018-07-25  Martin Liska  <mliska@suse.cz>
6463         PR middle-end/86645
6464         * dumpfile.c: And excluded values with TDF_ALL_VALUES.
6465         * dumpfile.h (enum dump_flag): Defince TDF_ALL_VALUES.
6467 2018-07-25  Martin Liska  <mliska@suse.cz>
6469         PR sanitizer/79635
6470         * params.def: Explain ASan abbreviation and provide
6471         a documentation link.
6473 2018-07-24  Martin Sebor  <msebor@redhat.com>
6475         PR tree-optimization/86622
6476         PR tree-optimization/86532
6477         * builtins.h (string_length): Declare.
6478         * builtins.c (c_strlen): Correct handling of non-constant offsets.
6479         (check_access): Be prepared for non-constant length ranges.
6480         (string_length): Make extern.
6481         * expr.c (string_constant): Only handle the minor non-constant
6482         array index.  Use string_constant to compute the length of
6483         a generic string constant.
6485 2018-07-24  Richard Sandiford  <richard.sandiford@arm.com>
6487         PR tree-optimization/86618
6488         * tree-vect-stmts.c (vectorizable_call): Don't take the address
6489         of LOOP_VINFO_MASKS (loop_vinfo) when loop_vinfo is null.
6491 2018-07-24  David Malcolm  <dmalcolm@redhat.com>
6493         PR tree-optimization/86636
6494         * json.cc (json::object::set): Fix comment.  Add assertions.
6495         (json::array::append): Move here from json.h.  Add comment and an
6496         assertion.
6497         (json::string::string): Likewise.
6498         * json.h (json::array::append): Move to json.cc.
6499         (json::string::string): Likewise.
6500         * optinfo-emit-json.cc
6501         (optrecord_json_writer::impl_location_to_json): Assert that we
6502         aren't attempting to write out UNKNOWN_LOCATION, or an ad-hoc
6503         wrapper around it.  Expand the location once, rather than three
6504         times.
6505         (optrecord_json_writer::inlining_chain_to_json): Fix the check for
6506         UNKNOWN_LOCATION, to use LOCATION_LOCUS to look through ad-hoc
6507         wrappers.
6508         (optrecord_json_writer::optinfo_to_json): Likewise, in four
6509         places.  Fix some overlong lines.
6511 2018-07-24  Matthew Malcomson  <matthew.malcomson@arm.com>
6513         * config/aarch64/aarch64-simd.md
6514         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>): Split into...
6515         (aarch64_<ANY_EXTEND:su>subw<mode>): ... This...
6516         (aarch64_<ANY_EXTEND:su>addw<mode>): ... And this.
6517         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w<mode>_internal): Split into...
6518         (aarch64_<ANY_EXTEND:su>subw<mode>_internal): ... This...
6519         (aarch64_<ANY_EXTEND:su>addw<mode>_internal): ... And this.
6520         (aarch64_<ANY_EXTEND:su><ADDSUB:optab>w2<mode>_internal): Split into...
6521         (aarch64_<ANY_EXTEND:su>subw2<mode>_internal): ... This...
6522         (aarch64_<ANY_EXTEND:su>addw2<mode>_internal): ... And this.
6524 2018-07-24  Jakub Jelinek  <jakub@redhat.com>
6526         PR middle-end/86627
6527         * expmed.c (expand_divmod): Punt if d == HOST_WIDE_INT_MIN
6528         and size > HOST_BITS_PER_WIDE_INT.  For size > HOST_BITS_PER_WIDE_INT
6529         and abs_d == d, do the power of two handling if profitable.
6531 2018-07-24  Richard Biener  <rguenther@suse.de>
6533         * match.pd: Add BIT_FIELD_REF canonicalizations.
6535 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6537         PR c/86617
6538         * genmatch.c (dt_operand::gen_match_op): Avoid folding volatile values.
6540 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6542         * gimple-fold.c (gimple_fold_builtin_printf): Don't create a not NUL
6543         terminated STRING_CST object.
6545 2018-07-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
6547         hsa-dump.c (dump_hsa_symbol): Avoid out of scope access to buf.
6549 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
6551         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Adjust.
6552         * config/rs6000/rs6000-protos.h (rs6000_split_v4si_init): Delete.
6553         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Always force
6554         the elements into a register.
6555         (rs6000_split_v4si_init_di_reg): Delete.
6556         (rs6000_split_v4si_init): Delete.
6557         * config/rs6000/vsx.md (unspec): Delete UNSPEC_VSX_VEC_INIT.
6558         (vsx_init_v4si): Rewrite as a define_expand.
6560 2018-07-23  Segher Boessenkool  <segher@kernel.crashing.org>
6562         * config/rs6000/rs6000.md (splitters for rldimi and rlwimi with the
6563         zero_extend argument from memory): New.
6565 2018-07-22  Martin Sebor  <msebor@redhat.com>
6567         PR bootstrap/86621
6568         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Avoid
6569         diagnosing calls with unknown arguments unless -Walloca-larger-than
6570         is restricted to less than PTRDIFF_MAX bytes.
6572 2018-07-22  Gerald Pfeifer  <gerald@pfeifer.com>
6574         * doc/gcov.texi (Invoking Gcov): Editorial changes.
6576 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
6578         * pretty-print.c (text_info::set_location): Remove redundant
6579         "line_table" parameter from call to rich_location::set_range.
6581 2018-07-20  Martin Sebor  <msebor@redhat.com>
6583         PR middle-end/82063
6584         * builtins.c (expand_builtin_alloca): Adjust.
6585         * calls.c (alloc_max_size): Simplify.
6586         * cgraphunit.c (cgraph_node::expand): Adjust.
6587         * common.opt (larger_than_size, warn_frame_larger_than): Remove
6588         variables.
6589         (frame_larger_than_size): Same.
6590         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Change options
6591         to take a HOST_WIDE_INT argument and accept a byte-size suffix.
6592         Initialize.
6593         * doc/invoke.texi (GCC Command Options): Document option arguments.
6594         Explain byte-size arguments and suffixes.
6595         (-Wvla-larger-than, -Wno-alloc-size-larger-than): Update.
6596         (-Wno-alloca-larger-than, -Wno-vla-larger-than): Same.
6597         (-Wframe-larger-than, -Wlarger-than, -Wstack-usage): Same.
6598         * doc/options.texi (UInteger): Expand.
6599         (Host_Wide_Int, ByteSize): Document new properties.
6600         * final.c (final_start_function_1): Include sizes in an error message.
6601         * function.c (frame_offset_overflow): Same.
6602         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Adjust.
6603         (alloca_call_type_by_arg): Change function argument to HOST_WIDE_INT.
6604         Diagnose unbounded alloca calls only for limits of less than
6605         PTRDIFF_MAX.
6606         (alloca_call_type): Adjust.  Diagnose possibly out-of-bounds alloca
6607         calls and VLA size only for limits of less than PTRDIFF_MAX.  Same
6608         for alloca(0).
6609         (pass_walloca::execute): Adjust.  Diagnose alloca calls in loops
6610         only for limits of less than PTRDIFF_MAX.
6611         * langhooks-def.h (lhd_handle_option): Change function argument
6612         to HOST_WIDE_INT.
6613         * langhooks.c (lhd_handle_option): Same.
6614         * langhooks.h (handle_option): Same.
6615         * opt-functions.awk (switch_bit_fields): Handle Host_Wide_Int and
6616         ByteSize flags.
6617         (var_type, var_type_struct): Same.
6618         (var_set): Handle ByteSize flag.
6619         * optc-gen.awk: Add comments to output to ease debugging.  Make
6620         use of HOST_WIDE_INT where appropriate.
6621         * opts-gen-save.awk:  Use %lx to format unsigned long.
6622         * opth-gen.awk: Change function argument to HOST_WIDE_INT.
6623         * opts-common.c (integral_argument): Return HOST_WIDE_INT and add
6624         arguments.  Parse bytes-size suffixes.
6625         (enum_arg_to_value): Change function argument to HOST_WIDE_INT.
6626         (enum_value_to_arg): Same.
6627         (decode_cmdline_option): Handle cl_host_wide_int.  Adjust.
6628         (handle_option): Adjust.
6629         (generate_option): Change function argument to HOST_WIDE_INT.
6630         (cmdline_handle_error): Adjust.
6631         (read_cmdline_option): Change function argument to HOST_WIDE_INT.
6632         (set_option): Change function argument to HOST_WIDE_INT.
6633         (option_enabled): Handle cl_host_wide_int.
6634         (get_option_state): Handle CLVC_SIZE.
6635         (control_warning_option): Same.
6636         * opts.c (common_handle_option): Change function argument to
6637         HOST_WIDE_INT.  Remove handling of OPT_Walloca_larger_than_ and
6638         OPT_Wvla_larger_than_.
6639         * opts.h (enum cl_var_type): Add an enumerator.
6640         * stor-layout.c (layout_decl): Print a more meaningful warning.
6641         * toplev.c (output_stack_usage): Adjust.
6643 2018-07-20  Qing Zhao  <qing.zhao@oracle.com>
6645         * builtins.c (expand_builtin_memcmp): Delete the last parameter for
6646         call to inline_expand_builtin_string_cmp.
6647         (expand_builtin_strcmp): Likewise.
6648         (expand_builtin_strncmp): Likewise.
6649         (inline_string_cmp): Delete the last parameter, change char_type_node
6650         to unsigned_char_type_node for strcmp/strncmp, add conversions to the
6651         two operands.
6652         (inline_expand_builtin_string_cmp): Delete the last parameter, give up
6653         the inlining expansion on target where the type of the call has same or
6654         narrower precision than unsigned char.
6656 2018-07-20  David Malcolm  <dmalcolm@redhat.com>
6658         * Makefile.in (OBJS): Add json.o and optinfo-emit-json.o.
6659         (CFLAGS-optinfo-emit-json.o): Define TARGET_NAME.
6660         * common.opt (fsave-optimization-record): New option.
6661         * coretypes.h (struct kv_pair): Move here from dumpfile.c.
6662         * doc/invoke.texi (-fsave-optimization-record): New option.
6663         * dumpfile.c: Include "optinfo-emit-json.h".
6664         (struct kv_pair): Move to coretypes.h.
6665         (optgroup_options): Make non-static.
6666         (dump_context::end_scope): Call
6667         optimization_records_maybe_pop_dump_scope.
6668         * dumpfile.h (optgroup_options): New decl.
6669         * json.cc: New file.
6670         * json.h: New file.
6671         * optinfo-emit-json.cc: New file.
6672         * optinfo-emit-json.h: New file.
6673         * optinfo.cc: Include "optinfo-emit-json.h".
6674         (optinfo::emit): Call optimization_records_maybe_record_optinfo.
6675         (optinfo_enabled_p): Check optimization_records_enabled_p.
6676         (optinfo_wants_inlining_info_p): Likewise.
6677         * optinfo.h: Update comment.
6678         * profile-count.c (profile_quality_as_string): New function.
6679         * profile-count.h (profile_quality_as_string): New decl.
6680         (profile_count::quality): New accessor.
6681         * selftest-run-tests.c (selftest::run_tests): Call json_cc_tests
6682         and optinfo_emit_json_cc_tests.
6683         * selftest.h (selftest::json_cc_tests): New decl.
6684         (selftest::optinfo_emit_json_cc_tests): New decl.
6685         * toplev.c: Include "optinfo-emit-json.h".
6686         (compile_file): Call optimization_records_finish.
6687         (do_compile): Call optimization_records_start.
6688         * tree-ssa-live.c: Include optinfo.h.
6689         (remove_unused_scope_block_p): Retain inlining information if
6690         optinfo_wants_inlining_info_p returns true.
6692 2018-07-20  Richard Biener  <rguenther@suse.de>
6694         PR debug/86585
6695         * dwarf2out.c (dwarf2out_die_ref_for_decl): Test in_lto_p
6696         to cover -flto-partition=none.
6698 2018-07-20  Martin Liska  <mliska@suse.cz>
6700         * tree.h (DECL_LOCATION_RANGE): Remove unused macro.
6701         (get_decl_source_range): Remove unused function.
6703 2018-07-20  Richard Biener  <rguenther@suse.de>
6705         * tree-ssa-sccvn.h (struct vn_nary_op_s): Add next member.
6706         (struct vn_phi_s): Likewise.
6707         (struct vn_reference_s): Likewise.
6708         * tree-ssa-sccvn.c (vn_nary_op_hasher::equal): Add shortcut
6709         for searching the slot of an entry known to be in the hash itself.
6710         (vn_phi_hasher::equal): Likewise.
6711         (vn_reference_hasher::equal): Likewise.
6712         (last_inserted_ref, last_inserted_phi, last_inserted_nary): New
6713         globals.
6714         (optimistic_info, current_info): Remove, keeping only valid_info.
6715         (vn_reference_lookup_1): Remove fallback lookup.
6716         (vn_reference_lookup_2): Likewise.
6717         (vn_nary_op_lookup_1): Likewise.
6718         (vn_phi_lookup): Likewise.
6719         (vn_nary_build_or_lookup_1): Make sure to not chain the built
6720         hash element.
6721         (vn_reference_insert): Adjust, chain the inserted hash element
6722         at last_inserted_ref.
6723         (vn_reference_insert_pieces): Likewise.
6724         (visit_reference_op_call): Likewise.
6725         (vn_nary_op_insert_into): Chain the inserted hash element at
6726         last_inserted_nary.
6727         (vn_nary_op_insert_pieces): Adjust.
6728         (vn_nary_op_insert): Likewise.
6729         (vn_nary_op_insert_stmt): Likewise.
6730         (vn_phi_insert): Adjust, chain the inserted hash element at
6731         last_inserted_phi.
6732         (process_scc): Remove clearing and copying the optimistic
6733         table.  Instead remove elements inserted during an optimistic
6734         iteration from the single table we maintain.
6735         (init_scc_vn): Adjust.
6736         (free_scc_vn): Likewise.
6737         (sccvn_dom_walker::record_cond): Likewise.
6738         (sccvn_dom_walker::after_dom_children): Likewise.
6740 2018-07-19  Martin Sebor  <msebor@redhat.com>
6742         PR tree-optimization/84047
6743         PR tree-optimization/83776
6744         * tree-vrp.c (vrp_prop::check_mem_ref): New function.
6745         (check_array_bounds): Call it.
6747 2018-07-19  Martin Sebor  <msebor@redhat.com>
6749         * align.h (align_flags): Use member initialization.
6751 2018-07-19  David Malcolm  <dmalcolm@redhat.com>
6753         * Makefile.in (OBJS): Add optinfo.o.
6754         * coretypes.h (class symtab_node): New forward decl.
6755         (struct cgraph_node): New forward decl.
6756         (class varpool_node): New forward decl.
6757         * dump-context.h: New file.
6758         * dumpfile.c: Include "optinfo.h", "dump-context.h", "cgraph.h",
6759         "tree-pass.h".
6760         (refresh_dumps_are_enabled): Use optinfo_enabled_p.
6761         (set_dump_file): Call dumpfile_ensure_any_optinfo_are_flushed.
6762         (set_alt_dump_file): Likewise.
6763         (dump_context::~dump_context): New dtor.
6764         (dump_gimple_stmt): Move implementation to...
6765         (dump_context::dump_gimple_stmt): ...this new member function.
6766         Add the stmt to any pending optinfo, creating one if need be.
6767         (dump_gimple_stmt_loc): Move implementation to...
6768         (dump_context::dump_gimple_stmt_loc): ...this new member function.
6769         Start a new optinfo and add the stmt to it.
6770         (dump_gimple_expr): Move implementation to...
6771         (dump_context::dump_gimple_expr): ...this new member function.
6772         Add the stmt to any pending optinfo, creating one if need be.
6773         (dump_gimple_expr_loc): Move implementation to...
6774         (dump_context::dump_gimple_expr_loc): ...this new member function.
6775         Start a new optinfo and add the stmt to it.
6776         (dump_generic_expr): Move implementation to...
6777         (dump_context::dump_generic_expr): ...this new member function.
6778         Add the tree to any pending optinfo, creating one if need be.
6779         (dump_generic_expr_loc): Move implementation to...
6780         (dump_context::dump_generic_expr_loc): ...this new member
6781         function.  Add the tree to any pending optinfo, creating one if
6782         need be.
6783         (dump_printf): Move implementation to...
6784         (dump_context::dump_printf_va): ...this new member function.  Add
6785         the text to any pending optinfo, creating one if need be.
6786         (dump_printf_loc): Move implementation to...
6787         (dump_context::dump_printf_loc_va): ...this new member function.
6788         Start a new optinfo and add the stmt to it.
6789         (dump_dec): Move implementation to...
6790         (dump_context::dump_dec): ...this new member function.  Add the
6791         value to any pending optinfo, creating one if need be.
6792         (dump_context::dump_symtab_node): New member function.
6793         (dump_context::get_scope_depth): New member function.
6794         (dump_context::begin_scope): New member function.
6795         (dump_context::end_scope): New member function.
6796         (dump_context::ensure_pending_optinfo): New member function.
6797         (dump_context::begin_next_optinfo): New member function.
6798         (dump_context::end_any_optinfo): New member function.
6799         (dump_context::s_current): New global.
6800         (dump_context::s_default): New global.
6801         (dump_scope_depth): Delete global.
6802         (dumpfile_ensure_any_optinfo_are_flushed): New function.
6803         (dump_symtab_node): New function.
6804         (get_dump_scope_depth): Reimplement in terms of dump_context.
6805         (dump_begin_scope): Likewise.
6806         (dump_end_scope): Likewise.
6807         (selftest::temp_dump_context::temp_dump_context): New ctor.
6808         (selftest::temp_dump_context::~temp_dump_context): New dtor.
6809         (selftest::verify_item): New function.
6810         (ASSERT_IS_TEXT): New macro.
6811         (ASSERT_IS_TREE): New macro.
6812         (ASSERT_IS_GIMPLE): New macro.
6813         (selftest::test_capture_of_dump_calls): New test.
6814         (selftest::dumpfile_c_tests): Call it.
6815         * dumpfile.h (dump_printf, dump_printf_loc, dump_basic_block)
6816         (dump_generic_expr_loc, dump_generic_expr, dump_gimple_stmt_loc)
6817         (dump_gimple_stmt, dump_dec): Gather these related decls and add a
6818         descriptive comment.
6819         (dump_function, print_combine_total_stats, enable_rtl_dump_file)
6820         (dump_node, dump_bb): Move these unrelated decls.
6821         (class dump_manager): Add leading comment.
6822         * optinfo.cc: New file.
6823         * optinfo.h: New file.
6825 2018-07-19  Michael Collison  <michael.collison@arm.com>
6826             Richard Henderson <rth@redhat.com>
6828         * config/aarch64/aarch64.md (subv<GPI>4, usubv<GPI>4): New patterns.
6829         (subti): Handle op1 zero.
6830         (subvti4, usub4ti4): New.
6831         (*sub<GPI>3_compare1_imm): New.
6832         (sub<GPI>3_carryinCV): New.
6833         (*sub<GPI>3_carryinCV_z1_z2, *sub<GPI>3_carryinCV_z1): New.
6834         (*sub<GPI>3_carryinCV_z2, *sub<GPI>3_carryinCV): New.
6836 2018-07-19  Michael Collison  <michael.collison@arm.com>
6837             Richard Henderson <rth@redhat.com>
6839         * config/aarch64/aarch64.md: (addv<GPI>4, uaddv<GPI>4): New.
6840         (addti3): Create simpler code if low part is already known to be 0.
6841         (addvti4, uaddvti4): New.
6842         (*add<GPI>3_compareC_cconly_imm): New.
6843         (*add<GPI>3_compareC_cconly): New.
6844         (*add<GPI>3_compareC_imm): New.
6845         (*add<GPI>3_compareC): Rename from add<GPI>3_compare1; do not
6846         handle constants within this pattern..
6847         (*add<GPI>3_compareV_cconly_imm): New.
6848         (*add<GPI>3_compareV_cconly): New.
6849         (*add<GPI>3_compareV_imm): New.
6850         (add<GPI>3_compareV): New.
6851         (add<GPI>3_carryinC, add<GPI>3_carryinV): New.
6852         (*add<GPI>3_carryinC_zero, *add<GPI>3_carryinV_zero): New.
6853         (*add<GPI>3_carryinC, *add<GPI>3_carryinV): New.
6854         ((*add<GPI>3_compareC_cconly_imm): Replace 'ne' operator
6855         with 'comparison' operator.
6856         (*add<GPI>3_compareV_cconly_imm): Ditto.
6857         (*add<GPI>3_compareV_cconly): Ditto.
6858         (*add<GPI>3_compareV_imm): Ditto.
6859         (add<GPI>3_compareV): Ditto.
6860         (add<mode>3_carryinC): Ditto.
6861         (*add<mode>3_carryinC_zero): Ditto.
6862         (*add<mode>3_carryinC): Ditto.
6863         (add<mode>3_carryinV): Ditto.
6864         (*add<mode>3_carryinV_zero): Ditto.
6865         (*add<mode>3_carryinV): Ditto.
6867 2018-07-19  Michael Collison  <michael.collison@arm.com>
6868             Richard Henderson <rth@redhat.com>
6870         * config/aarch64/aarch64-modes.def (CC_V): New.
6871         * config/aarch64/aarch64-protos.h
6872         (aarch64_addti_scratch_regs): Declare
6873         (aarch64_subvti_scratch_regs): Declare.
6874         (aarch64_expand_subvti): Declare.
6875         (aarch64_gen_unlikely_cbranch): Declare
6876         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Test
6877         for signed overflow using CC_Vmode.
6878         (aarch64_get_condition_code_1): Handle CC_Vmode.
6879         (aarch64_gen_unlikely_cbranch): New function.
6880         (aarch64_addti_scratch_regs): New function.
6881         (aarch64_subvti_scratch_regs): New function.
6882         (aarch64_expand_subvti): New function.
6884 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
6886         * config/aarch64/aarch64-option-extensions.def: New entry for profile
6887         extension.
6888         * config/aarch64/aarch64.h (AARCH64_FL_PROFILE): New.
6889         * doc/invoke.texi (aarch64-feature-modifiers): New entry for profile
6890         extension.
6892 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
6894         PR target/83009
6895         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
6896         address check not strict.
6898 2018-07-19  Andre Vieira  <andre.simoesdiasvieira@arm.com>
6900         * config/aarch64/aarch64-simd.md (aarch64_simd_mov<VQ:mode>): Replace
6901         Umq with Umn.
6902         (store_pair_lanes<mode>): Likewise.
6903         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): Add new
6904         enum value 'ADDR_QUERY_LDP_STP_N'.
6905         * config/aarch64/aarch64.c (aarch64_addr_query_type): Likewise.
6906         (aarch64_print_address_internal): Add declaration.
6907         (aarch64_print_ldpstp_address): Remove.
6908         (aarch64_classify_address): Adapt mode for 'ADDR_QUERY_LDP_STP_N'.
6909         (aarch64_print_operand): Change printing of 'y'.
6910         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Use
6911         new enum value 'ADDR_QUERY_LDP_STP_N', don't hardcode mode and use
6912         'true' rather than '1'.
6913         * config/aarch64/constraints.md (Uml): Likewise.
6914         (Uml): Rename to Umn.
6915         (Umq): Remove.
6917 2018-07-19  Richard Biener  <rguenther@suse.de>
6919         * tree-ssa-sccvn.h (struct vn_phi_s): Make phiargs member
6920         a trailing array.
6921         * tree-ssa-sccvn.c: Remove alloc-pool.h use.
6922         (vn_phi_hasher): Derive from nofree_ptr_hash and remove remove method.
6923         (vn_reference_hasher): Likewise.
6924         (struct vn_tables_s): Remove obstack and alloc-pool members.
6925         (vn_tables_obstack, vn_tables_insert_obstack): New global obstacks.
6926         (vn_nary_build_or_lookup_1): Manually build in vn_tables_insert_obstack.
6927         (vn_reference_insert): Allocate from obstack instead of from alloc-pool.
6928         (vn_reference_insert_pieces): Likewise.
6929         (alloc_vn_nary_op_noinit): Adjust.
6930         (vn_nary_op_insert_stmt): Allocate phiargs in-place.
6931         (vn_phi_eq): Adjust.
6932         (shared_lookup_phiargs): Remove.
6933         (vn_phi_lookup): Allocate temporary vn_phi_s on the stack.
6934         (vn_phi_insert): Allocate from obstack instead of from alloc-pool.
6935         (visit_reference_op_call): Likewise.
6936         (copy_nary, copy_phi, copy_reference): Remove.
6937         (process_scc): Rewind the obstack when iterating.  Do not
6938         copy the elements to valid_info but just move them from one
6939         hashtable to the other.
6940         (allocate_vn_table): Adjust.
6941         (free_vn_table): Likewise.
6942         (init_scc_vn): Likewise.
6943         (free_scc_vn): Likewise.
6945 2018-07-19  H.J. Lu  <hongjiu.lu@intel.com>
6947         PR target/86560
6948         * config/i386/i386.c (rest_of_insert_endbranch): Lookup
6949         indirect_return as function type attribute.
6950         (ix86_attribute_table): Change indirect_return to function
6951         type attribute.
6952         * doc/extend.texi: Update indirect_return attribute.
6954 2018-07-19  Aldy Hernandez  <aldyh@redhat.com>
6956         * wide-int.h (widest2_int): New.
6957         * gimple-fold.c (arith_overflowed_p): Use it.
6958         * tree.h (widest2_int_cst): New.
6959         * tree-vrp.c (wide_int_binop_overflow): Rename from
6960         vrp_int_const_binop.
6961         Rewrite to work on trees.
6962         (extract_range_from_multiplicative_op_1): Abstract code to...
6963         (wide_int_range_min_max): ...here.
6964         (wide_int_range_cross_product): ...and here.
6965         (extract_range_from_binary_expr_1): Abstract overflow code to...
6966         (wide_int_range_mult_wrapping): ...here.
6967         * tree-vrp.h (wide_int_range_cross_product): New.
6968         (wide_int_range_mult_wrapping): New.
6970 2018-07-19  Andrew Senkevich  <andrew.senkevich@intel.com>
6971             Julia Koval  <julia.koval@intel.com>
6973         * config/i386/x86-tune-costs.h (skylake_memcpy,
6974         skylake_memset): Replace rep_prefix with unrolling for size 512.
6976 2018-07-18  Kugan Vivekanandarajah  <kuganv@linaro.org>
6978         PR middle-end/86544
6979         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): Handle
6980         comparision with EQ_EXPR in last stmt.
6982 2018-07-18  Kelvin Nilsen  <kelvin@gcc.gnu.org>
6984         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Rename
6985         this subsection to "PowerPC AltiVec/VSX Built-in Functions".
6986         (PowerPC AltiVec/VSX Built-in Functions): New name for subsection
6987         previously known as "PowerPC AltiVec Built-in Functions".  Move
6988         some material to new subsubsections "PowerPC AltiVec Built-in
6989         Functions on ISA 2.06" and "PowerPC AltiVec Built-in Functions on
6990         ISA 2.07".
6991         (PowerPC Altivec Built-in Functions on ISA 2.05): New subsubsection.
6992         (PowerPC Altivec Built-in Functions on ISA 2.06): Likewise.
6993         (PowerPC Altivec Built-in Functions on ISA 2.07): Likewise.
6994         (PowerPC Altivec Built-in Functions on ISA 3.0): Likewise.
6996 2018-07-18  Richard Biener  <rguenther@suse.de>
6998         PR tree-optimization/86557
6999         * tree-vect-patterns.c (vect_recog_divmod_pattern): Also handle
7000         EXACT_DIV_EXPR.
7002 2018-07-18  Ilya Leoshkevich  <iii@linux.ibm.com>
7004         * config/s390/s390.c (s390_function_profiler): Generate CFI.
7006 2018-07-17  Jeff Law  <law@redhat.com>
7008         * config/arm/arm.c (get_label_padding): Update for recent
7009         changes to label_to_alignment.
7011         PR tree-optimization/86010
7012         * tree-ssa-dse.c (compute_trims): Fix typo/thinko.
7014         * config/mips/mips.c (vr4130_align_insns): Update for recent
7015         changes to label_to_alignment.
7017         * config/frv/frv.c (frv_label_align): Update for recent changes
7018         to label_to_alignment.
7020         * config/nios2/nios2.c (nios2_label_align): Update for recent
7021         changes which dropped ALIGN_LABELS_LOG.
7023 2018-07-17  Andreas Schwab  <schwab@linux-m68k.org>
7025         * config/m68k/m68k.md (umulsi3_highpart+1, const_umulsi3_highpart)
7026         (smulsi3_highpart+1, const_smulsi3_highpart): Add CC_STATUS_INIT.
7028 2018-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
7030         * config/arc/arc.c (arc_label_align): Use align_labels instead of
7031         deprecated align_labels_log.
7033 2018-07-17  Richard Biener  <rguenther@suse.de>
7035         PR lto/86456
7036         * dwarf2out.c (init_sections_and_labels): Always generate
7037         a debug_line_str_section for early LTO debug.
7038         (dwarf2out_finish): Reset debug_line_str_hash output early.
7039         Bump counter for extra dwarf5 .debug_loc labels to not conflict
7040         with fat LTO part.
7041         (dwarf2out_early_finish): Output debug_line_str.
7043 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
7045         * config/s390/s390.c (preferred_la_operand_p): Do not use LA with
7046         index register on z196 or later.
7048 2018-07-17  Robin Dapp  <rdapp@linux.ibm.com>
7050         * config/s390/s390.c (s390_default_align): Set default function
7051         alignment to 16.
7052         (s390_override_options_after_change): Call s390_default align.
7053         (s390_option_override_internal): Call s390_default_align.
7054         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define.
7056 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
7058         PR middle-end/86542
7059         * omp-low.c (create_task_copyfn): Copy over also fields corresponding
7060         to _looptemp_ clauses, other than the first two.
7062 2018-07-17  Martin Liska  <mliska@suse.cz>
7064         * opts.c: Do not enable OPT_falign_* for -Os.
7066 2018-07-17  Martin Liska  <mliska@suse.cz>
7068         * align.h (MAX_CODE_ALIGN): New.
7069         (MAX_CODE_ALIGN_VALUE): New.
7070         * common/config/i386/i386-common.c (ix86_handle_option):
7071         (MAX_CODE_ALIGN): Moved to align.h.
7072         * final.c (MAX_CODE_ALIGN): Likewise.
7073         * opts.c (parse_and_check_align_values):
7074         (MAX_CODE_ALIGN): Likewise.
7075         (MAX_CODE_ALIGN_VALUE): Likewise.
7077 2018-07-17  Martin Liska  <mliska@suse.cz>
7079         * config/i386/att.h (ASM_OUTPUT_ALIGN): Fix spacing
7080         in order to fulfil coding style.
7081         * config/i386/cygming.h (ASM_OUTPUT_ALIGN): Likewise.
7082         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7083         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
7084         * config/iq2000/iq2000.h (ASM_OUTPUT_ALIGN): Likewise.
7085         * config/pa/pa.h (ASM_OUTPUT_ALIGN): Likewise.
7086         * config/sparc/sol2.h (ASM_OUTPUT_ALIGN_WITH_NOP): Likewise.
7087         * config/sparc/sparc.h (ASM_OUTPUT_ALIGN): Likewise.
7088         * config/visium/visium.h (ASM_OUTPUT_ALIGN): Likewise.
7089         (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
7091 2018-07-17  Martin Liska  <mliska@suse.cz>
7093         * align.h: New file.
7094         * config/alpha/alpha.c (alpha_align_insns_1): Use align_functions
7095         directly.
7096         * config/i386/i386.c (ix86_avoid_jump_mispredicts): Use new return type
7097         align_flags of label_to_alignment.
7098         * config/m32r/m32r.h (LOOP_ALIGN): Wrap returned values into
7099         align_flags class.
7100         * config/m68k/m68k.c: Do not use removed align_labels_value and
7101         align_loops_value.
7102         * config/nds32/nds32.h (JUMP_ALIGN): Wrap result into align_flags class.
7103         (LOOP_ALIGN): Likewise.
7104         (LABEL_ALIGN): Likewise.
7105         * config/powerpcspe/powerpcspe.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7106         Remove not used macro.
7107         (rs6000_loop_align): Change return type to align_flags.
7108         (rs6000_loop_align_max_skip): Remove.
7109         * config/rs6000/rs6000-protos.h (rs6000_loop_align):
7110         Change return type to align_flags.
7111         * config/rs6000/rs6000.c (TARGET_ASM_LOOP_ALIGN_MAX_SKIP):
7112         Remove not used macro.
7113         (rs6000_loop_align):  Change return type to align_flags.
7114         (rs6000_loop_align_max_skip): Remove.
7115         * config/rx/rx.h (JUMP_ALIGN): Wrap integer values
7116         * config/rx/rx-protos.h (rx_align_for_label): Make it
7117         static function.
7118         * config/rx/rx.c (rx_align_for_label): Change return type
7119         to align_flags.
7120         (rx_max_skip_for_label): Remove TARGET_ASM_*_ALIGN_MAX_SKIP
7121         macro definitions.
7122         into align_flags class.
7123         (LABEL_ALIGN): Likewise.
7124         (LOOP_ALIGN): Likewise.
7125         * config/s390/s390.c (s390_label_align): Use align_flags
7126         class member.
7127         (s390_asm_output_function_label): Likewise.
7128         * config/sh/sh.c (sh_override_options_after_change):
7129         Use align_flags class directly without macros.
7130         (find_barrier): Likewise.
7131         (barrier_align): Likewise.
7132         (sh_loop_align): Likewise.
7133         * config/spu/spu.c (spu_option_override):
7134         Use align_flags_tuple::get_value instead of removed macros.
7135         (spu_sched_init): Likewise.
7136         * config/spu/spu.h (GTY): Likewise.
7137         * config/visium/visium.c (visium_option_override):
7138         Set "8" as default secondary alignment.
7139         * config/visium/visium.h (SUBALIGN_LOG): Define to 3
7140         in order to guarantee secondary alignment of 8.
7141         * coretypes.h: Include align.h header file.
7142         * doc/tm.texi: Remove TARGET_ASM_JUMP_ALIGN_MAX_SKIP,
7143         TARGET_ASM_LOOP_ALIGN_MAX_SKIP, TARGET_ASM_LABEL_ALIGN_MAX_SKIP
7144         and TARGET_ASM_LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP macros.
7145         * doc/tm.texi.in: Likewise.
7146         * final.c (struct label_alignment): Remove not used structure.
7147         (LABEL_ALIGN): Change type to align_flags.
7148         (LOOP_ALIGN): Likewise.
7149         (JUMP_ALIGN): Likewise.
7150         (default_loop_align_max_skip): Remove.
7151         (default_label_align_max_skip): Likewise.
7152         (default_jump_align_max_skip): Likewise.
7153         (default_label_align_after_barrier_max_skip):
7154         (LABEL_TO_ALIGNMENT): Change to access label_align vector.
7155         (LABEL_TO_MAX_SKIP): Remove.
7156         (label_to_alignment): Return align_flags type instead of integer.
7157         (label_to_max_skip): Remove.
7158         (align_fuzz): Use align_flags type.
7159         (compute_alignments): Use align_flags type and use align_flags::max
7160         to combine multiple alignments.
7161         (grow_label_align): Grow vec instead of C array.
7162         (update_alignments): Assign just LABEL_TO_ALIGNMENT.
7163         (shorten_branches):  Use align_flags type and use align_flags::max
7164         to combine multiple alignments.
7165         (final_scan_insn_1): Remove usage of secondary alignment that comes
7166         from label alignment, but instead use proper secondary alignment
7167         which is computed in grow_label_align.
7168         * flags.h (struct align_flags_tuple): Move to align.h.
7169         (struct align_flags): Likewise.
7170         (state_align_loops): Rename to align_loops.
7171         (state_align_jumps): Rename to align_jumps.
7172         (state_align_labels): Rename to align_labels.
7173         (state_align_functions): Rename to align_functions.
7174         (align_loops_log): Remove.
7175         (align_jumps_log): Remove.
7176         (align_labels_log): Remove.
7177         (align_functions_log): Remove.
7178         (align_loops_max_skip): Remove.
7179         (align_jumps_max_skip): Remove.
7180         (align_labels_max_skip): Remove.
7181         (align_functions_max_skip): Remove.
7182         (align_loops_value): Remove.
7183         (align_jumps_value): Remove.
7184         (align_labels_value): Remove.
7185         (align_functions_value): Remove.
7186         * output.h (label_to_alignment): Change return type to align_flags.
7187         (label_to_max_skip): Remove.
7188         * target.def: Remove loop_align_max_skip, label_align_max_skip,
7189         jump_align_max_skip macros.
7190         * targhooks.h (default_loop_align_max_skip): Remove.
7191         (default_label_align_max_skip): Likewise.
7192         (default_jump_align_max_skip): Likewise.
7193         (default_label_align_after_barrier_max_skip): Remove.
7194         * toplev.c (read_log_maxskip): Use ::normalize function.
7195         (parse_N_M): Remove not used argument and also call ::normalize.
7196         (parse_alignment_opts): Do not pass unused arguments.
7197         * varasm.c (assemble_start_function): Use directly align_functions
7198         instead of removed macros.
7199         * system.h: Do not poison removed macros.
7201 2018-07-17  Jakub Jelinek  <jakub@redhat.com>
7203         PR middle-end/86539
7204         * gimplify.c (gimplify_omp_for): Ensure taskloop firstprivatized init
7205         and cond temporaries don't have reference type if iterator has
7206         pointer type.  For init use &for_pre_body instead of pre_p if
7207         for_pre_body is non-empty.
7209 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
7211         * config/rs6000/rs6000.md (trunc<mode>sf2): Expand truncates of
7212         double-double modes to SFmode directly directly.
7213         (trunc<mode>sf2_fprs): Delete.
7215 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
7217         * config/rs6000/rs6000.c (init_float128_ibm): Use the correct names
7218         for conversions between IFmode and the decimal floating point modes.
7219         (init_float128_ieee): Use the correct names for conversions between
7220         KFmode and the decimal floating point modes.
7222 2018-07-16  Segher Boessenkool  <segher@kernel.crashing.org>
7224         * config/rs6000/rs6000.c (init_float128_ibm): Use more correct names
7225         for the conversions between TDmode and IFmode.
7226         (init_float128_ieee): Use more correct names for the conversions
7227         between TDmode and KFmode.
7229 2018-07-16  Jakub Jelinek  <jakub@redhat.com>
7231         PR tree-optimization/86526
7232         * builtins.c (expand_builtin_memcmp): Formatting fixes.
7233         (inline_expand_builtin_string_cmp): Likewise.
7234         (inline_string_cmp): Likewise.  Use c_readstr instead of
7235         builtin_memcpy_read_str.  Add unit_mode temporary.
7237 2018-07-16  Bernd Edlinger  <bernd.edlinger@hotmail.de>
7239         PR middle-end/86528
7240         * builtins.c (check_access): Bail out if range[0] is no INTEGER_CST.
7241         * expr.c (string_constant): Fix the element size of ARRAY_TYPE.
7243 2018-07-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
7245         * doc/extend.texi (PowerPC AltiVec Built-in Functions):
7246         Alphabetize prototypes of built-in functions, separating out
7247         built-in functions that are listed in this section but should be
7248         described elsewhere.
7250 2018-07-16  Uros Bizjak  <ubizjak@gmail.com>
7252         PR target/86511
7253         * expmed.c (emit_store_flag): Do not emit setcc followed by a
7254         conditional move when trapping comparison was split to a
7255         non-trapping one (and vice versa).
7257 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
7259         * config/s390/s390.c (s390_function_profiler): Generate nops
7260         instead of profiler call sequences.
7261         * config/s390/s390.opt: Add the new option.
7263 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
7265         * config/s390/s390.c (s390_function_profiler): Generate
7266         __mcount_loc section.
7267         * config/s390/s390.opt: Add the new option.
7269 2018-07-16  Ilya Leoshkevich  <iii@linux.ibm.com>
7271         * common.opt: Add the new warning.
7272         * config/s390/s390.c (s390_function_profiler): Emit "brasl
7273         %r0,__fentry__" when -mfentry is specified.
7274         (s390_option_override_internal): Disallow -mfentry for 31-bit
7275         CPUs.
7276         * config/s390/s390.opt: Add the new option.
7278 2018-07-16  Richard Biener  <rguenther@suse.de>
7280         PR lto/86523
7281         * dwarf2out.c (dwarf2out_register_external_die): Assign DIE parents
7282         for function-local FUNCTION_DECL and RESULT_DECL immediately.
7284 2018-07-16  Martin Liska  <mliska@suse.cz>
7286         PR ipa/86529
7287         * ipa-pure-const.c (malloc_candidate_p): Revert ::get
7288         to ::get_create.
7290 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
7292         * config/arc/arcHS.md: Update ARCHS scheduling rules.
7294 2017-07-16  Claudiu Zissulescu  <claziss@synopsys.com>
7296         * config/arc/arc-arch.h (arc_tune_attr): Add new tune parameters
7297         for ARCHS4x.
7298         * config/arc/arc-cpus.def (hs4x): New cpu.
7299         (hs4xd): Likewise.
7300         * config/arc/arc-tables.opt: Regenerate.
7301         * config/arc/arc.c (arc_sched_issue_rate): New function.
7302         (TARGET_SCHED_ISSUE_RATE): Define.
7303         (TARGET_SCHED_EXPOSED_PIPELINE): Likewise.
7304         * config/arc/arc.md (attr type): Add fpu_fuse, fpu_sdiv, fpu_ddiv,
7305         fpu_cvt.
7306         (attr tune): Add ARCHS4x tune values.
7307         (attr tune_dspmpy): Define.
7308         (*tst): Correct instruction type.
7309         * config/arc/arcHS.md: Don't use this automaton for ARCHS4x cpus.
7310         * config/arc/arcHS4x.md: New file.
7311         * config/arc/fpu.md: Update instruction type attributes.
7312         * config/arc/t-multilib: Regenerate.
7314 2018-07-16  Tom de Vries  <tdevries@suse.de>
7316         PR debug/86455
7317         * var-tracking.c (vt_initialize): Fix pre_dec handling.
7319 2018-07-16  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
7321         * config/aarch64/atomics.md (aarch64_store_execlusive<mode>): Add
7322         early clobber.
7324 2018-07-16  Eric Botcazou  <ebotcazou@adacore.com>
7326         PR tree-optimization/86514
7327         * tree-ssa-reassoc.c (init_range_entry) <CASE_CONVERT>: Return for a
7328         conversion to a boolean type from a type with greater precision.
7330 2018-07-16  Tom de Vries  <tdevries@suse.de>
7332         * var-tracking.c (vt_initialize): Print adjusted insn slim if
7333         dump_flags request TDF_SLIM.
7335 2018-07-16  Aldy Hernandez  <aldyh@redhat.com>
7337         * fold-const.c (int_const_binop_1): Abstract...
7338         (wide_int_binop): ...wide int code here.
7339         (poly_int_binop): ...poly int code here.
7340         Abstract the rest of int_const_binop_1 into int_const_binop.
7341         * fold-const.h (wide_int_binop): New.
7342         * tree-vrp.c (vrp_int_const_binop): Call wide_int_binop.
7343         Remove useless PLUS/MINUS_EXPR case.
7344         (zero_nonzero_bits_from_vr): Move wide int code...
7345         (zero_nonzero_bits_from_bounds): ...here.
7346         (extract_range_from_binary_expr_1): Move mask optimization code...
7347         (range_easy_mask_min_max): ...here.
7348         * tree-vrp.h (zero_nonzero_bits_from_bounds): New.
7349         (range_easy_mask_min_max): New.
7351 2018-07-15  Jeff Law  <law@redhat.com>
7353         PR target/85993
7354         * config/sh/sh.c (output_mi_thunk): Remove dead conditional
7355         block.
7357 2018-07-14  Jim Wilson  <jimw@sifive.com>
7359         * config/riscv/linux.h (TARGET_ASM_FILE_END): New.
7361 2018-07-14  Paul Koning  <ni1d@arrl.net>
7363         * config/pdp11/pdp11.c (pdp11_rtx_costs): Bugfixes.
7365 2018-07-13  Jan Hubicka  <hubicka@ucw.cz>
7367         * lto-streamer-out.c (copy_function_or_variable): Dump info about
7368         copying section.
7370 2018-07-13  Bill Schmidt  <wschmidt@linux.ibm.com>
7371             Steve Munroe  <munroesj52@gmail.com>
7373         * config/rs6000/emmintrin.h (_mm_and_si128): New function.
7374         (_mm_andnot_si128): Likewise.
7375         (_mm_or_si128): Likewise.
7376         (_mm_xor_si128): Likewise.
7378 2018-07-13  Qing Zhao  <qing.zhao@oracle.com>
7380         PR middle-end/78809
7381         * builtins.c (expand_builtin_memcmp): Inline the calls first
7382         when result_eq is false.
7383         (expand_builtin_strcmp): Inline the calls first.
7384         (expand_builtin_strncmp): Likewise.
7385         (inline_string_cmp): New routine. Expand a string compare
7386         call by using a sequence of char comparison.
7387         (inline_expand_builtin_string_cmp): New routine. Inline expansion
7388         a call to str(n)cmp/memcmp.
7389         * doc/invoke.texi (--param builtin-string-cmp-inline-length):
7390         New option.
7391         * params.def (BUILTIN_STRING_CMP_INLINE_LENGTH): New.
7393 2018-07-13  Richard Earnshaw  <rearnsha@arm.com>
7395         * config/arm/driver-arm.c: Include arm-native.h.
7396         (host_detect_local_cpu): Use auto-generated data tables.
7397         (vendors, arm_cpu_table): Delete.  Move part information to ...
7398         * config/arm/arm-cpus.in: ... here.
7399         * config/arm/parsecpu.awk (gen_native): New function.
7400         (vendor, part): New CPU fields.
7401         (END): Add support for building the native CPU detection tables.
7402         * config/arm/t-arm (arm-native.h): Add build rule.
7403         (driver-arm.o): Add dependency on arm-native.h.
7405 2018-07-13  Richard Biener  <rguenther@suse.de>
7407         PR middle-end/85974
7408         * match.pd (addr1 - addr2): Allow either of the operand to
7409         have a conversion.
7411 2018-07-13  Tom de Vries  <tdevries@suse.de>
7413         * tree-inline.c (remap_ssa_name): Save and reuse debug exprs generated
7414         in remap_ssa_name.
7416 2018-07-13  Jackson Woodruff  <jackson.woodruff@arm.com>
7418         * config/aarch64/aarch64.c (aarch64_operands_adjust_ok_for_ldpstp): Use
7419         arrays instead of numbered variables.
7421 2018-07-13  Eric Botcazou  <ebotcazou@adacore.com>
7423         * config/sparc/sparc-protos.h (sparc_compute_frame_size): Delete.
7424         * config/sparc/sparc.c (sparc_compute_frame_size): Make static.
7426 2018-07-13  Richard Biener  <rguenther@suse.de>
7428         PR debug/86452
7429         * dwarf2out.c (gen_type_die_with_usage): Use scope_die_for
7430         instead of get_context_die.
7432 2018-07-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
7433             Richard Biener  <rguenther@suse.de>
7435         PR middle-end/86489
7436         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Check
7437         that the loop latch destination where phi is defined.
7439 2018-07-12  Kito Cheng  <kito.cheng@gmail.com>
7441         * config/riscv/riscv.c (enum riscv_privilege_levels): Add UNKNOWN_MODE.
7442         (riscv_expand_epilogue): Add assertion to check interrupt mode.
7443         (riscv_set_current_function): Extract getting interrupt type to new
7444         function.
7445         (riscv_get_interrupt_type): New function.
7446         (riscv_merge_decl_attributes): New function, checking interrupt type is
7447         same.
7448         (TARGET_MERGE_DECL_ATTRIBUTES): Define.
7450 2018-07-12  Paul Koning  <ni1d@arrl.net>
7452         * config/pdp11/pdp11.c (pdp11_output_def): Fix typo in .set
7453         directive.
7455 2018-07-12  Paul Koning  <ni1d@arrl.net>
7457         * doc/rtl.texi (REG_NONNEG): Remove decrement and branch until
7458         zero reference, add doloop_end instead.
7459         * doc/md.texi (decrement_and_branch_until_zero): Remove.
7460         (Looping patterns): Remove decrement_and_branch_until_zero.  Add
7461         detail for doloop_end.
7463 2018-07-12  Martin Sebor  <msebor@redhat.com>
7465         PR c/86453
7466         * attribs.c (decl_attributes): Reject conflicting attributes before
7467         calling attribute handlers.
7469 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
7471         * dumpfile.c (gcc::dump_manager::get_dump_file_name): Add PART
7472          parameter.
7473         (gcc::dump_manager::get_dump_file_name): likewise.
7474         (dump_begin): Likewise.
7475         * dumpfile.h (dump_begin): Update prototype.
7476         (gcc::dump_manager::get_dump_file_name,
7477         gcc::dump_manager::get_dump_file_name): Update prototype.
7479 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
7481         * internal-fn.h (vectorizable_internal_fn_p): New function.
7482         * tree-vect-slp.c (compatible_calls_p): Likewise.
7483         (vect_build_slp_tree_1): Remove nops argument.  Handle calls
7484         to internal functions.
7485         (vect_build_slp_tree_2): Update call to vect_build_slp_tree_1.
7487 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
7489         * fold-const.h (inverse_conditions_p): Declare.
7490         * fold-const.c (inverse_conditions_p): New function.
7491         * match.pd: Use inverse_conditions_p.  Add folds of view_converts
7492         that test the inverse condition of a conditional internal function.
7493         * internal-fn.h (vectorized_internal_fn_supported_p): Declare.
7494         * internal-fn.c (internal_fn_mask_index): Handle conditional
7495         internal functions.
7496         (vectorized_internal_fn_supported_p): New function.
7497         * tree-if-conv.c: Include internal-fn.h and fold-const.h.
7498         (any_pred_load_store): Replace with...
7499         (need_to_predicate): ...this new variable.
7500         (redundant_ssa_names): New variable.
7501         (ifcvt_can_use_mask_load_store): Move initial checks to...
7502         (ifcvt_can_predicate): ...this new function.  Handle tree codes
7503         for which a conditional internal function exists.
7504         (if_convertible_gimple_assign_stmt_p): Use ifcvt_can_predicate
7505         instead of ifcvt_can_use_mask_load_store.  Update after variable
7506         name change.
7507         (predicate_load_or_store): New function, split out from
7508         predicate_mem_writes.
7509         (check_redundant_cond_expr): New function.
7510         (value_available_p): Likewise.
7511         (predicate_rhs_code): Likewise.
7512         (predicate_mem_writes): Rename to...
7513         (predicate_statements): ...this.  Use predicate_load_or_store
7514         and predicate_rhs_code.
7515         (combine_blocks, tree_if_conversion): Update after above name changes.
7516         (ifcvt_local_dce): Handle redundant_ssa_names.
7517         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): Handle
7518         general conditional functions.
7519         * tree-vect-stmts.c (vectorizable_call): Likewise.
7521 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
7522             Alan Hayward  <alan.hayward@arm.com>
7523             David Sherwood  <david.sherwood@arm.com>
7525         * internal-fn.h (can_interpret_as_conditional_op_p): Declare.
7526         * internal-fn.c (can_interpret_as_conditional_op_p): New function.
7527         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Handle conditional
7528         plus and minus and convert them into IFN_COND_FMA-based sequences.
7529         (convert_mult_to_fma): Handle conditional plus and minus.
7531 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
7533         * doc/md.texi (cond_fma, cond_fms, cond_fnma, cond_fnms): Document.
7534         * optabs.def (cond_fma_optab, cond_fms_optab, cond_fnma_optab)
7535         (cond_fnms_optab): New optabs.
7536         * internal-fn.def (COND_FMA, COND_FMS, COND_FNMA, COND_FNMS): New
7537         internal functions.
7538         (FMA): Use DEF_INTERNAL_FLT_FN rather than DEF_INTERNAL_FLT_FLOATN_FN.
7539         * internal-fn.h (get_conditional_internal_fn): Declare.
7540         (get_unconditional_internal_fn): Likewise.
7541         * internal-fn.c (cond_ternary_direct): New macro.
7542         (expand_cond_ternary_optab_fn): Likewise.
7543         (direct_cond_ternary_optab_supported_p): Likewise.
7544         (FOR_EACH_COND_FN_PAIR): Likewise.
7545         (get_conditional_internal_fn): New function.
7546         (get_unconditional_internal_fn): Likewise.
7547         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 5.
7548         (gimple_match_op::gimple_match_op): Add a new overload for 5
7549         operands.
7550         (gimple_match_op::set_op): Likewise.
7551         (gimple_resimplify5): Declare.
7552         * genmatch.c (decision_tree::gen): Generate simplifications for
7553         5 operands.
7554         * gimple-match-head.c (gimple_simplify): Define an overload for
7555         5 operands.  Handle calls with 5 arguments in the top-level overload.
7556         (convert_conditional_op): Handle conversions from unconditional
7557         internal functions to conditional ones.
7558         (gimple_resimplify5): New function.
7559         (build_call_internal): Pass a fifth operand.
7560         (maybe_push_res_to_seq): Likewise.
7561         (try_conditional_simplification): Try converting conditional
7562         internal functions to unconditional internal functions.
7563         Handle 3-operand unconditional forms.
7564         * match.pd (UNCOND_TERNARY, COND_TERNARY): Operator lists.
7565         Define ternary equivalents of the current rules for binary conditional
7566         internal functions.
7567         * config/aarch64/aarch64.c (aarch64_preferred_else_value): Handle
7568         ternary operations.
7569         * config/aarch64/iterators.md (UNSPEC_COND_FMLA, UNSPEC_COND_FMLS)
7570         (UNSPEC_COND_FNMLA, UNSPEC_COND_FNMLS): New unspecs.
7571         (optab): Handle them.
7572         (SVE_COND_FP_TERNARY): New int iterator.
7573         (sve_fmla_op, sve_fmad_op): New int attributes.
7574         * config/aarch64/aarch64-sve.md (cond_<optab><mode>)
7575         (*cond_<optab><mode>_2, *cond_<optab><mode_4)
7576         (*cond_<optab><mode>_any): New SVE_COND_FP_TERNARY patterns.
7578 2018-07-12  Richard Sandiford  <richard.sandiford@linaro.org>
7580         * target.def (preferred_else_value): New target hook.
7581         * doc/tm.texi.in (TARGET_PREFERRED_ELSE_VALUE): New hook.
7582         * doc/tm.texi: Regenerate.
7583         * targhooks.h (default_preferred_else_value): Declare.
7584         * targhooks.c (default_preferred_else_value): New function.
7585         * internal-fn.h (conditional_internal_fn_code): Declare.
7586         * internal-fn.c (FOR_EACH_CODE_MAPPING): New macro.
7587         (get_conditional_internal_fn): Use it.
7588         (conditional_internal_fn_code): New function.
7589         * gimple-match.h (gimple_match_cond): New struct.
7590         (gimple_match_op): Add a cond member function.
7591         (gimple_match_op::gimple_match_op): Update all forms to take a
7592         gimple_match_cond.
7593         * genmatch.c (expr::gen_transform): Use the same condition as res_op
7594         for the suboperation, but don't specify a particular else_value.
7595         * tree-ssa-sccvn.c (vn_nary_simplify, vn_reference_lookup_3)
7596         (visit_nary_op, visit_reference_op_load): Pass
7597         gimple_match_cond::UNCOND to the gimple_match_op constructor.
7598         * gimple-match-head.c: Include tree-eh.h
7599         (convert_conditional_op): New function.
7600         (maybe_resimplify_conditional_op): Likewise.
7601         (gimple_resimplify1): Call maybe_resimplify_conditional_op.
7602         (gimple_resimplify2): Likewise.
7603         (gimple_resimplify3): Likewise.
7604         (gimple_resimplify4): Likewise.
7605         (maybe_push_res_to_seq): Return null for conditional operations.
7606         (try_conditional_simplification): New function.
7607         (gimple_simplify): Call it.  Pass conditions to the gimple_match_op
7608         constructor.
7609         * match.pd: Fold VEC_COND_EXPRs of an IFN_COND_* call to a new
7610         IFN_COND_* call.
7611         * config/aarch64/aarch64.c (aarch64_preferred_else_value): New
7612         function.
7613         (TARGET_PREFERRED_ELSE_VALUE): Redefine.
7615 2018-07-12  Jan Hubicka  <hubicka@ucw.cz>
7617         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
7618         DECL_FCONTEXT
7619         (hash_tree): Do not hash DECL_FCONTEXT
7620         * tree-streamer-in.c (lto_input_ts_field_decl_tree_pointers):
7621         Do not stream DECL_FCONTEXT.
7622         * tree-streamer-out.c (write_ts_field_decl_tree_pointers): Likewise.
7623         * tree.c (free_lang_data_in_decl): Free DECL_FCONTEXT.
7625 2018-07-12  Richard Biener  <rguenther@suse.de>
7627         PR debug/86462
7628         * dwarf2out.c (gen_block_die): Only output blocks when they have
7629         at least one !DECL_IGNORED_P variable.
7631 2018-07-12  Richard Biener  <rguenther@suse.de>
7633         PR target/84829
7634         * config/gnu-user.h (GNU_USER_TARGET_NO_PTHREADS_LIB_SPEC):
7635         Remove -mieee-fp handling.
7637 2018-07-12  Richard Biener  <rguenther@suse.de>
7639         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Remove bogus
7640         left-over from last patch.
7642 2018-07-12  Jakub Jelinek  <jakub@redhat.com>
7644         PR tree-optimization/86492
7645         * gimple-ssa-store-merging.c
7646         (imm_store_chain_info::coalesce_immediate_stores): Call
7647         check_no_overlap even for the merge_overlapping case.  Formatting fix.
7649 2018-07-12  Richard Biener  <rguenther@suse.de>
7651         PR middle-end/86479
7652         * fold-const.c (fold_binary_op_with_conditional_arg): Do not
7653         move possibly trapping operations into the conditional.
7655 2018-07-12  Richard Biener  <rguenther@suse.de>
7657         * tree-ssa-sccvn.c (mprts_hook_cnt): Remove.
7658         (vn_lookup_simplify_result): Remove recursion limit applied
7659         here.
7660         (vn_nary_build_or_lookup_1): Adjust.
7661         (try_to_simplify): Likewise.
7662         * gimple-match-head.c (gimple_resimplify1): Instead apply one
7663         here.
7664         (gimple_resimplify2): Likewise.
7665         (gimple_resimplify3): Likewise.
7666         (gimple_resimplify4): Likewise.
7668 2018-07-11  Jakub Jelinek  <jakub@redhat.com>
7670         * config/i386/avx512bitalgintrin.h (_mm512_mask_bitshuffle_epi64_mask):
7671         Use __mmask64 type instead of __mmask8 for __M argument.
7672         * config/i386/avx512fintrin.h (_mm512_mask_xor_epi64,
7673         _mm512_maskz_xor_epi64): Use __mmask8 type instead of __mmask16 for
7674         __U argument.
7675         (_mm512_mask_cmpneq_epi64_mask): Use __mmask8 type instead of
7676         __mmask16 for __M argument.
7677         (_mm512_maskz_insertf32x4, _mm512_maskz_inserti32x4,
7678         _mm512_mask_insertf32x4, _mm512_mask_inserti32x4): Cast last argument
7679         to __mmask16 instead of __mmask8.
7680         * config/i386/avx512vlintrin.h (_mm_mask_add_ps, _mm_maskz_add_ps,
7681         _mm256_mask_add_ps, _mm256_maskz_add_ps, _mm_mask_sub_ps,
7682         _mm_maskz_sub_ps, _mm256_mask_sub_ps, _mm256_maskz_sub_ps,
7683         _mm256_maskz_cvtepi32_ps, _mm_maskz_cvtepi32_ps): Use __mmask8 type
7684         instead of __mmask16 for __U argument.
7685         * config/i386/avx512vlbwintrin.h (_mm_mask_cmp_epi8_mask): Use
7686         __mmask16 instead of __mmask8 for __U argument.
7687         (_mm256_mask_cmp_epi8_mask): Use __mmask32 instead of __mmask16 for
7688         __U argument.
7689         (_mm256_cmp_epi8_mask): Use __mmask32 return type instead of
7690         __mmask16.
7691         (_mm_mask_cmp_epu8_mask): Use __mmask16 instead of __mmask8 for __U
7692         argument.
7693         (_mm256_mask_cmp_epu8_mask): Use __mmask32 instead of __mmask16 for
7694         __U argument.
7695         (_mm256_cmp_epu8_mask): Use __mmask32 return type instead of
7696         __mmask16.
7697         (_mm_mask_cmp_epi16_mask): Cast last argument to __mmask8 instead
7698         of __mmask16.
7699         (_mm256_mask_cvtepi8_epi16): Use __mmask16 instead of __mmask32 for
7700         __U argument.
7701         (_mm_mask_cvtepi8_epi16): Use __mmask8 instead of __mmask32 for
7702         __U argument.
7703         (_mm256_mask_cvtepu8_epi16): Use __mmask16 instead of __mmask32 for
7704         __U argument.
7705         (_mm_mask_cvtepu8_epi16): Use __mmask8 instead of __mmask32 for
7706         __U argument.
7707         (_mm256_mask_cmpneq_epu8_mask, _mm256_mask_cmplt_epu8_mask,
7708         _mm256_mask_cmpge_epu8_mask, _mm256_mask_cmple_epu8_mask): Change
7709         return type as well as __M argument type and all casts from __mmask8
7710         to __mmask32.
7711         (_mm256_mask_cmpneq_epu16_mask, _mm256_mask_cmplt_epu16_mask,
7712         _mm256_mask_cmpge_epu16_mask, _mm256_mask_cmple_epu16_mask): Change
7713         return type as well as __M argument type and all casts from __mmask8
7714         to __mmask16.
7715         (_mm256_mask_cmpneq_epi8_mask, _mm256_mask_cmplt_epi8_mask,
7716         _mm256_mask_cmpge_epi8_mask, _mm256_mask_cmple_epi8_mask): Change
7717         return type as well as __M argument type and all casts from __mmask8
7718         to __mmask32.
7719         (_mm256_mask_cmpneq_epi16_mask, _mm256_mask_cmplt_epi16_mask,
7720         _mm256_mask_cmpge_epi16_mask, _mm256_mask_cmple_epi16_mask): Change
7721         return type as well as __M argument type and all casts from __mmask8
7722         to __mmask16.
7723         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_shrdi_epi32,
7724         _mm_mask_shldi_epi32): Cast last argument to __mmask8 instead of
7725         __mmask16.
7727 2018-07-11  Grazvydas Ignotas  <notasas@gmail.com>
7729         * config/i386/avx512bwintrin.h: (_mm512_mask_cmp_epi8_mask,
7730         _mm512_mask_cmp_epu8_mask): Use __mmask64 type instead of __mmask32
7731         for __U argument.
7733 2018-07-11  Paul Koning  <ni1d@arrl.net>
7735         * doc/md.texi (define_subst): Document how multiple occurrences of
7736         the same argument in the replacement pattern are handled.
7738 2018-07-11  Paul Koning  <ni1d@arrl.net>
7740         * doc/extend.texi (Common Variable Attributes): Move "mode" into
7741         alphabetical order.
7742         (Common Type Attributes): Add "mode" attribute.
7744 2018-07-11  Jan Hubicka  <hubicka@ucw.cz>
7746         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not
7747         stream DECL_ORIGINAL_TYPE.
7748         (DFS::DFS_write_tree_body): Drop hack handling local external decls.
7749         (hash_tree): Do not walk DECL_ORIGINAL_TYPE.
7750         * tree-streamer-in.c (lto_input_ts_decl_non_common_tree_pointers):
7751         Do not walk original type.
7752         * tree-streamer-out.c (streamer_write_chain): Drop hack handling
7753         external decls.
7754         (write_ts_decl_non_common_tree_pointers): Do not stream
7755         DECL_ORIGINAL_TYPE
7756         * tree.c (free_lang_data_in_decl): Clear DECL_ORIGINAL_TYPE.
7757         (find_decls_types_r): Do not walk DEC_ORIGINAL_TYPE.
7759 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
7761         * tree-ssa-threadupdate.c (thread_through_all_blocks): Do not jump
7762         thread twice from the same starting edge.
7764 2018-07-11  Aldy Hernandez  <aldyh@redhat.com>
7766         * vr-values.c (gimple_stmt_nonzero_p): Abstract common code to...
7767         * gimple.c (gimple_call_nonnull_result_p): ...here...
7768         (gimple_call_nonnull_arg): ...and here.
7769         * gimple.h (gimple_call_nonnull_result_p): New.
7770         (gimple_call_nonnull_arg): New.
7772 2018-07-11  Richard Earnshaw  <rearnsha@arm.com>
7774         * config/arm/arm-cpus.in: Move information from fpu field of each
7775         cpu definition to the isa field.
7776         * config/arm/parsecpu.awk (fpu): Delete match rule.
7777         (gen_comm_data): Don't add bits from the CPU's FPU entry.
7779 2018-07-11  Richard Biener  <rguenther@suse.de>
7781         PR debug/86457
7782         * dwarf2out.c (init_sections_and_labels): Use
7783         output_asm_line_debug_info consistently.
7784         (dwarf2out_early_finish): Likewise.
7785         (dwarf2out_finish): Remove DW_AT_stmt_list from early generated
7786         type units.
7788 2018-07-11  Richard Biener  <rguenther@suse.de>
7790         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1):
7791         Rework father_bb setting in a way to avoid propagating constants
7792         multiple times on a loop body.
7794 2018-07-10  Mark Wielaard  <mark@klomp.org>
7796         PR debug/86459
7797         * dwarf2out.c (output_macinfo_op): Fix dwarf_FORM typo in gcc_assert.
7799 2018-07-10  Richard Biener  <rguenther@suse.de>
7801         * hash-map.h (hash_map::iterator::operator*): Return
7802         references to key and value.
7804 2018-07-10  Jakub Jelinek  <jakub@redhat.com>
7806         PR c++/86443
7807         * gimplify.c (find_combined_omp_for): Add DATA argument, in addition
7808         to finding the inner OMP_FOR/OMP_SIMD stmt find non-trivial wrappers,
7809         BLOCKs with BLOCK_VARs, OMP_PARALLEL in between, OMP_FOR in between.
7810         (gimplify_omp_for): For composite loops, move outer
7811         OMP_{DISTRIBUTE,TASKLOOP,FOR,PARALLEL} right around innermost
7812         OMP_FOR/OMP_SIMD if there are any non-trivial wrappers.  For class
7813         iterators add any needed clauses.  Allow OMP_FOR_ORIG_DECLS to contain
7814         TREE_LIST for both the original class iterator and the "last" helper
7815         var.  Gimplify OMP_FOR_PRE_BODY before the outermost composite
7816         loop, remember has_decl_expr from outer composite loops for the
7817         innermost OMP_SIMD in TREE_PRIVATE bit on OMP_FOR_INIT.
7819 2018-07-09  Martin Sebor  <msebor@redhat.com>
7821         PR middle-end/77357
7822         PR middle-end/86428
7823         * builtins.c (c_strlen): Avoid out-of-bounds warnings when
7824         accessing implicitly initialized array elements.
7825         * expr.c (string_constant): Handle string initializers of
7826         character arrays within aggregates.
7827         * gimple-fold.c (fold_array_ctor_reference): Add argument.
7828         Store element offset.  As a special case, handle zero size.
7829         (fold_nonarray_ctor_reference): Same.
7830         (fold_ctor_reference): Add argument.  Store subobject offset.
7831         * gimple-fold.h (fold_ctor_reference): Add argument.
7833 2018-07-09  Paul Koning  <ni1d@arrl.net>
7835         * config/pdp11/pdp11.c (pdp11_addr_cost): New function.
7836         (pdp11_insn_cost): New function.
7837         (pdp11_md_asm_adjust): New function.
7838         (TARGET_INVALID_WITHIN_DOLOOP): Define.
7839         (pdp11_rtx_costs): Update to match machine better.
7840         (output_addr_const_pdp11): Correct format mismatch warnings.
7841         * config/pdp11/pdp11.h (SLOW_BYTE_ACCESS): Correct definition.
7842         * config/pdp11/pdp11.md: General change to add base_cost and/or
7843         length attributes for use by new pdp11_insn_cost function.
7844         (MIN_BRANCH): Correct definition.
7845         (MIN_SOB): Ditto.
7846         (doloop_end): Use standard pattern name for looping pattern.
7847         (doloop_end_nocc): New.
7848         (movsf): Add another constraint alternative.
7849         (zero_extendqihi2): Add constraint alternatives for not in place
7850         extend.
7851         (zero_extendhisi2): Remove.
7852         (shift patterns): Add CC handling variants.
7853         (bswaphi2): New.
7854         (bswapsi2): New.
7855         (rothi3): New.
7856         (define_peephole2): New peephole to recognize mov that sets CC for
7857         subsequent test.
7859 2018-07-09  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7861         * config/sparc/sparc.c (sparc_fold_builtin) <SPARC_BUILTIN_PDIST,
7862         SPARC_BUILTIN_PDISTN>: Adapt for signature change of wi::neg,
7863         wi::add.
7865 2018-07-09  Jakub Jelinek  <jakub@redhat.com>
7867         PR c/86420
7868         * real.c (real_nextafter): Return true if result is denormal.
7870 2018-07-09  Martin Liska  <mliska@suse.cz>
7872         * common.opt: Add back wrongly removed attribute.
7874 2018-07-09  Richard Biener  <rguenther@suse.de>
7876         PR debug/86413
7877         * dwarf2out.c (gen_block_die): For an early generated DIE
7878         always output high/low PC attributes.
7880 2018-07-09  Tom de Vries  <tdevries@suse.de>
7882         * cfgexpand.c (expand_debug_source_expr): Handle VAR_DECL.
7883         * tree-inline.c (remap_ssa_name): Handle default def ssa_name mapping
7884         onto VAR_DECL with abstract origin.
7886 2018-07-07  Jim Wilson  <jimw@sifive.com>
7888         * config/riscv/riscv.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): New.
7890 2018-07-07  Tom de Vries  <tdevries@suse.de>
7892         * tree-dfa.c (dump_enumerated_decls): Handle cfun->cfg == NULL.
7894 2018-07-07  Aldy Hernandez  <aldyh@redhat.com>
7896         * tree-vrp.c (vrp_int_const_binop): Change overflow type to
7897         overflow_type.
7898         (combine_bound): Use wide-int overflow calculation instead of
7899         rolling our own.
7900         * calls.c (maybe_warn_alloc_args_overflow): Change overflow type to
7901         overflow_type.
7902         * fold-const.c (int_const_binop_2): Same.
7903         (extract_muldiv_1): Same.
7904         (fold_div_compare): Same.
7905         (fold_abs_const): Same.
7906         * match.pd: Same.
7907         * poly-int.h (add): Same.
7908         (sub): Same.
7909         (neg): Same.
7910         (mul): Same.
7911         * predict.c (predict_iv_comparison): Same.
7912         * profile-count.c (slow_safe_scale_64bit): Same.
7913         * simplify-rtx.c (simplify_const_binary_operation): Same.
7914         * tree-chrec.c (tree_fold_binomial): Same.
7915         * tree-data-ref.c (split_constant_offset_1): Same.
7916         * tree-if-conv.c (idx_within_array_bound): Same.
7917         * tree-scalar-evolution.c (iv_can_overflow_p): Same.
7918         * tree-ssa-phiopt.c (minmax_replacement): Same.
7919         * tree-vect-loop.c (is_nonwrapping_integer_induction): Same.
7920         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Same.
7921         * vr-values.c (vr_values::adjust_range_with_scev): Same.
7922         * wide-int.cc (wi::add_large): Same.
7923         (wi::mul_internal): Same.
7924         (wi::sub_large): Same.
7925         (wi::divmod_internal): Same.
7926         * wide-int.h: Change overflow type to overflow_type for neg, add,
7927         mul, smul, umul, div_trunc, div_floor, div_ceil, div_round,
7928         mod_trunc, mod_ceil, mod_round, add_large, sub_large,
7929         mul_internal, divmod_internal.
7930         (overflow_type): New enum.
7931         (accumulate_overflow): New.
7933 2018-07-06  Kugan Vivekanandarajah  <kugan.vivekanandarajah@linaro.org>
7935         * tree-ssa-phiopt.c (cond_removal_in_popcount_pattern): New.
7936         (tree_ssa_phiopt_worker): Call cond_removal_in_popcount_pattern.
7938 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
7940         * tree-ssa-loop-niter.c (number_of_iterations_popcount): If popcount
7941         argument is checked for zero before entering loop, avoid checking again.
7943 2018-07-06  Kugan Vivekanandarajah  <kuganv@linaro.org>
7945         * gimplify.h (generic_expr_could_trap_p): Set as global function.
7946         * gimplify.h (generic_expr_could_trap_p): Likwise.
7947         * tree-scalar-evolution.c (expression_expensive_p): Handle COND_EXPR.
7949 2018-07-06  Jakub Jelinek  <jakub@redhat.com>
7951         PR tree-optimization/86401
7952         * fold-const.c (fold_binary_loc) <case BIT_AND_EXPR>: Move the
7953         ((A & N) + B) & M -> (A + B) & M etc. optimization into ...
7954         (fold_bit_and_mask): ... here.  New helper function for match.pd.
7955         * fold-const.h (fold_bit_and_mask): Declare.
7956         * match.pd (((A & N) + B) & M -> (A + B) & M): New optimization.
7958 2018-07-06  Peter Bergner  <bergner@linux.ibm.com>
7960         PR target/86324
7961         * target.def (translate_mode_attribute): New hook.
7962         * targhooks.h (default_translate_mode_attribute): Declare.
7963         * targhooks.c (default_translate_mode_attribute): New function.
7964         * doc/tm.texi.in (TARGET_TRANSLATE_MODE_ATTRIBUTE): New hook.
7965         * doc/tm.texi: Regenerate.
7966         * config/rs6000/rs6000.c (TARGET_TRANSLATE_MODE_ATTRIBUTE): Define.
7967         (rs6000_translate_mode_attribute): New function.
7969 2018-07-06  Paul Koning  <ni1d@arrl.net>
7971         * doc/md.texi (define_split): Document DONE and FAIL.
7972         (define_peephole2): Ditto.
7974 2018-07-05  Jeff Law  <law@redhat.com>
7976         PR tree-optimization/86010
7977         * tree-ssa-dse.c (compute_trims): More aggressively trim at
7978         both the head and tail of mem* and str* calls.
7980 2018-07-05  Jim Wilson  <jimw@sifive.com>
7982         * config.gcc (riscv*-*-*): When setting xlen, handle riscv-*.
7984 2018-07-05  Indu Bhagat  <indu.bhagat@oracle.com>
7986         * config/aarch64/aarch64-simd.md: correct flags text for
7987         MIN_EXPR replacement.
7989 2018-07-05  James Clarke  <jrtc27@jrtc27.com>
7991         * configure: Regenerated.
7993 2018-07-05  Carl Love  <cel@us.ibm.com>
7995         * config/rs6000/rs6000-c.c: Map ALTIVEC_BUILTIN_VEC_UNPACKH for
7996         float argument to VSX_BUILTIN_DOUBLEH_V4SF.
7997         Map ALTIVEC_BUILTIN_VEC_UNPACKL for float argument to
7998         VSX_BUILTIN_DOUBLEL_V4SF.
8000 2018-07-05  Martin Sebor  <msebor@redhat.com>
8002         PR c++/86400
8003         * tree-ssa-strlen.c (maybe_set_strlen_range): Use type size rather
8004         than its domain to compute its the upper bound of a char array.
8006 2018-07-05  Nathan Sidwell  <nathan@acm.org>
8008         Replace NO_IMPLICIT_EXTERN_C with SYSTEM_IMPLICIT_EXTERN_C.
8009         * doc/cpp.texi: Update comment.
8010         * doc/tm.texi: Rebuilt.
8011         * doc/tm.texi.in (NO_IMPLICIT_EXTERN_C): Replace with ...
8012         (SYSTEM_IMPLICIT_EXTERN_C): ... this, opposite sense.
8013         * doc/extend.texi (Backwards Compatibility): Clarify it is system
8014         headers affected by extern "C".
8015         * system.h: Poison NO_IMPLICIT_EXTERN_C.
8016         * config/alpha/alpha.h, config/arm/uclinux-elf.h,
8017         config/bfin/elf.h, config/cris/cris.h, config/darwin.h,
8018         config/dragonfly.h, config/freebsd.h, config/gnu-user.h,
8019         config/i386/cygming.h, config/i386/djgpp.h, config/i386/nto.h,
8020         config/ia64/hpux.h, config/lm32/lm32.h, config/lm32/uclinux-elf.h,
8021         config/lynx.h, config/mips/elf.h, config/mmix/mmix.h,
8022         config/netbsd.h, config/pa/pa-hpux.h, config/powerpcspe/sysv4.h,
8023         config/riscv/elf.h, config/rs6000/sysv4.h, config/rtems.h,
8024         config/s390/tpf.h, config/sh/newlib.h, config/sol2.h,
8025         config/sparc/openbsd64.h, config/sparc/sp-elf.h,
8026         config/sparc/sp64-elf.h, config/spu/spu.h,
8027         config/stormy16/stormy16.h, config/v850/v850.h,
8028         config/visium/visium.h, config/vx-common.h, config/xtensa/elf.h: Don't
8029         define NO_IMPLICIT_EXTERN_C.
8030         * config/rs6000/aix.h: Set SYSTEM_IMPLICIT_EXTERN_C.
8032 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
8034         PR target/84711
8035         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8036         instead of GET_MODE_SIZE when comparing Units.
8038 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
8040         PR target/84711
8041         * rtlanal.c (set_noop_p): Constrain on mode change,
8042         include hard-reg-set.h
8044 2018-07-05  Tamar Christina  <tamar.christina@arm.com>
8046         * config/aarch64/aarch64.c (aarch64_expand_movmem): Fix mode size.
8048 2018-07-05  Jakub Jelinek  <jakub@redhat.com>
8050         Revert
8051         2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
8053         PR sanitizer/84250
8054         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8055         libasan.
8056         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8058 2018-07-04  Maxim Ostapenko  <m.ostapenko@samsung.com>
8060         PR sanitizer/84250
8061         * config/gnu-user.h (LIBASAN_EARLY_SPEC): Pass -lstdc++ for static
8062         libasan.
8063         * gcc.c: Do not pass LIBUBSAN_SPEC if ASan is enabled with UBSan.
8065 2018-07-04  Eric Botcazou  <ebotcazou@adacore.com>
8067         PR middle-end/86380
8068         * expmed.c (choose_multiplier): Fix incorrect comparison with mask.
8070 2018-07-04  Aldy Hernandez  <aldyh@redhat.com>
8072         * tree-vrp.c (extract_range_from_binary_expr_1): Initialize
8073         neg_*_op* variables.
8075 2018-07-04  Martin Liska  <mliska@suse.cz>
8077         * tree-switch-conversion.c: Define
8078         max_ratio_for_speed and max_ratio_for_size constants.
8080 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
8081             Martin Liska  <mliska@suse.cz>
8083         PR middle-end/66240
8084         PR target/45996
8085         PR c/84100
8086         * common.opt: Rename align options with 'str_' prefix.
8087         * common/config/i386/i386-common.c (set_malign_value): New
8088         function.
8089         (ix86_handle_option): Use it to set -falign-* options/
8090         * config/aarch64/aarch64-protos.h (struct tune_params): Change
8091         type from int to string.
8092         * config/aarch64/aarch64.c: Update default values from int
8093         to string.
8094         * config/alpha/alpha.c (alpha_override_options_after_change):
8095         Likewise.
8096         * config/arm/arm.c (arm_override_options_after_change_1): Likewise.
8097         * config/i386/dragonfly.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8098         max skip conditionally.
8099         * config/i386/freebsd.h (SUBALIGN_LOG): New.
8100         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8101         max skip conditionally.
8102         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8103         max skip conditionally.
8104         * config/i386/gnu-user.h (SUBALIGN_LOG): New.
8105         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8106         max skip conditionally.
8107         * config/i386/i386.c (struct ptt): Change type from int to
8108         string.
8109         (ix86_default_align): Set default values.
8110         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Print
8111         max skip conditionally.
8112         * config/i386/iamcu.h (SUBALIGN_LOG): New.
8113         (ASM_OUTPUT_MAX_SKIP_ALIGN):
8114         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN):
8115         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8116         max skip conditionally.
8117         * config/i386/openbsdelf.h (SUBALIGN_LOG): New.
8118         (ASM_OUTPUT_MAX_SKIP_ALIGN) Print max skip conditionally.:
8119         * config/i386/x86-64.h (SUBALIGN_LOG): New.
8120         (ASM_OUTPUT_MAX_SKIP_ALIGN): Print
8121         max skip conditionally.
8122         (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8123         * config/ia64/ia64.c (ia64_option_override): Set default values
8124         for alignment options.
8125         * config/m68k/m68k.c: Handle new str_align_* options.
8126         * config/mips/mips.c (mips_set_compression_mode): Change
8127         type of constants.
8128         (mips_option_override): Set default values for options.
8129         * config/powerpcspe/powerpcspe.c (rs6000_option_override_internal):
8130         Likewise.
8131         * config/rs6000/rs6000.c (rs6000_option_override_internal):
8132         Likewise.
8133         * config/rx/rx.c (rx_option_override): Likewise.
8134         * config/rx/rx.h (JUMP_ALIGN): Use align_jumps_log.
8135         (LABEL_ALIGN): Use align_labels_log.
8136         (LOOP_ALIGN): Use align_loops_align.
8137         * config/s390/s390.c (s390_asm_output_function_label): Use new
8138         macros.
8139         * config/sh/sh.c (sh_override_options_after_change):
8140         Change type of constants.
8141         * config/spu/spu.c (spu_sched_init): Likewise.
8142         * config/sparc/sparc.c (sparc_option_override): Set default
8143         values for options.
8144         * config/visium/visium.c (visium_option_override): Likewise.
8145         * config/visium/visium.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Do not
8146         emit p2align format with last argument if it's not needed.
8147         * doc/invoke.texi: Document extended format of -falign-*.
8148         * final.c: Use align_labels alignment.
8149         * flags.h (struct target_flag_state): Change type to use
8150         align_flags.
8151         (struct align_flags_tuple): New.
8152         (struct align_flags): Likewise.
8153         (align_loops_log): Redefine macro to use new types.
8154         (align_loops_max_skip): Redefine macro to use new types.
8155         (align_jumps_log): Redefine macro to use new types.
8156         (align_jumps_max_skip): Redefine macro to use new types.
8157         (align_labels_log): Redefine macro to use new types.
8158         (align_labels_max_skip): Redefine macro to use new types.
8159         (align_functions_log): Redefine macro to use new types.
8160         (align_loops): Redefine macro to use new types.
8161         (align_jumps): Redefine macro to use new types.
8162         (align_labels): Redefine macro to use new types.
8163         (align_functions): Redefine macro to use new types.
8164         (align_functions_max_skip): Redefine macro to use new types.
8165         (align_loops_value): New macro.
8166         (align_jumps_value): New macro.
8167         (align_labels_value): New macro.
8168         (align_functions_value): New macro.
8169         * function.c (invoke_set_current_function_hook): Propagate
8170         alignment values from flags to global variables default in
8171         topleev.h.
8172         * ipa-icf.c (sem_function::equals_wpa): Use
8173         cl_optimization_option_eq instead of memcmp.
8174         * lto-streamer.h (cl_optimization_stream_out): Support streaming
8175         of string types.
8176         (cl_optimization_stream_in): Likewise.
8177         * optc-save-gen.awk: Support strings in cl_optimization.
8178         * opth-gen.awk: Likewise.
8179         * opts.c (finish_options): Remove error checking of invalid
8180         value ranges.
8181         (MAX_CODE_ALIGN): Remove.
8182         (MAX_CODE_ALIGN_VALUE): Likewise.
8183         (parse_and_check_align_values): New function.
8184         (check_alignment_argument): Likewise.
8185         (common_handle_option): Use check_alignment_argument.
8186         * opts.h (parse_and_check_align_values): Declare.
8187         * toplev.c (init_alignments): Remove.
8188         (read_log_maxskip): New.
8189         (parse_N_M): Likewise.
8190         (parse_alignment_opts): Likewise.
8191         (backend_init_target): Remove usage of init_alignments.
8192         * toplev.h (parse_alignment_opts): Declare.
8193         * tree-streamer-in.c (streamer_read_tree_bitfields): Add new
8194         argument.
8195         * tree-streamer-out.c (streamer_write_tree_bitfields): Likewise.
8196         * tree.c (cl_option_hasher::equal): New.
8197         * varasm.c: Use new global macros.
8199 2018-07-04  Denys Vlasenko  <dvlasenk@redhat.com>
8201         * config/i386/dragonfly.h: (ASM_OUTPUT_MAX_SKIP_ALIGN):
8202         Use a simpler align directive also if MAXSKIP = ALIGN-1.
8203         * config/i386/gas.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8204         * config/i386/lynx.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8205         * config/i386/netbsd-elf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8206         * config/i386/i386.h (ASM_OUTPUT_MAX_SKIP_PAD): Likewise.
8207         * config/i386/freebsd.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Remove "If N
8208         is large, do at least 8 byte alignment" code. Add SUBALIGN_LOG
8209         define. Use a simpler align directive also if MAXSKIP = ALIGN-1.
8210         * config/i386/gnu-user.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8211         * config/i386/iamcu.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8212         * config/i386/openbsdelf.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8213         * config/i386/x86-64.h (ASM_OUTPUT_MAX_SKIP_ALIGN): Likewise.
8215 2018-07-04  Martin Liska  <mliska@suse.cz>
8216             Jonathan Wakely  <jwakely@redhat.com>
8218         * coverage.c: Use correct type.
8219         * doc/invoke.texi: Language correction.
8221 2018-07-03  H.J. Lu  <hongjiu.lu@intel.com>
8223         PR target/85620
8224         * config/i386/i386.c (rest_of_insert_endbranch): Also generate
8225         ENDBRANCH for non-tail call which may return via indirect branch.
8226         * doc/extend.texi: Document indirect_return attribute.
8228 2018-07-03  Martin Sebor  <msebor@redhat.com>
8230         PR tree-optimization/86274
8231         * gimple-ssa-sprintf.c (fmtresult::type_max_digits): Verify
8232         precondition.
8233         (format_floating): Correct handling of infinities and NaNs.
8235 2018-07-03  Martin Sebor  <msebor@redhat.com>
8237         * print-tree.c (print_real_cst): New function.
8238         (print_node_brief): Call it.
8239         (print_node): Ditto.
8241 2018-07-03  Jeff Law  <law@redhat.com>
8243         * config/h8300/h8300.md (logical<mode>3_sn, logical<mode>3): Merge
8244         into a single pattern.
8246         * config/h8300/h8300.md (ors code_iterator): New.
8247         (bsetqi_msx, bnotqi_msx patterns and splitters): Consolidate into
8248         a single pattern and single splitter.
8249         (bsethi_msx, bnothi_msx patterns): Consolidate into a single pattern.
8250         (iorqi3_1, xorqi3_1): Likewise.
8251         (iorqi3, xorqi3 expanders): Similarly.
8253         * config/h8300/h8300.md (movmd_internal_normal): Consolidated with
8254         (movmd_internal) into a single pattern using the P mode iterator.
8255         (movmd splitters): Similarly.
8256         (stpcpy_internal_normal, stpcpy_internal): Similarly for thes patterns.
8257         (movsd splitters): Similarly.
8259         * config/h8300/h8300.c (h8300_insn_length_from_table): Consolidate
8260         ADDB, ADDW and ADDL into a single ADD attribute which selects the
8261         right table based on the size of the operand.
8262         * config/h8300/h8300.md (length_table): Corresponding changes. All
8263         references to "addb", "addw" and "addl" changed to "add".
8264         (btst patterns): Merge two variants into a single pattern.
8265         (tstqi, tsthi): Likewise.
8266         (addhi3_incdec, addsi3_incdec): Likewise.
8267         (subhi3_h8300hs, subsi3_h8300hs): Likewise.
8268         (mulhi3, mulsi3): Likewise.
8269         (udivhi3, udivsi3): Likewise.
8270         (divhi3, divsi3): Likewise.
8271         (andorqi3, andorhi3, andorsi3): Likewise.
8273 2018-07-03  Uros Bizjak  <ubizjak@gmail.com>
8275         PR target/85694
8276         * config/i386/sse.md (uavg<mode>3_ceil): New expander.
8277         (<sse2_avx2>_uavg<mode>3<mask_name>): Simplify expander.
8279 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
8281         PR tree-optimization/85694
8282         * config/aarch64/iterators.md (HADD, RHADD): New int iterators.
8283         (u): Handle UNSPEC_SHADD, UNSPEC_UHADD, UNSPEC_SRHADD and
8284         UNSPEC_URHADD.
8285         * config/aarch64/aarch64-simd.md (<u>avg<mode>3_floor)
8286         (<u>avg<mode>3_ceil): New patterns.
8288 2018-07-03  David Malcolm  <dmalcolm@redhat.com>
8290         * gcc.dg/vect/slp-perm-1.c: Remove "note: " prefix from
8291         scan-tree-dump directive.
8292         * gcc.dg/vect/slp-perm-2.c: Likewise.
8293         * gcc.dg/vect/slp-perm-3.c: Likewise.
8294         * gcc.dg/vect/slp-perm-5.c: Likewise.
8295         * gcc.dg/vect/slp-perm-6.c: Likewise.
8296         * gcc.dg/vect/slp-perm-7.c: Likewise.
8297         * gcc.dg/vect/slp-perm-8.c: Likewise.
8299 2018-07-03  Marek Polacek  <polacek@redhat.com>
8301         PR middle-end/86202
8302         * gimple-fold.c (size_must_be_zero_p): Check the type of the size.
8304 2018-07-03  Richard Biener  <rguenther@suse.de>
8306         PR ipa/86389
8307         * tree-ssa-structalias.c (find_func_clobbers): Properly
8308         handle indirect calls.
8310 2018-07-03  Jeff Law  <law@redhat.com>
8312         * config/h8300/h8300.md (HSI, QHSI, QHSIF): New mode iterators.
8313         (shifts): New code iterator.
8314         (movqi, movhi, movsi, movsf expanders): Consolidate into a single
8315         expander.  Fix HImode handling on H8/SX.
8316         (addqi3, addhi3, addsi3 expanders): Consolidate into a single expander.
8317         (subqi3, subhi3, subsi3 expanders): Likewise.
8318         (andqi3, andhi3, andsi3 expanders): Likewise.
8319         (iorqi3, iorhi3, iorsi3 expanders): Likewise.
8320         (xorqi3, xorhi3, xorsi3 expanders): Likewise.
8321         (negqi2, neghi2, negsi2, negsf2 expanders): Likewise.
8322         (one_cmplqi2, one_cmplhi2, one_cmplsi2): Likewise.
8323         (zero_extendqihi2, zero_extendqisi2): Likewise.
8324         (extendqihi2, extendqisi2): Likewise.
8325         (rotlqi3, rotlhi3, rotlsi3): Likewise.
8326         (neghi2_h8300, negsi2_h8300): Likewise for these patterns.
8327         (rotlqi3_1, rotlhi3_1): Likewise.
8328         (logicalhi3_sn, logicalsi3_sn): Likewise.
8329         (logicalhi3, logicalsi3): Likewise.
8331 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
8333         * tree-vect-patterns.c (vect_recog_rotate_pattern)
8334         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
8335         (vect_recog_mixed_size_cond_pattern, adjust_bool_pattern_cast)
8336         (adjust_bool_pattern, vect_recog_bool_pattern): Pass the vector
8337         type to append_pattern_def_seq instead of creating a stmt_vec_info
8338         directly.
8339         (build_mask_conversion): Likewise.  Remove vinfo argument.
8340         (vect_add_conversion_to_patterm): Likewise, renaming to...
8341         (vect_add_conversion_to_pattern): ...this.
8342         (vect_recog_mask_conversion_pattern): Update call to
8343         build_mask_conversion.  Pass the vector type to
8344         append_pattern_def_seq here too.
8345         (vect_recog_gather_scatter_pattern): Update call to
8346         vect_add_conversion_to_pattern.
8348 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
8350         * tree-vect-patterns.c (new_pattern_def_seq): Delete.
8351         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8352         (vect_recog_widen_op_pattern, vect_recog_over_widening_pattern)
8353         (vect_recog_rotate_pattern, vect_synth_mult_by_constant): Don't set
8354         STMT_VINFO_PATTERN_DEF_SEQ to null here.
8355         (vect_recog_pow_pattern, vect_recog_vector_vector_shift_pattern)
8356         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern): Use
8357         append_pattern_def_seq instead of new_pattern_def_seq.
8358         (vect_recog_divmod_pattern): Do both of the above.
8359         (vect_pattern_recog_1): Assert that STMT_VINO_PATTERN_DEF_SEQ
8360         is null.
8362 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
8364         * tree-vect-patterns.c (vect_recog_dot_prod_pattern):
8365         (vect_recog_sad_pattern, vect_recog_widen_op_pattern)
8366         (vect_recog_widen_mult_pattern, vect_recog_pow_pattern):
8367         (vect_recog_widen_sum_pattern, vect_recog_over_widening_pattern)
8368         (vect_recog_average_pattern, vect_recog_cast_forwprop_pattern)
8369         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
8370         (vect_recog_vector_vector_shift_pattern, vect_synth_mult_by_constant)
8371         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
8372         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
8373         (vect_recog_mask_conversion_pattern): Replace vec<gimple *>
8374         parameter with a single stmt_vec_info.
8375         (vect_recog_func_ptr): Likewise.
8376         (vect_recog_gather_scatter_pattern): Likewise, folding in...
8377         (vect_try_gather_scatter_pattern): ...this.
8378         (vect_pattern_recog_1): Remove stmts_to_replace and just pass
8379         the stmt_vec_info of the statement to be matched.  Don't clear
8380         STMT_VINFO_RELATED_STMT.
8381         (vect_pattern_recog): Update call accordingly.
8383 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
8385         PR tree-optimization/85694
8386         * doc/md.texi (avgM3_floor, uavgM3_floor, avgM3_ceil)
8387         (uavgM3_ceil): Document new optabs.
8388         * doc/sourcebuild.texi (vect_avg_qi): Document new target selector.
8389         * internal-fn.def (IFN_AVG_FLOOR, IFN_AVG_CEIL): New internal
8390         functions.
8391         * optabs.def (savg_floor_optab, uavg_floor_optab, savg_ceil_optab)
8392         (savg_ceil_optab): New optabs.
8393         * tree-vect-patterns.c (vect_recog_average_pattern): New function.
8394         (vect_vect_recog_func_ptrs): Add it.
8395         * tree-vect-stmts.c (vectorizable_call): Get the type of the zero
8396         constant directly from the associated lhs.
8398 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
8400         * tree-vect-patterns.c (vect_split_statement): New function.
8401         (vect_convert_input): Use it to try to split an existing cast.
8403 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
8405         * poly-int.h (print_hex): New function.
8406         * dumpfile.h (dump_dec, dump_hex): Declare.
8407         * dumpfile.c (dump_dec, dump_hex): New poly_wide_int functions.
8408         * tree-vectorizer.h (_stmt_vec_info): Add min_output_precision,
8409         min_input_precision, operation_precision and operation_sign.
8410         * tree-vect-patterns.c (vect_get_range_info): New function.
8411         (vect_same_loop_or_bb_p, vect_single_imm_use)
8412         (vect_operation_fits_smaller_type): Delete.
8413         (vect_look_through_possible_promotion): Add an optional
8414         single_use_p parameter.
8415         (vect_recog_over_widening_pattern): Rewrite to use new
8416         stmt_vec_info infomration.  Handle one operation at a time.
8417         (vect_recog_cast_forwprop_pattern, vect_narrowable_type_p)
8418         (vect_truncatable_operation_p, vect_set_operation_type)
8419         (vect_set_min_input_precision): New functions.
8420         (vect_determine_min_output_precision_1): Likewise.
8421         (vect_determine_min_output_precision): Likewise.
8422         (vect_determine_precisions_from_range): Likewise.
8423         (vect_determine_precisions_from_users): Likewise.
8424         (vect_determine_stmt_precisions, vect_determine_precisions): Likewise.
8425         (vect_vect_recog_func_ptrs): Put over_widening first.
8426         Add cast_forwprop.
8427         (vect_pattern_recog): Call vect_determine_precisions.
8429 2018-07-03  Richard Sandiford  <richard.sandiford@arm.com>
8431         * tree-vect-patterns.c (vect_mark_pattern_stmts): Remove pattern
8432         statements that have been replaced by further pattern statements.
8433         (vect_pattern_recog_1): Clear STMT_VINFO_PATTERN_DEF_SEQ on failure.
8435 2018-07-03  Richard Biener  <rguenther@suse.de>
8437         * tree-vect-stmts.c (vect_is_simple_use): Consolidate dumping,
8438         always set *dt.  Dump vectype in vectype overload.
8439         * dumpfile.h (dump_gimple_expr): New function.
8440         (dump_gimple_expr_loc): Likewise.
8441         * dumpfile.c (dump_gimple_expr): New function.
8442         (dump_gimple_expr_loc): Likewise.
8444 2018-07-02  Jeff Law  <law@redhat.com>
8446         * config/h8300/h8300.md (movqi_h8300, movqi_h8300hs): Consolidate
8447         the H8/300, H8/300H and H8/S variants into a single pattern.
8448         (movhi_h8300, movqi_h8300hs): Similarly.
8449         (pushqi_h8300hs, pushhi_h8300hs): Consolidate into a single pattern.
8450         (QHI mode iterator): New.
8452         * config/h8300/h8300.md: Remove trailing whitespace.
8454 2018-07-02  Jim Wilson  <jimw@sifive.com>
8456         * config/riscv/riscv.c (riscv_expand_epilogue): Use emit_jump_insn
8457         instead of emit_insn for interrupt returns.
8458         * config/riscv/riscv.md (riscv_met): Add (return) to rtl.
8459         (riscv_sret, riscv_uret): Likewise.
8461 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
8463         * pretty-print.c (selftest::test_pp_format): Move save and restore
8464         of quotes to class auto_fix_quotes, and add an instance.
8465         * selftest.c: Include "intl.h".
8466         (selftest::auto_fix_quotes::auto_fix_quotes): New ctor.
8467         (selftest::auto_fix_quotes::~auto_fix_quotes): New dtor.
8468         * selftest.h (selftest::auto_fix_quotes): New class.
8470 2018-07-02  Richard Henderson  <richard.henderson@linaro.org>
8472         * config/aarch64/aarch64-protos.h, config/aarch64/aarch64.c
8473         (aarch64_sve_prepare_conditional_op): Remove.
8474         * config/aarch64/aarch64-sve.md (cond_<SVE_INT_BINARY><SVE_I>):
8475         Allow aarch64_simd_reg_or_zero as select operand; remove
8476         the aarch64_sve_prepare_conditional_op call.
8477         (cond_<SVE_INT_BINARY_SD><SVE_SDI>): Likewise.
8478         (cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
8479         (*cond_<SVE_INT_BINARY><SVE_I>_z): New pattern.
8480         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_z): New pattern.
8481         (*cond_<SVE_COND_FP_BINARY><SVE_F>_z): New pattern.
8482         (*cond_<SVE_INT_BINARY><SVE_I>_any): New pattern.
8483         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_any): New pattern.
8484         (*cond_<SVE_COND_FP_BINARY><SVE_F>_any): New pattern
8485         and a splitters to match all of the *_any patterns.
8486         * config/aarch64/predicates.md (aarch64_sve_any_binary_operator): New.
8488         * config/aarch64/iterators.md (SVE_INT_BINARY_REV): Remove.
8489         (SVE_COND_FP_BINARY_REV): Remove.
8490         (sve_int_op_rev, sve_fp_op_rev): New.
8491         * config/aarch64/aarch64-sve.md (*cond_<SVE_INT_BINARY><SVE_I>_0): New.
8492         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_0): New.
8493         (*cond_<SVE_COND_FP_BINARY><SVE_F>_0): New.
8494         (*cond_<SVE_INT_BINARY><SVE_I>_2): Rename, add movprfx alternative.
8495         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_2): Similarly.
8496         (*cond_<SVE_COND_FP_BINARY><SVE_F>_2): Similarly.
8497         (*cond_<SVE_INT_BINARY><SVE_I>_3): Similarly; use sve_int_op_rev.
8498         (*cond_<SVE_INT_BINARY_SD><SVE_SDI>_3): Similarly.
8499         (*cond_<SVE_COND_FP_BINARY><SVE_F>_3): Similarly; use sve_fp_op_rev.
8501         * config/aarch64/aarch64-sve.md (cond_<SVE_COND_FP_BINARY><SVE_F>):
8502         Remove match_dup 1 from the inner unspec.
8503         (*cond_<SVE_COND_FP_BINARY><SVE_F>): Likewise.
8505         * config/aarch64/aarch64.md (movprfx): New attr.
8506         (length): Default movprfx to 8.
8507         * config/aarch64/aarch64-sve.md (*mul<SVE_I>3): Add movprfx alt.
8508         (*madd<SVE_I>, *msub<SVE_I): Likewise.
8509         (*<su>mul<SVE_I>3_highpart): Likewise.
8510         (*<SVE_INT_BINARY_SD><SVE_SDI>3): Likewise.
8511         (*v<ASHIFT><SVE_I>3): Likewise.
8512         (*<su><MAXMIN><SVE_I>3): Likewise.
8513         (*<su><MAXMIN><SVE_F>3): Likewise.
8514         (*fma<SVE_F>4, *fnma<SVE_F>4): Likewise.
8515         (*fms<SVE_F>4, *fnms<SVE_F>4): Likewise.
8516         (*div<SVE_F>4): Likewise.
8518 2018-07-02  Richard Sandiford  <richard.sandiford@arm.com>
8520         * tree-vect-patterns.c (vect_recog_widen_shift_pattern): Fix typo
8521         in dump string.
8523 2018-07-02  Richard Biener  <rguenther@suse.de>
8525         PR tree-optimization/86363
8526         * tree-ssa-sccvn.c (vn_reference_lookup_3): Check the
8527         memset argument refers to a non-variable address.
8529 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
8531         * tree-vrp.c (extract_range_from_binary_expr_1): Abstract a lot of the
8532         {PLUS,MINUS}_EXPR code to...
8533         (adjust_symbolic_bound): ...here,
8534         (combine_bound): ...here,
8535         (set_value_range_with_overflow): ...and here.
8537 2018-07-02  Aldy Hernandez  <aldyh@redhat.com>
8539         * tree-vrp.c (extract_range_from_unary_expr): Abstract ABS_EXPR
8540         code...
8541         (extract_range_from_abs_expr): ...here.
8543 2018-07-02  Eric Botcazou  <ebotcazou@adacore.com>
8545         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Do not overrule
8546         -fno-omit-frame-pointer when not optimizing.
8548 2018-07-02  Martin Liska  <mliska@suse.cz>
8550         PR ipa/86279
8551         * ipa-pure-const.c (malloc_candidate_p): Revert usage of ::get.
8552         (propagate_nothrow): Likewise.
8554 2018-07-02  Martin Liska  <mliska@suse.cz>
8556         PR ipa/86323
8557         * ipa-inline.c (early_inliner): Revert wrongly added ::get call.
8559 2018-07-02  David Malcolm  <dmalcolm@redhat.com>
8561         * dumpfile.c (dump_generic_expr_loc): Undo removal of this
8562         function in r262149, changing "loc" param from source_location to
8563         const dump_location_t &.
8564         * dumpfile.h (dump_generic_expr_loc): Undo removal of this
8565         declaration, as above.
8567 2018-07-01  Paul Koning  <ni1d@arrl.net>
8569         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
8570         -munit-asm, -mgnu-asm, -mdec-asm.
8571         * config/pdp11/pdp11-protos.h (pdp11_gen_int_label): New.
8572         (pdp11_output_labelref): New.
8573         (pdp11_output_def): New.
8574         (pdp11_output_addr_vec_elt): New.
8575         * config/pdp11/pdp11.c: Use tab between opcode and operands.  Use
8576         %# and %@ format codes.
8577         (pdp11_option_override): New.
8578         (TARGET_ASM_FILE_START_FILE_DIRECTIVE): Define.
8579         (pdp11_output_ident): New.
8580         (pdp11_asm_named_section): New.
8581         (pdp11_asm_init_sections): New.
8582         (pdp11_file_start): New.
8583         (pdp11_file_end): New.
8584         (output_ascii): Use .ascii/.asciz for -mdec-asm.
8585         (pdp11_asm_print_operand): Update %# and %$ for -mdec-asm.  Add
8586         %o, like %c but octal.
8587         (pdp11_option_override): New.
8588         * config/pdp11/pdp11.h (TEXT_SECTION_ASM_OP): Update for
8589         -mdec-asm.
8590         (DATA_SECTION_ASM_OP): Ditto.
8591         (READONLY_DATA_SECTION_ASM_OP): New.
8592         (IS_ASM_LOGICAL_LINE_SEPARATOR): New.
8593         (ASM_GENERATE_INTERNAL_LABEL): Use new function.
8594         (ASM_OUTPUT_LABELREF): Ditto.
8595         (ASM_OUTPUT_DEF): Ditto.
8596         (ASM_OUTPUT_EXTERNAL): New.
8597         (ASM_OUTPUT_SOURCE_FILENAME): New.
8598         (ASM_OUTPUT_ADDR_VEC_ELT): Use new function.
8599         (ASM_OUTPUT_SKIP): Update for -mdec-asm.
8600         * config/pdp11/pdp11.md: Use tab between opcode and operands.  Use
8601         %# and %@ format codes.
8602         * config/pdp11/pdp11.opt (mgnu-asm): New.
8603         (mdec-asm): Conflicts with -mgnu-asm and -munix-asm.
8604         (munix-asm): Conflicts with -mdec-asm and -mgnu-asm.
8605         * doc/invoke.txt (PDP-11 Options): Add -mgnu-asm.
8607 2018-07-01  Aldy Hernandez  <aldyh@redhat.com>
8609         * tree-ssa-threadupdate.c (mark_threaded_blocks): Avoid
8610         dereferencing path[] beyond its length.
8611         (debug_path): New.
8612         (debug_all_paths): New.
8613         (rewire_first_differing_edge): New.
8614         (adjust_paths_after_duplication): New.
8615         (duplicate_thread_path): Call adjust_paths_after_duplication.
8616         Add new argument.
8617         (thread_through_all_blocks): Add new argument to
8618         duplicate_thread_path.
8620 2018-06-30  Jim Wilson  <jimw@sifive.com>
8622         * config/riscv/predicates.md (p2m1_shift_operand): New.
8623         (high_mask_shift_operand): New.
8624         * config/riscv/riscv.md (lshrsi3_zero_extend_3+1): New combiner
8625         pattern using p2m1_shift_operand.
8626         (lshsi3_zero_extend_3+2): New combiner pattern using
8627         high_mask_shift_operand.
8629 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
8631         * tree-vect-patterns.c (vect_get_external_def_edge): New function,
8632         split out from...
8633         (vect_recog_rotate_pattern): ...here.
8634         (vect_convert_input): Try to insert casts of invariants in the
8635         preheader.
8636         * tree-vect-loop-manip.c (vect_loop_versioning): Don't require the
8637         preheader to be empty.
8639 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
8641         * tree-vect-patterns.c (append_pattern_def_seq): Take an optional
8642         vector type.  If given, install it in the new statement's
8643         STMT_VINFO_VECTYPE.
8644         (vect_element_precision): New function.
8645         (vect_unpromoted_value): New struct.
8646         (vect_unpromoted_value::vect_unpromoted_value): New function.
8647         (vect_unpromoted_value::set_op): Likewise.
8648         (vect_look_through_possible_promotion): Likewise.
8649         (vect_joust_widened_integer, vect_joust_widened_type): Likewise.
8650         (vect_widened_op_tree, vect_convert_input): Likewise.
8651         (vect_convert_inputs, vect_convert_output): Likewise.
8652         (vect_recog_dot_prod_pattern): Use vect_look_through_possible_promotion
8653         to handle the optional cast of the multiplication result and
8654         vect_widened_op_tree to detect the widened multiplication itself.
8655         Do not require the input and output of promotion casts to have
8656         the same sign, but base the signedness of the operation on the
8657         input rather than the result.  If the pattern includes two
8658         promotions, check that those promotions have the same sign.
8659         Do not restrict the MULT_EXPR handling to a double-width result;
8660         handle quadruple-width results and wider.  Use vect_convert_inputs
8661         to convert the inputs to the common type.
8662         (vect_recog_sad_pattern):  Use vect_look_through_possible_promotion
8663         to handle the optional cast of the ABS result.  Also allow a sign
8664         change or a sign extension between the ABS and MINUS.
8665         Use vect_widened_op_tree to detect the widened subtraction and use
8666         vect_convert_inputs to convert the inputs to the common type.
8667         (vect_handle_widen_op_by_const): Delete.
8668         (vect_recog_widen_op_pattern): New function.
8669         (vect_recog_widen_mult_pattern): Use it.
8670         (vect_recog_widen_shift_pattern): Likewise.
8671         (vect_recog_widen_sum_pattern): Use
8672         vect_look_through_possible_promotion to handle the promoted
8673         PLUS_EXPR operand.
8675 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
8677         * gimple-iterator.c (gsi_for_stmt): Add a new overload that takes
8678         the containing gimple_seq *.
8679         * gimple-iterator.h (gsi_for_stmt): Declare it.
8680         * tree-vect-patterns.c (vect_recog_dot_prod_pattern)
8681         (vect_recog_sad_pattern, vect_recog_widen_sum_pattern)
8682         (vect_recog_widen_shift_pattern, vect_recog_rotate_pattern)
8683         (vect_recog_vector_vector_shift_pattern, vect_recog_divmod_pattern)
8684         (vect_recog_mask_conversion_pattern): Remove STMT_VINFO_IN_PATTERN_P
8685         checks.
8686         (vect_init_pattern_stmt, vect_set_pattern_stmt): New functions,
8687         split out from...
8688         (vect_mark_pattern_stmts): ...here.  Handle cases in which the
8689         statement being replaced is part of an existing pattern
8690         definition sequence, inserting the new pattern statements before
8691         the original one.
8692         (vect_pattern_recog_1): Don't return a bool.  If the statement
8693         is already part of a pattern, instead apply pattern matching
8694         to the pattern definition statements.  Don't clear the
8695         STMT_VINFO_RELATED_STMT if is_pattern_stmt_p.
8696         (vect_pattern_recog): Don't break after the first match;
8697         continue processing the pattern definition statements instead.
8698         Don't bail out for STMT_VINFO_IN_PATTERN_P here.
8700 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
8702         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
8703         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
8704         (vect_recog_widen_sum_pattern): Use it.
8706 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
8708         * tree-vect-loop.c (vectorizable_reduction): Assert that the
8709         phi is not a pattern statement and has not been replaced by
8710         a pattern statement.
8711         * tree-vect-patterns.c (type_conversion_p): Don't check
8712         STMT_VINFO_IN_PATTERN_P.
8713         (vect_recog_vector_vector_shift_pattern): Likewise.
8714         (vect_recog_dot_prod_pattern): Expect vect_is_simple_use to return
8715         the pattern statement rather than the original statement; check
8716         directly for a WIDEN_MULT_EXPR here.
8717         * tree-vect-slp.c (vect_get_and_check_slp_defs): Expect
8718         vect_is_simple_use to return the pattern statement rather
8719         than the original statement; use is_pattern_stmt_p to check
8720         for such a pattern statement.
8721         * tree-vect-stmts.c (process_use): Expect vect_is_simple_use
8722         to return the pattern statement rather than the original statement;
8723         don't do the same transformation here.
8724         (vect_is_simple_use): If the defining statement has been replaced
8725         by a pattern statement, return the pattern statement instead.
8726         Remove the corresponding (local) transformation from the vectype
8727         overload.
8729 2018-06-30  Richard Sandiford  <richard.sandiford@arm.com>
8731         * tree-vectorizer.h (vect_is_simple_use): Move the gimple ** to the
8732         end and default to null.
8733         * tree-vect-loop.c (vect_create_epilog_for_reduction)
8734         (vectorizable_reduction): Update calls accordingly, dropping the
8735         gimple ** argument if the passed-back statement isn't needed.
8736         * tree-vect-patterns.c (vect_get_internal_def, type_conversion_p)
8737         (vect_recog_rotate_pattern): Likewise.
8738         (vect_recog_mask_conversion_pattern): Likewise.
8739         * tree-vect-slp.c (vect_get_and_check_slp_defs): Likewise.
8740         (vect_mask_constant_operand_p): Likewise.
8741         * tree-vect-stmts.c (is_simple_and_all_uses_invariant, process_use):
8742         (vect_model_simple_cost, vect_get_vec_def_for_operand): Likewise.
8743         (get_group_load_store_type, get_load_store_type): Likewise.
8744         (vect_check_load_store_mask, vect_check_store_rhs): Likewise.
8745         (vectorizable_call, vectorizable_simd_clone_call): Likewise.
8746         (vectorizable_conversion, vectorizable_assignment): Likewise.
8747         (vectorizable_shift, vectorizable_operation): Likewise.
8748         (vectorizable_store, vect_is_simple_cond): Likewise.
8749         (vectorizable_condition, vectorizable_comparison): Likewise.
8750         (get_same_sized_vectype, vect_get_mask_type_for_stmt): Likewise.
8751         (vect_is_simple_use): Rename the def_stmt argument to def_stmt_out
8752         and move it to the end.  Cope with null def_stmt_outs.
8754 2018-06-30  Bernd Edlinger  <bernd.edlinger@hotmail.de>
8756         * Makefile.in (FLAGS_TO_PASS): Add RANLIB_FOR_TARGET.
8758 2018-06-29  Jeff Law  <law@redhat.com>
8760         * config/v850/v850.c (v850_legitimate_address_p): Handle large
8761         displacements for TARGET_V850E2V3 and newer.
8762         (TARGET_LRA_P): Remove.  Defaults to LRA now.
8763         * config/v850/v850.md (sign23byte_load): Remove.
8764         (unsign23byte_load, sign23hword_load, unsign23hword_load): Likewise.
8765         (23word_load, 23byte_store, 23hword_store, 23word_store): Likewise.
8767 2018-06-29  Martin Liska  <mliska@suse.cz>
8769         PR lto/85759
8770         * coverage.c (coverage_init): Mangle full path name.
8771         * doc/invoke.texi: Document the change.
8772         * gcov-io.c (mangle_path): New.
8773         * gcov-io.h (mangle_path): Likewise.
8774         * gcov.c (mangle_name): Use mangle_path for path mangling.
8776 2018-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8778         * config/arm/arm.c (output_move_double): Don't allow STRD instructions
8779         if starting source register is not even.
8781 2018-06-29  Martin Liska  <mliska@suse.cz>
8783         PR tree-optimization/86263
8784         * tree-switch-conversion.c (switch_decision_tree::try_switch_expansion):
8785         Make edge redirection.
8787 2018-06-29  David Malcolm  <dmalcolm@redhat.com>
8789         * dumpfile.c (dump_loc): Add indentation based on scope depth.
8790         (dump_scope_depth): New variable.
8791         (get_dump_scope_depth): New function.
8792         (dump_begin_scope): New function.
8793         (dump_end_scope): New function.
8794         * dumpfile.h (get_dump_scope_depth): New declaration.
8795         (dump_begin_scope): New declaration.
8796         (dump_end_scope): New declaration.
8797         (class auto_dump_scope): New class.
8798         (AUTO_DUMP_SCOPE): New macro.
8799         * tree-vectorizer.h (DUMP_VECT_SCOPE): Reimplement in terms of
8800         AUTO_DUMP_SCOPE.
8802 2018-06-29  Richard Biener  <rguenther@suse.de>
8804         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Assert
8805         compute_all_dependences succeeds.
8806         * tree-vect-loop.c (vect_get_datarefs_in_loop): Fail early if we
8807         exceed --param loop-max-datarefs-for-datadeps.
8809 2018-06-29  Jakub Jelinek  <jakub@redhat.com>
8811         * config/rs6000/t-rs6000: Append rs6000-modes.h to TM_H.
8813 2018-06-28  Uros Bizjak  <ubizjak@gmail.com>
8815         PR target/86348
8816         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): Use
8817         alternative 0 in preferred_for_speed attribute.
8819 2018-06-28  Paul Koning  <ni1d@arrl.net>
8821         * config/pdp11/pdp11-protos.h (pdp11_shift_length): New function.
8822         * config/pdp11/pdp11.c (pdp11_shift_length): New function.
8823         * config/pdp11/pdp11.h (ADJUST_INSN_LENGTH): Remove.
8824         * config/pdp11/pdp11.md: Correct "length" attribute calculation
8825         for shift insn patterns.
8827 2018-06-28  David Malcolm  <dmalcolm@redhat.com>
8829         * cgraph.c (cgraph_node::get_body): Replace assignments to
8830         "dump_file" with calls to set_dump_file.
8831         * dumpfile.c (alt_dump_file): Make static, and group with...
8832         (alt_flags): ...this definition.
8833         (dumps_are_enabled): New variable.
8834         (refresh_dumps_are_enabled): New function.
8835         (set_dump_file): New function.
8836         (set_alt_dump_file): New function.
8837         (gcc::dump_manager::dump_start): Replace assignments to
8838         "dump_file" and "alt_dump_file" with calls to set_dump_file and
8839         set_alt_dump_file.
8840         (gcc::dump_manager::dump_finish): Likewise.
8841         * dumpfile.h (alt_dump_file): Delete decl.
8842         (dumps_are_enabled): New variable decl.
8843         (set_dump_file): New function decl.
8844         (dump_enabled_p): Rewrite in terms of new "dumps_are_enabled"
8845         global.
8846         * tree-nested.c (lower_nested_functions): Replace assignments to
8847         "dump_file" with calls to set_dump_file.
8849 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
8851         * tree-cfg.c (verify_gimple_in_cfg): Call verify_location on the
8852         goto_locus of each outgoing edge of each basic block.
8854 2018-06-28  Richard Biener  <rguenther@suse.de>
8856         * dwarf2out.c (decl_scope_table): Remove.
8857         (push_decl_scope): Likewise.
8858         (pop_decl_scope): Likewise.
8859         (gen_type_die_for_member): Do not call push/pop_decl_scope.
8860         (gen_struct_or_union_type_die): Likewise.
8861         (gen_tagged_type_die): Likewise.
8862         (dwarf2out_init): Do not initialize decl_scope_table.
8863         (dwarf2out_c_finalize): Do not free it.
8865 2018-06-28  Richard Biener  <rguenther@suse.de>
8867         * dwarf2out.c (gen_subprogram_die): Use is_unit_die when
8868         deciding whether to not re-use a DIE.
8870 2018-06-28  Richard Biener  <rguenther@suse.de>
8872         * dwarf2out.c (gen_subprogram_die): Always re-use DIEs with an
8873         DW_AT_abstract_origin attribute.
8875 2018-06-28  Martin Liska  <mliska@suse.cz>
8877         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
8878         Use newly introduced constants.
8879         * tree-switch-conversion.h (struct jump_table_cluster):
8880         Define max_ratio_for_size and max_ratio_for_speed.
8882 2018-06-28  Martin Liska  <mliska@suse.cz>
8884         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
8885         Add new checking assert to catch invalid state.
8886         (jump_table_cluster::can_be_handled): Handle single case
8887         clusters.
8888         (jump_table_cluster::is_beneficial): Bail out for such case.
8889         (bit_test_cluster::find_bit_tests):
8890         Add new checking assert to catch invalid state.
8891         (bit_test_cluster::can_be_handled): Handle single case
8892         clusters.
8893         (bit_test_cluster::is_beneficial): Bail out for such case.
8894         (switch_decision_tree::analyze_switch_statement):
8895         Fix comment.
8897 2018-06-28  Martin Liska  <mliska@suse.cz>
8899         * common.opt: Introduce -completion option.
8900         * gcc.c (driver_handle_option): Handle it.
8901         (driver::main): Print completions if completion
8902         is set.
8903         * opt-suggestions.c (option_proposer::get_completions):
8904         New function.
8905         (option_proposer::suggest_completion): Likewise.
8906         (option_proposer::find_param_completions): Likewise.
8907         (verify_autocompletions): Likewise.
8908         (test_completion_valid_options): Likewise.
8909         (test_completion_valid_params): Likewise.
8910         (in_completion_p): Likewise.
8911         (empty_completion_p): Likewise.
8912         (test_completion_partial_match): Likewise.
8913         (test_completion_garbage): Likewise.
8914         (opt_proposer_c_tests): Likewise.
8915         * opt-suggestions.h: Declare new functions.
8916         * opts.c (common_handle_option): Handle OPT__completion_.
8917         * selftest-run-tests.c (selftest::run_tests): Add
8918         opt_proposer_c_tests.
8919         * selftest.c (assert_str_startswith): New.
8920         * selftest.h (assert_str_startswith): Likewise.
8921         (opt_proposer_c_tests): New.
8922         (ASSERT_STR_STARTSWITH): Likewise.
8924 2018-06-28  Martin Liska  <mliska@suse.cz>
8926         * Makefile.in: Add opt-suggestions.o.
8927         * gcc-main.c: Include opt-suggestions.h.
8928         * gcc.c (driver::driver): Likewise.
8929         (driver::~driver): Remove m_option_suggestions.
8930         (driver::build_option_suggestions): Moved to option_proposer.
8931         (driver::suggest_option): Likewise.
8932         (driver::handle_unrecognized_options): Use option_proposer.
8933         * gcc.h (class driver): Add new memver m_option_proposer.
8934         * opt-suggestions.c: New file.
8935         * opt-suggestions.h: New file.
8937 2018-06-28  Martin Liska  <mliska@suse.cz>
8939         * vec.h (class auto_string_vec): New (moved from auto_argvec).
8940         (auto_string_vec::~auto_string_vec): Likewise.
8942 2018-06-28  Eric Botcazou  <ebotcazou@adacore.com>
8944         * tree-inline.h (struct copy_body_data): Move remapping_type_depth and
8945         prevent_decl_creation_for_types fields up and add reset_location field.
8946         * tree-inline.c (remap_gimple_stmt): Force input_location on the new
8947         statement if id->reset_location is true.
8948         (copy_edges_for_bb): Do not set goto_locus on the new edges if
8949         id->reset_location is true.
8950         (copy_phis_for_bb): Force input_location on the arguments if
8951         id->reset_location is true.
8952         (expand_call_inline): Set id->reset_location if DECL_IGNORED_P
8953         is set on the function to be inlined.
8955 2018-06-27  Stephan Bergmann  <sbergman@redhat.com>
8957         * doc/invoke.texi (Debugging Options): Add -gsplit-dwarf.
8959 2018-06-27  Dimitar Dimitrov  <dimitar@dinux.eu>
8961         * lra-eliminations.c (update_reg_eliminate): Mark all spanning hard
8962         registers for Pmode.
8963         * lra-lives.c (check_pseudos_live_through_calls): Mark all spanning
8964         hard registers for the clobbered pseudo.
8966 2018-06-27  Paul Koning  <ni1d@arrl.net>
8968         * common/config/pdp11/pdp11-common.c (pdp11_handle_option): Handle
8969         mutually exclusive options.
8970         * config/pdp11/constraints.md (h): New constraint.
8971         (O): Update definition to match shift code generation.
8972         (D): New constraint.
8973         * config/pdp11/pdp11-modes.def (CCNZ): Define mode.
8974         (CCFP): Remove.
8975         * config/pdp11/pdp11-protos.h (int_no_side_effect_operand): New
8976         function.
8977         (output_jump): Change arguments.
8978         (pdp11_fixed_cc_regs): New function.
8979         (pdp11_cc_mode): Ditto.
8980         (pdp11_expand_shift): Ditto.
8981         (pdp11_assemble_shift): Ditto.
8982         (pdp11_small_shift): Ditto.
8983         (pdp11_branch_cost): Remove.
8984         * config/pdp11/pdp11.c (pdp11_assemble_integer): Remove comments
8985         from output.
8986         (pdp11_register_move_cost): Update for CC registers.
8987         (pdp11_rtx_costs): Add case for LSHIFTRT.
8988         (pdp11_output_jump): Add CCNZ mode conditional branches.
8989         (notice_update_cc_on_set): Remove.
8990         (pdp11_cc_mode): New function.
8991         (simple_memory_operand): Correct pre/post decrement case.
8992         (no_side_effect_operand): New function.
8993         (pdp11_regno_reg_class): Add CC_REGS class.
8994         (pdp11_fixed_cc_regs): New function.
8995         (pdp11_small_shift): New function.
8996         (pdp11_expand_shift): New function to expand shift insns.
8997         (pdp11_assemble_shift): New function to output shifts.
8998         (pdp11_branch_cost): Remove.
8999         (pdp11_modes_tieable_p): Make QI/HI modes tieable.
9000         * config/pdp11/pdp11.h (SIZE_TYPE): Ensure 16-bit type.
9001         (WCHAR_TYPE): Ditto.
9002         (PTRDIFF_TYPE): Ditto.
9003         (ADJUST_INSN_LENGTH): New macro.
9004         (FIXED_REGISTERS): Add CC registers.
9005         (CALL_USED_REGISTERS): Ditto.
9006         (reg_class): Ditto.
9007         (REG_CLASS_NAMES): Ditto.
9008         (REG_CLASS_CONTENTS): Ditto.
9009         (SELECT_CC_MODE): Use new function.
9010         (TARGET_FLAGS_REGNUM): New macro.
9011         (TARGET_FIXED_CONDITION_CODE_REGS): Ditto.
9012         (cc0_reg_rtx): Remove.
9013         (CC_STATUS_MDEP): Remove.
9014         (CC_STATUS_MDEFP_INIT): Remove.
9015         (CC_IN_FPU): Remove.
9016         (NOTICE_UPDATE_CC): Remove.
9017         (REGISTER_NAMES): Add CC registers.
9018         (BRANCH_COST): Change to constant 1.
9019         * config/pdp11/pdp11.md: Rewrite for CCmode condition code
9020         handling.
9021         * config/pdp11/pdp11.opt (mbcopy): Remove.
9022         (mbcopy-builtin): Remove.
9023         (mbranch-cheap): Remove.
9024         (mbranch-expensive): Remove.
9025         * config/pdp11/predicates.md (expand_shift_operand): Update to
9026         match shift code generation.
9027         (ccnz_operator): New predicate.
9028         * doc/invoke.texi (PDP-11 Options): Remove deleted options
9029         -mbcopy, -mbcopy-builtin, -mbranch-cheap, -mbranch-expensive.
9030         Remove non-existent option -mabshi, -mno-abshi.  Document mutually
9031         exclusive options.
9032         * doc/md.texi (PDP-11): Document new D and h constraints.  Update
9033         description of O constraint.
9035 2018-06-27  Jeff Law  <law@redhat.com>
9036             Austin Law  <austinklaw@gmail.com>
9038         * config/v850/v850.md (addsi3_set_flags): New pattern.
9039         (subsi3_set_flags, negsi2_set_flags, andsi3_set_flags): Likewise.
9040         (iorsi3_set_flags, xorsi3_set_flags, one_cmplsi2_set_flags): Likewise.
9041         (zero_extendhisi2_v850_set_flags): Likewise.
9042         (zero_extendqisi2_v850_set_flags): Likewise.
9043         (ashlsi3_set_flags, ashlsi3_v850e2_set_flags): Likewise.
9044         (lshrsi3_set_flags, lshrsi3_v850e2_set_flags): Likewise.
9045         (ashrsi3_set_flags, ashrsi3_v850e2_set_flags): Likewise.
9047         * config/v850/v850-protos.h (notice_update_cc): Remove.
9048         * config/v850/v850.c (v850_compare_op0, v850_compare_op1): Remove.
9049         (v850_print_operand): Handle 'D' and "d".
9050         (v850_select_cc_mode): Remove ATTRIBUTE_UNUSED for last argument.
9051         Add handling of arithmetic/logical operations compared against zero.
9052         (v850_gen_float_compare): Remove ATTRIBUTE_UNUSED for last argument.
9053         Do not look at v850_compare_op, instead get mode from last argument.
9054         (v850_gen_compare): Remove
9055         (increment_stack): Use addsi3_clobber_flags to avoid splitting failure
9056         after reload for prologue insns.
9057         (expand_prologue): Account for CLOBBER of CC_REGNUM in various
9058         patterns.
9059         (construct_save_jarl): Likewise.
9060         (TARGET_FLAGS_REGNUM): Define.
9061         * config/v850/v850.h (v850_compare_op0, v850_compare_op1): Remove.
9062         (NOTICE_UPDATE_CC): Remove.
9063         * config/v850/v850.md (v850_tst1): Use (reg:CCZ CC_REGNUM) rather
9064         than cc0.  Conditionalize on reload_completed.
9065         (cmpsi_insn, setfcc_insn): Likewise.
9066         (tst1 splitter): Turn into define_and_split which sets the flags
9067         after reload.
9068         (cstoresi4, cbranchsf4, cbranchdf4, cbranchsi4_insn): Likewise.
9069         (cbranchsi4, branch_normal, branch_invert): Do not expose cc0 here.
9070         (cstoresf4, cstoredf4): Clobber the flags.
9071         (cmpsi, cmpsf, cmpdf): Remove expanders.
9072         (setf_insn): Remove pattern.
9073         (addsi3): Turn into define_and_split which clobbers the flags after
9074         reload and a suitable pattern (addsi3_clobber_flags) for use after
9075         reload.
9076         (subsi3, negsi2, andsi3, iorsi3, xorsi3, one_cmplsi2) Likewise.
9077         (ashlsi3, ashlsi3_v850e2, lshrsi3, lsh4si3_v850e2): Likewise.
9078         (ashrsi3, ashrsi3_v850e2): Likewise.
9079         (bins): Clobber the flags.
9080         (movsicc_normal_cc, movsicc_normal, movsicc_tst1): Likewise.
9081         (movsicc_tst1_revesed, sasf, swap and rotate patterns): Likewise.
9082         (fix_loop_counter, call_internal_short, call_internal_long): Likewise.
9083         (call_value_internal_short, call_value_internal_long): Likewise.
9084         (callt_save_interrupt, callt_return_interrupt): Likewise.
9085         (save_interrupt, return_interrupt): Likewise.
9086         (callt_save_all_interrupt, save_all_interrupt): Likewise.
9087         (_save_all_interrupt, callt_restore_all_interrupt): Likewise.
9088         (restore_all_interrupt, _restore_all_interrupt): Likewise.
9089         (All FP comparisons): Only allow after reload has completed.
9090         (trfsr): Likewise.
9091         (divh, divhu): Tweak output template.
9092         (branch_z_normal, branch_z_invert): Remove
9093         (branch_nz_normal, branch_nz_invert): Likewise.
9094         (extendhisi_insn, extendqisi_insn): Do not clobber flags.
9096         * config/v850/v850-modes.def (CCZ, CCNZ): Add new modes.
9097         * config/v850/v850.c (notice_update_cc): Remove.
9098         * config/v850/v850.h  (CC_OVERFLOW_UNUSABLE): Remove
9099         (CC_NO_CARRY): Likewise.
9100         (NOTICE_UPDATE_CC): Define to nothing.
9101         * config/v850/v850.md: Remove block comment on cc0 handling
9102         Remove "cc" attribute from all patterns.  Remove cc_status handling
9103         from all patterns.  Minor formatting fixes.
9105 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9107         * config/aarch64/aarch64-cores.def (cortex-a76): New entry.
9108         (cortex-a76.cortex-a55): Likewise.
9109         * config/aarch64/aarch64-tune.md: Regenerate.
9110         * doc/invoke.texi (AArch64 Options): Document cortex-a76 and
9111         cortex-a76.cortex-a55.
9113 2018-06-27  Jeff Law  <law@redhat.com>
9115         * config/v850/t-v850 (MULTILIB_OPTIONS): Remove 8byte-align.
9116         (MULTILIB_DIRNAMES): Similarly.
9118 2018-06-27  Eric Botcazou  <ebotcazou@adacore.com>
9120         * gimple.h (gimple_return_retbnd): Delete.
9121         (gimple_return_set_retbnd): Likewise.
9122         * cgraphunit.c (cgraph_node::expand_thunk): Remove call to
9123         gimple_return_set_retbnd.
9124         * gimple-pretty-print.c (dump_gimple_return): Remove call to
9125         gimple_return_retbnd and adjust.
9126         * tree-inline.h (struct copy_body_data): Remove retbnd field.
9127         * tree-inline.c (remap_gimple_stmt): Remove handling of retbnd.
9128         Explicitly return NULL in a couple more cases.  Move assertion
9129         on debug statements and remove unreachable code.
9130         (reset_debug_binding): Do not test id->retbnd.
9131         (expand_call_inline): Do not set it.
9133 2018-06-27  Rasmus Villemoes  <rv@rasmusvillemoes.dk>
9135         * configure.ac: Add --disable-gcov option.
9136         * configure: Regenerate.
9137         * Makefile.in: Honour @enable_gcov@.
9138         * doc/install.texi: Document --disable-gcov.
9140 2018-06-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9142         * config/arm/arm-cpus.in (cortex-a76): New entry.
9143         (cortex-a76.cortex-a55): Likewise.
9144         * config/arm/arm-tables.opt: Regenerate.
9145         * config/arm/arm-tune.md: Likewise.
9146         * config/arm/driver-arm.c (arm_cpu_table): Add Cortex-A76 entry.
9147         * doc/invoke.texi (ARM Options): Document cortex-a76 and
9148         cortex-a76.cortex-a55.
9150 2018-06-27  Tamar Christina  <tamar.christina@arm.com>
9152         PR target/85769
9153         * config/aarch64/aarch64.md (*movhf_aarch64): Add dup v0.4h pattern.
9155 2018-06-27  Siddhesh Poyarekar  <siddhesh@sourceware.org>
9157         * config/aarch64/aarch64.h (CALL_USE_REGISTERS): Fix obsolete
9158         comment.
9159         (EPILOGUE_USES): Likewise.
9161 2018-06-26  Eric Botcazou  <ebotcazou@adacore.com>
9163         * tree-inline.c (remap_location): New function extracted from...
9164         (copy_edges_for_bb): Add ID parameter.  Remap goto_locus.
9165         (copy_phis_for_bb): ...here.  Call remap_location.
9166         (copy_cfg_body): Adjust call to copy_edges_for_bb.
9168 2018-06-26  Aaron Sawdey  <acsawdey@linux.ibm.com>
9170         * config/rs6000/rs6000-string.c (expand_block_clear): Don't use
9171         unaligned vsx for 16B memset.
9173 2018-06-26  Segher Boessenkool  <segher@kernel.crashing.org>
9175         PR target/86285
9176         * config/rs6000/rs6000.c (rs6000_init_builtins): Do not set
9177         ieee128_float_type_node to long_double_type_node unless
9178         TARGET_LONG_DOUBLE_128 is set.
9180 2018-06-26  David Malcolm  <dmalcolm@redhat.com>
9182         * cfgloop.c (get_loop_location): Convert return type from
9183         location_t to dump_user_location_t, replacing INSN_LOCATION lookups
9184         by implicit construction from rtx_insn *, and using
9185         dump_user_location_t::from_function_decl for the fallback case.
9186         * cfgloop.h (get_loop_location): Convert return type from
9187         location_t to dump_user_location_t.
9188         * cgraphunit.c (walk_polymorphic_call_targets): Update call to
9189         dump_printf_loc to pass in a dump_location_t rather than a
9190         location_t, via the gimple stmt.
9191         * coverage.c (get_coverage_counts): Update calls to
9192         dump_printf_loc to pass in dump_location_t rather than a
9193         location_t.
9194         * doc/optinfo.texi (Dump types): Convert example of
9195         dump_printf_loc from taking "locus" to taking "insn".  Update
9196         description of the "_loc" calls to cover dump_location_t.
9197         * dumpfile.c: Include "backend.h", "gimple.h", "rtl.h", and
9198         "selftest.h".
9199         (dump_user_location_t::dump_user_location_t): New constructors,
9200         from gimple *stmt and rtx_insn *.
9201         (dump_user_location_t::from_function_decl): New function.
9202         (dump_loc): Make static.
9203         (dump_gimple_stmt_loc): Convert param "loc" from location_t to
9204         const dump_location_t &.
9205         (dump_generic_expr_loc): Delete.
9206         (dump_printf_loc): Convert param "loc" from location_t to
9207         const dump_location_t &.
9208         (selftest::test_impl_location): New function.
9209         (selftest::dumpfile_c_tests): New function.
9210         * dumpfile.h: Include "profile-count.h".
9211         (class dump_user_location_t): New class.
9212         (struct dump_impl_location_t): New struct.
9213         (class dump_location_t): New class.
9214         (dump_printf_loc): Convert 2nd param from source_location to
9215         const dump_location_t &.
9216         (dump_generic_expr_loc): Delete.
9217         (dump_gimple_stmt_loc): Convert 2nd param from source_location to
9218         const dump_location_t &.
9219         * gimple-fold.c (fold_gimple_assign): Update call to
9220         dump_printf_loc to pass in a dump_location_t rather than a
9221         location_t, via the gimple stmt.
9222         (gimple_fold_call): Likewise.
9223         * gimple-loop-interchange.cc
9224         (loop_cand::analyze_iloop_reduction_var): Update for change to
9225         check_reduction_path.
9226         (tree_loop_interchange::interchange): Update for change to
9227         find_loop_location.
9228         * graphite-isl-ast-to-gimple.c (scop_to_isl_ast): Update for
9229         change in return-type of find_loop_location.
9230         (graphite_regenerate_ast_isl): Likewise.
9231         * graphite-optimize-isl.c (optimize_isl): Likewise.
9232         * graphite.c (graphite_transform_loops): Likewise.
9233         * ipa-devirt.c (ipa_devirt): Update call to dump_printf_loc to
9234         pass in a dump_location_t rather than a location_t, via the
9235         gimple stmt.
9236         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
9237         * ipa.c (walk_polymorphic_call_targets): Likewise.
9238         * loop-unroll.c (report_unroll): Convert "locus" param from
9239         location_t to dump_location_t.
9240         (decide_unrolling): Update for change to get_loop_location's
9241         return type.
9242         * omp-grid.c (struct grid_prop): Convert field "target_loc" from
9243         location_t to dump_user_location_t.
9244         (grid_find_single_omp_among_assignments_1): Updates calls to
9245         dump_printf_loc to pass in a dump_location_t rather than a
9246         location_t, via the gimple stmt.
9247         (grid_parallel_clauses_gridifiable): Convert "tloc" from
9248         location_t to dump_location_t.  Updates calls to dump_printf_loc
9249         to pass in a dump_location_t rather than a location_t, via the
9250         gimple stmt.
9251         (grid_inner_loop_gridifiable_p): Likewise.
9252         (grid_dist_follows_simple_pattern): Likewise.
9253         (grid_gfor_follows_tiling_pattern): Likewise.
9254         (grid_target_follows_gridifiable_pattern): Likewise.
9255         (grid_attempt_target_gridification): Convert initialization
9256         of local "grid" from memset to zero-initialization; FIXME: does
9257         this require C++11?  Update call to dump_printf_loc to pass in a
9258         optinfo_location rather than a location_t, via the gimple stmt.
9259         * profile.c (read_profile_edge_counts): Updates call to
9260         dump_printf_loc to pass in a dump_location_t rather than a
9261         location_t
9262         (compute_branch_probabilities): Likewise.
9263         * selftest-run-tests.c (selftest::run_tests): Call
9264         dumpfile_c_tests.
9265         * selftest.h (dumpfile_c_tests): New decl.
9266         * tree-loop-distribution.c (pass_loop_distribution::execute):
9267         Update for change in return type of find_loop_location.
9268         * tree-parloops.c (parallelize_loops): Likewise.
9269         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Convert
9270         "locus" from location_t to dump_user_location_t.
9271         (canonicalize_loop_induction_variables): Likewise.
9272         * tree-ssa-loop-ivopts.c (tree_ssa_iv_optimize_loop): Update
9273         for change in return type of find_loop_location.
9274         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update call
9275         to dump_printf_loc to pass in a dump_location_t rather than a
9276         location_t, via the stmt.
9277         * tree-ssa-sccvn.c (eliminate_dom_walker::before_dom_children):
9278         Likewise.
9279         * tree-vect-loop-manip.c (find_loop_location): Convert return
9280         type from source_location to dump_user_location_t.
9281         (vect_do_peeling): Update for above change.
9282         (vect_loop_versioning): Update for change in type of
9283         vect_location.
9284         * tree-vect-loop.c (check_reduction_path): Convert "loc" param
9285         from location_t to dump_user_location_t.
9286         (vect_estimate_min_profitable_iters): Update for change in type
9287         of vect_location.
9288         * tree-vect-slp.c (vect_print_slp_tree): Convert param "loc" from
9289         location_t to dump_location_t.
9290         (vect_slp_bb): Update for change in type of vect_location.
9291         * tree-vectorizer.c (vect_location): Convert from source_location
9292         to dump_user_location_t.
9293         (try_vectorize_loop_1): Update for change in vect_location's type.
9294         (vectorize_loops): Likewise.
9295         (increase_alignment): Likewise.
9296         * tree-vectorizer.h (vect_location): Convert from source_location
9297         to dump_user_location_t.
9298         (find_loop_location): Convert return type from source_location to
9299         dump_user_location_t.
9300         (check_reduction_path): Convert 1st param from location_t to
9301         dump_user_location_t.
9302         * value-prof.c (check_counter): Update call to dump_printf_loc to
9303         pass in a dump_user_location_t rather than a location_t; update
9304         call to error_at for change in type of "locus".
9305         (check_ic_target): Update call to dump_printf_loc to
9306         pass in a dump_user_location_t rather than a location_t, via the
9307         call_stmt.
9309 2018-06-26  Robin Dapp  <rdapp@linux.vnet.ibm.com>
9311         * config/s390/s390.h (enum processor_flags): Do not use
9312         default tune parameter when -march was specified.
9314 2018-06-26  Jakub Jelinek  <jakub@redhat.com>
9316         PR target/86314
9317         * config/i386/i386.md (setcc + movzbl to xor + setcc peephole2s):
9318         Check reg_overlap_mentioned_p in addition to reg_set_p with the same
9319         operands.
9321 2018-06-26  Richard Biener  <rguenther@suse.de>
9323         PR tree-optimization/86287
9324         PR bootstrap/86316
9325         * tree-vect-loop.c (vect_transform_loop_stmt): Fix read-after-free.
9326         (vect_analyze_loop): Initialize n_stmts.
9328 2018-06-26  Richard Biener  <rguenther@suse.de>
9330         PR middle-end/86271
9331         * fold-const.c (fold_convertible_p): Pointer extension
9332         isn't valid.
9334 2018-06-26  Alexandre Oliva <aoliva@redhat.com>
9336         PR debug/86064
9337         * dwarf2out.c (loc_list_has_views): Adjust comments.
9338         (dw_loc_list): Split single cross-partition range with
9339         nonzero locview.
9341 2018-06-25  Jeff Law  <law@redhat.com>
9343         * common/config/v850/v850-common.c (TARGET_DEFAULT_TARGET_FLAGS): Turn
9344         on -mbig-switch by default.
9346         * config/v850/predicates.md (const_float_1_operand): Fix match_code
9347         test.
9348         (const_float_0_operand): Remove unused predicate.
9349         * config/v850/v850.md (define_constants): Remove UNSPEC_LOOP.
9350         (define_c_enum unspec): Add LOOP, RCP and RSQRT constants.
9351         (recipsf2): New expander.  Original pattern now called
9352         (recipsf2_insn).
9353         (recipdf2, recipdf2_insn): Similarly.
9354         (rsqrtsf2, rsqrtsf2_insn): Similarly
9355         (rsqrtdf2, rsqrtdf2_insn): Similarly
9357 2018-06-26  Gerald Pfeifer  <gerald@pfeifer.com>
9359         * ginclude/stddef.h: Remove an obsolete comment on FreeBSD 5.
9360         Simplify logic for FreeBSD (twice).
9362 2018-06-25  Martin Sebor  <msebor@redhat.com>
9364         PR tree-optimization/86204
9365         * tree-ssa-strlen.c (handle_builtin_strlen): Avoid storing
9366         a strnlen result if it's less than the length of the string.
9368 2018-06-25  Martin Sebor  <msebor@redhat.com>
9370         PR tree-optimization/85700
9371         * gimple-fold.c (gimple_fold_builtin_strncat): Adjust comment.
9372         * tree-ssa-strlen.c (is_strlen_related_p): Handle integer subtraction.
9373         (maybe_diag_stxncpy_trunc): Distinguish strncat from strncpy.
9375 2018-06-25  Martin Sebor  <msebor@redhat.com>
9377         * doc/extend.texi (Zero-length arrays): Update and clarify.
9379 2018-06-25  Michael Meissner  <meissner@linux.ibm.com>
9381         * config.gcc (powerpc64le*): Revert January 16th, 2018 patch that
9382         added IEEE/IBM long double multilib support on PowerPC little
9383         endian Linux systems.
9384         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Likewise.
9385         (MULTILIB_DEFAULTS): Likewise.
9386         * config/rs6000/rs6000.c (rs6000_option_override_internal):
9387         Likewise.
9388         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Likewise.
9389         * config/rs6000/t-ldouble-linux64le-ibm: Delete, no longer used.
9390         * config/rs6000/t-ldouble-linux64le-ieee: Delete, no longer used.
9392 2018-06-25  Alexander Monakov  <amonakov@ispras.ru>
9394         PR middle-end/86311
9395         * sort.cc (REORDER_23): Avoid memcpy with same destination and source.
9396         (REORDER_45): Likewise.
9398 2018-06-25  Jeff Law  <law@redhat.com>
9400         * config/v850/v850.md (divmodhi4): Make sure to sign extend the
9401         dividend to 32 bits.  Adjust length.
9402         (udivmodhi4): Cleanup output template.  Fix length.
9404 2018-06-25  Carl Love  <cel@us.ibm.com>
9406         * config/rs6000/vsx.md: Change word selector to prefered location.
9408 2018-06-25  Richard Biener  <rguenther@suse.de>
9410         PR tree-optimization/86304
9411         * tree-vectorizer.c (vectorize_loops): Walk over new possibly
9412         epilogue-if-converted loops as well.
9414 2018-06-25  Jan Hubicka  <hubicka@ucw.cz>
9416         * lto-section-out.c (lto_begin_section): Do not print section
9417         name for noaddr and unnumbered dumps.
9419 2018-06-25  Richard Biener  <rguenther@suse.de>
9421         * tree-vectorizer.h (struct vec_info_shared): New structure
9422         with parts split out from struct vec_info and loop_nest from
9423         struct _loop_vec_info.
9424         (struct vec_info): Adjust accordingly.
9425         (struct _loop_vec_info): Likewise.
9426         (LOOP_VINFO_LOOP_NEST): Adjust.
9427         (LOOP_VINFO_DATAREFS): Likewise.
9428         (LOOP_VINFO_DDRS): Likewise.
9429         (struct _bb_vec_info): Likewise.
9430         (BB_VINFO_DATAREFS): Likewise.
9431         (BB_VINFO_DDRS): Likewise.
9432         (struct _stmt_vec_info): Add dr_aux member.
9433         (DR_VECT_AUX): Adjust to refer to member of DR_STMTs vinfo.
9434         (DR_MISALIGNMENT_UNINITIALIZED): New.
9435         (set_dr_misalignment): Adjust.
9436         (dr_misalignment): Assert misalign isn't DR_MISALIGNMENT_UNINITIALIZED.
9437         (vect_analyze_loop): Adjust prototype.
9438         (vect_analyze_loop_form): Likewise.
9439         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
9440         Compute dependences lazily.
9441         (vect_record_base_alignments): Use shared datarefs/ddrs.
9442         (vect_verify_datarefs_alignment): Likewise.
9443         (vect_analyze_data_refs_alignment): Likewise.
9444         (vect_analyze_data_ref_accesses): Likewise.
9445         (vect_analyze_data_refs): Likewise.
9446         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Add
9447         constructor parameter for shared part.
9448         (vect_analyze_loop_form): Pass in shared part and adjust.
9449         (vect_analyze_loop_2): Pass in storage for the number of
9450         stmts.  Move loop nest finding to the caller.  Compute
9451         datarefs lazily.
9452         (vect_analyze_loop): Pass in shared part.
9453         (vect_transform_loop): Verify shared datarefs are unchanged.
9454         * tree-vect-slp.c (_bb_vec_info::_bb_vec_info): Add
9455         constructor parameter for shared part.
9456         (vect_slp_analyze_bb_1): Pass in shared part and adjust.
9457         (vect_slp_bb): Verify shared datarefs are unchanged before
9458         transform.
9459         * tree-vect-stmts.c (ensure_base_align): Adjust for DR_AUX
9460         change.
9461         (new_stmt_vec_info): Initialize DR_AUX misalignment to
9462         DR_MISALIGNMENT_UNINITIALIZED.
9463         * tree-vectorizer.c (vec_info::vec_info): Add constructor
9464         parameter for shared part.
9465         (vec_info::~vec_info): Adjust.
9466         (vec_info_shared::vec_info_shared): New.
9467         (vec_info_shared::~vec_info_shared): Likewise.
9468         (vec_info_shared::save_datarefs): Likewise.
9469         (vec_info_shared::check_datarefs): Likewise.
9470         (try_vectorize_loop_1): Construct shared part live for analyses
9471         of a single loop for multiple vector sizes.
9472         * tree-parloops.c (gather_scalar_reductions): Adjust.
9474 2018-06-25  Richard Biener  <rguenther@suse.de>
9476         * tree-vect-data-refs.c (vect_find_stmt_data_reference): Modify
9477         DR for SIMD lane accesses here and mark DR with (void *)-1 aux.
9478         (vect_analyze_data_refs): Remove similar code from here and
9479         simplify accordingly.
9481 2018-06-25  Richard Biener  <rguenther@suse.de>
9483         * tree-vect-data-refs.c (vect_check_gather_scatter): Fail
9484         for reverse storage order accesses rather than asserting
9485         they cannot happen here.
9487 2018-06-25  Tom de Vries  <tdevries@suse.de>
9489         PR debug/86257
9490         * config/i386/i386.md (define_insn "*tls_global_dynamic_64_<mode>"):
9491         Use data16 instead of .byte for insn prefix.
9493 2018-06-25  Andreas Krebbel  <krebbel@linux.ibm.com>
9495         PR C++/86082
9496         * parser.c (make_char_string_pack): Pass this literal chars
9497         through cpp_interpret_string.
9498         (cp_parser_userdef_numeric_literal): Check the result of
9499         make_char_string_pack.
9501 2018-06-24  Maya Rashish  <coypu@sdf.org>
9503         * ginclude/stddef.h: Simplify conditions around avoiding
9504         re-definition of __size_t.
9506 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
9508         * lto-streamer-out.c (tree_is_indexable): Make LABEL_DECL nonindexable
9509         unless it is forced or nonlocal; assert that we stream no IMPORTED_DECL.
9511 2018-06-22  Maya Rashish  <coypu@sdf.org>
9513         * doc/invoke.texi (mno-fancy-math-387): Update for changes
9514         made to OpenBSD and NetBSD through the years.
9516 2018-06-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9518         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
9519         behavior of vec_pack (vector double, vector double) to match
9520         behavior of vec_float2 (vector double, vector double).
9522 2018-06-22  Olivier Hainque  <hainque@adacore.com>
9524         * gimplify.c (gimplify_function_tree): Prevent creation
9525         of a trampoline for the address of the current function
9526         passed to entry/exit instrumentation hooks.
9528 2018-06-22  Aaron Sawdey  <acsawdey@linux.ibm.com>
9530         PR target/86222
9531         * config/rs6000/rs6000-string.c (expand_strn_compare): Handle -m32
9532         correctly.
9534 2018-06-22  Martin Liska  <mliska@suse.cz>
9536         PR tree-optimization/86263
9537         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9538         Bail out if is_enabled is false.
9539         * tree-switch-conversion.h (jump_table_cluster::is_enabled):
9540         New declaration.
9541         (jump_table_cluster::is_enabled): New function.
9543 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
9545         * lto-streamer-out.c (DFS::DFS_write_tree_body): Do not stream
9546         BINFO_BASE_ACCESSES and BINFO_VPTR_FIELD.
9547         * tree-streamer-in.c (streamer_read_tree_bitfields): Likewise.
9548         (lto_input_ts_binfo_tree_pointers): Likewise.
9549         * tree-streamer-out.c (streamer_write_tree_bitfields,
9550         write_ts_binfo_tree_pointers): Likewise.
9551         * tree.c (free_lang_data_in_binfo): Clear BINFO_VPTR_FIELD.
9553 2018-06-22  Jan Hubicka  <hubicka@ucw.cz>
9555         * tree.c (free_lang_data_in_type): Free all TYPE_VFIELDs.
9557 2018-06-22  Martin Liska  <mliska@suse.cz>
9559         * symbol-summary.h (get): Make it pure and inline move
9560         functionality from ::get function.
9561         (get): Remove and inline into ::get and ::get_create.
9562         (get_create): Move code from ::get function.
9564 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9566         PR target/85994
9567         * config/i386/sol2.h (CPP_SPEC): Don't pass -P for
9568         -x assembler-with-cpp.
9570 2018-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9572         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define
9573         _FILE_OFFSET_BITS=64 for C++.
9575 2018-06-21  Michael Meissner  <meissner@linux.ibm.com>
9577         * config/rs6000/rs6000.md (extendtfif2): Add missing 128-bit
9578         conversion insn that shows up when pr85657-3.c is compiled using
9579         IEEE 128-bit long double.
9580         (neg<mode>2_internal): Use the correct mode to check whether the
9581         mode is IBM extended.
9582         * config/rs6000/rs6000.c (init_float128_ieee): Prevent complex
9583         multiply and divide external functions from being created more
9584         than once.
9586 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
9588         * cfgrtl.c (fixup_reorder_chain): Do not emit NOPs in DECL_IGNORED_P
9589         functions.
9590         (rtl_merge_blocks): Likewise.  Do not emit a NOP if the location of
9591         the edge can be forwarded.
9592         (cfg_layout_merge_blocks): Likewise.
9594 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
9596         * except.c (finish_eh_generation): Commit edge insertions only after
9597         the EH edges have been redirected from post-landing to landing pads.
9599 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
9601         * tree-nested.c (get_frame_type): Use create_tmp_var_raw instead of
9602         create_tmp_var_for to create the FRAME decl.
9603         (finalize_nesting_tree_1): Do not unchain the FRAME decl.
9605 2018-06-21  Eric Botcazou  <ebotcazou@adacore.com>
9607         * tree-inline.c (copy_edges_for_bb): Minor tweak.
9608         (maybe_move_debug_stmts_to_successors): Also reset the locus of the
9609         debug statement when resetting its value.
9610         (expand_call_inline): Copy the locus of the call onto the assignment
9611         of the return value, if any.  Use local variable in more cases.
9613 2018-06-21  Martin Liska  <mliska@suse.cz>
9615         * ipa-pure-const.c (propagate_nothrow): Use
9616         funct_state_summaries->get.
9617         (dump_malloc_lattice): Likewise.
9618         (propagate_malloc): Likewise.
9620 2018-06-21  Richard Biener  <rguenther@suse.de>
9622         * lto-streamer-out.c (DFS::DFS_write_tree_body): Update outdated
9623         comment.  Follow BLOCK_ABSTRACT_ORIGIN unconditionally.
9624         * tree-streamer-in.c (lto_input_ts_block_tree_pointers): Update
9625         comment.
9626         * tree-streamer-out.c (write_ts_block_tree_pointers): Stream
9627         BLOCK_ABSTRACT_ORIGIN unconditionally.
9629 2018-06-21  David Malcolm  <dmalcolm@redhat.com>
9631         * ipa-cp.c (ipcp_driver): Set edge_clone_summaries to NULL after
9632         deleting it.
9633         * ipa-reference.c (ipa_reference_c_finalize): Delete
9634         ipa_ref_opt_sum_summaries and set it to NULL.
9636 2018-06-21  Tom de Vries  <tdevries@suse.de>
9638         PR tree-optimization/85859
9639         * tree-ssa-tail-merge.c (stmt_local_def): Copy gimple_is_call
9640         test with comment from bb_no_side_effects_p.
9642 2018-06-21  Richard Biener  <rguenther@suse.de>
9644         PR tree-optimization/86232
9645         * tree-ssa-loop-niter.c (number_of_iterations_popcount): Adjust
9646         max for constant niter.
9648 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9650         * config/aarch64/aarch64-simd.md
9651         (*aarch64_crypto_aes<aes_op>v16qi_xor_combine): New.
9653 2018-06-21  Andre Vieira  <andre.simoesdiasvieira@arm.com>
9655         * config/aarch64/aarch64-simd.md (aarch64_crypto_aes<aes_op>v16qi):
9656         Make opernads of the unspec commutative.
9658 2018-06-21  Richard Biener  <rguenther@suse.de>
9660         * tree-data-ref.c (dr_step_indicator): Handle NULL DR_STEP.
9661         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
9662         Avoid calling vect_mark_for_runtime_alias_test with gathers or scatters.
9663         (vect_analyze_data_ref_dependence): Re-order checks to deal with
9664         NULL DR_STEP.
9665         (vect_record_base_alignments): Do not record base alignment
9666         for gathers or scatters.
9667         (vect_compute_data_ref_alignment): Drop return value that is always
9668         true.  Bail out early for gathers or scatters.
9669         (vect_enhance_data_refs_alignment): Bail out early for gathers
9670         or scatters.
9671         (vect_find_same_alignment_drs): Likewise.
9672         (vect_analyze_data_refs_alignment): Remove dead code.
9673         (vect_slp_analyze_and_verify_node_alignment): Likewise.
9674         (vect_analyze_data_refs): For possible gathers or scatters do
9675         not create an alternate DR, just check their possible validity
9676         and mark them.  Adjust DECL_NONALIASED handling to not rely
9677         on DR_BASE_ADDRESS.
9678         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Do not
9679         update inits of gathers or scatters.
9680         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
9681         Also copy gather/scatter flag to pattern vinfo.
9683 2018-06-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
9685         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Change
9686         behavior of vec_packsu (vector unsigned long long, vector unsigned
9687         long long) to match behavior of vec_packs with same signature.
9689 2018-06-20  Chung-Lin Tang <cltang@codesourcery.com>
9690             Thomas Schwinge <thomas@codesourcery.com>
9691             Cesar Philippidis  <cesar@codesourcery.com>
9693         * gimplify.c (gimplify_scan_omp_clauses): Add support for
9694         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
9695         (gimplify_adjust_omp_clauses): Likewise.
9696         (gimplify_oacc_declare_1): Add support for GOMP_MAP_RELEASE, remove
9697         support for GOMP_MAP_FORCE_{ALLOC,TO,FROM,TOFROM}.
9698         (gimplify_omp_target_update): Update handling of acc update and
9699         enter/exit data.
9700         * omp-low.c (install_var_field): Remove unused parameter
9701         base_pointers_restrict.
9702         (scan_sharing_clauses): Remove base_pointers_restrict parameter.
9703         Update call to install_var_field. Handle OMP_CLAUSE_{IF_PRESENT,
9704         FINALIZE}
9705         (omp_target_base_pointers_restrict_p): Delete.
9706         (scan_omp_target): Update call to scan_sharing_clauses.
9707         * tree-core.h (enum omp_clause_code): Add OMP_CLAUSE_{IF_PRESENT,
9708         FINALIZE}.
9709         * tree-nested.c (convert_nonlocal_omp_clauses): Handle
9710         OMP_CLAUSE_{IF_PRESENT,FINALIZE}.
9711         (convert_local_omp_clauses): Likewise.
9712         * tree-pretty-print.c (dump_omp_clause): Likewise.
9713         * tree.c (omp_clause_num_ops): Add entries for  OMP_CLAUSE_{IF_PRESENT,
9714         FINALIZE}.
9715         (omp_clause_code_name): Likewise.
9717 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
9719         PR debug/86194
9720         * var-tracking.c (use_narrower_mode_test): Check if shift amount can
9721         be narrowed.
9723         PR tree-optimization/86231
9724         * tree-vrp.c (union_ranges): For (  [  )  ] or (   )[   ] range and
9725         anti-range don't overwrite *vr0min before using it to compute *vr0max.
9727 2018-06-20  Tom de Vries  <tdevries@suse.de>
9729         PR tree-optimization/86097
9730         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Also convert *nit to
9731         iv type if signedness of iv type is not the same as that of *nit.
9733 2018-06-20  Jakub Jelinek  <jakub@redhat.com>
9735         * cfgrtl.c (rtl_verify_edges): Formatting fix.  If bb->preds has any
9736         EDGE_EH edges, verify they are all EDGE_EH.
9738 2018-06-20  Maya Rashish  <coypu@sdf.org>
9740         * ginclude/stddef.h: Limit #include <machine/ansi.h> to NetBSD.
9742 2018-06-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
9744         * config/aarch64/aarch64-tuning-flags.def (no_ldp_stp_qregs): New.
9745         * config/aarch64/aarch64.c (xgene1_tunings): Add
9746         AARCH64_EXTRA_TUNE_NO_LDP_STP_QREGS to tune_flags.
9747         (aarch64_mode_valid_for_sched_fusion_p):
9748         Allow 16-byte modes.
9749         (aarch64_classify_address): Allow 16-byte modes for load_store_pair_p.
9750         * config/aarch64/aarch64-ldpstp.md: Add peepholes for LDP STP of
9751         128-bit modes.
9752         * config/aarch64/aarch64-simd.md (load_pair<VQ:mode><VQ2:mode>):
9753         New pattern.
9754         (vec_store_pair<VQ:mode><VQ2:mode>): Likewise.
9755         * config/aarch64/iterators.md (VQ2): New mode iterator.
9757 2018-06-20  Martin Liska  <mliska@suse.cz>
9759         * tree-switch-conversion.c (jump_table_cluster::can_be_handled):
9760         Change default ratio from 10 to 8.
9762 2018-06-20  Martin Liska  <mliska@suse.cz>
9764         * tree-switch-conversion.c (jump_table_cluster::find_jump_tables):
9765         New.
9766         (bit_test_cluster::find_bit_tests): Likewise.
9767         (switch_decision_tree::analyze_switch_statement): Find clusters.
9768         * tree-switch-conversion.h (struct jump_table_cluster): Document
9769         hierarchy.
9771 2018-06-20  Martin Liska  <mliska@suse.cz>
9773         * tree-switch-conversion.c (switch_conversion::collect):
9774         Record m_uniq property.
9775         (switch_conversion::expand): Bail out for special conditions.
9776         (group_cluster::~group_cluster): New.
9777         (group_cluster::group_cluster): Likewise.
9778         (group_cluster::dump): Likewise.
9779         (jump_table_cluster::emit): New.
9780         (switch_decision_tree::fix_phi_operands_for_edges): New.
9781         (struct case_node): Remove struct.
9782         (jump_table_cluster::can_be_handled): New.
9783         (case_values_threshold): Moved to header.
9784         (reset_out_edges_aux): Likewise.
9785         (jump_table_cluster::is_beneficial): New.
9786         (bit_test_cluster::can_be_handled): Likewise.
9787         (add_case_node): Remove.
9788         (bit_test_cluster::is_beneficial): New.
9789         (case_bit_test::cmp): New.
9790         (bit_test_cluster::emit): New.
9791         (expand_switch_as_decision_tree_p): Remove.
9792         (bit_test_cluster::hoist_edge_and_branch_if_true): New.
9793         (fix_phi_operands_for_edge): Likewise.
9794         (switch_decision_tree::analyze_switch_statement): New.
9795         (compute_cases_per_edge): Move ...
9796         (switch_decision_tree::compute_cases_per_edge): ... here.
9797         (try_switch_expansion): Likewise.
9798         (switch_decision_tree::try_switch_expansion): Likewise.
9799         (record_phi_operand_mapping): Likewise.
9800         (switch_decision_tree::record_phi_operand_mapping): Likewise.
9801         (emit_case_decision_tree): Likewise.
9802         (switch_decision_tree::emit): Likewise.
9803         (balance_case_nodes): Likewise.
9804         (switch_decision_tree::balance_case_nodes): Likewise.
9805         (dump_case_nodes): Likewise.
9806         (switch_decision_tree::dump_case_nodes): Likewise.
9807         (emit_jump): Likewise.
9808         (switch_decision_tree::emit_jump): Likewise.
9809         (emit_cmp_and_jump_insns): Likewise.
9810         (switch_decision_tree::emit_cmp_and_jump_insns): Likewise.
9811         (emit_case_nodes): Likewise.
9812         (switch_decision_tree::emit_case_nodes): Likewise.
9813         (conditional_probability): Remove.
9814         * tree-switch-conversion.h (enum cluster_type): New.
9815         (PRINT_CASE): New.
9816         (struct cluster): Likewise.
9817         (cluster::cluster): Likewise.
9818         (struct simple_cluster): Likewise.
9819         (simple_cluster::simple_cluster): Likewise.
9820         (struct group_cluster): Likewise.
9821         (struct jump_table_cluster): Likewise.
9822         (struct bit_test_cluster): Likewise.
9823         (struct min_cluster_item): Likewise.
9824         (struct case_tree_node): Likewise.
9825         (case_tree_node::case_tree_node): Likewise.
9826         (jump_table_cluster::case_values_threshold): Likewise.
9827         (struct case_bit_test): Likewise.
9828         (struct switch_decision_tree): Likewise.
9829         (struct switch_conversion): Likewise.
9830         (switch_decision_tree::reset_out_edges_aux): Likewise.
9832 2018-06-20  Martin Liska  <mliska@suse.cz>
9834         * tree-switch-conversion.c (MAX_CASE_BIT_TESTS): Remove.
9835         (hoist_edge_and_branch_if_true): Likewise.
9836         (expand_switch_using_bit_tests_p): Likewise.
9837         (struct case_bit_test): Likewise.
9838         (case_bit_test_cmp): Likewise.
9839         (emit_case_bit_tests): Likewise.
9840         (switch_conversion::switch_conversion): New class.
9841         (struct switch_conv_info): Remove old struct.
9842         (collect_switch_conv_info): More to ...
9843         (switch_conversion::collect): ... this.
9844         (check_range): Likewise.
9845         (switch_conversion::check_range): Likewise.
9846         (check_all_empty_except_final): Likewise.
9847         (switch_conversion::check_all_empty_except_final): Likewise.
9848         (check_final_bb): Likewise.
9849         (switch_conversion::check_final_bb): Likewise.
9850         (create_temp_arrays): Likewise.
9851         (switch_conversion::create_temp_arrays): Likewise.
9852         (free_temp_arrays): Likewise.
9853         (gather_default_values): Likewise.
9854         (switch_conversion::gather_default_values): Likewise.
9855         (build_constructors): Likewise.
9856         (switch_conversion::build_constructors): Likewise.
9857         (constructor_contains_same_values_p): Likewise.
9858         (switch_conversion::contains_same_values_p): Likewise.
9859         (array_value_type): Likewise.
9860         (switch_conversion::array_value_type): Likewise.
9861         (build_one_array): Likewise.
9862         (switch_conversion::build_one_array): Likewise.
9863         (build_arrays): Likewise.
9864         (switch_conversion::build_arrays): Likewise.
9865         (gen_def_assigns): Likewise.
9866         (switch_conversion::gen_def_assigns): Likewise.
9867         (prune_bbs): Likewise.
9868         (switch_conversion::prune_bbs): Likewise.
9869         (fix_phi_nodes): Likewise.
9870         (switch_conversion::fix_phi_nodes): Likewise.
9871         (gen_inbound_check): Likewise.
9872         (switch_conversion::gen_inbound_check): Likewise.
9873         (process_switch): Use the newly created class.
9874         (switch_conversion::expand): New.
9875         (switch_conversion::~switch_conversion): New.
9876         * tree-switch-conversion.h: New file.
9878 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9880         * tree-vectorizer.h (NUM_PATTERNS, vect_recog_func_ptr): Move to
9881         tree-vect-patterns.c.
9882         * tree-vect-patterns.c (vect_supportable_direct_optab_p): New function.
9883         (vect_recog_dot_prod_pattern): Use it.  Remove the type_in argument.
9884         (vect_recog_sad_pattern): Likewise.
9885         (vect_recog_widen_sum_pattern): Likewise.
9886         (vect_recog_pow_pattern): Likewise.  Check for a null vectype.
9887         (vect_recog_widen_shift_pattern): Remove the type_in argument.
9888         (vect_recog_rotate_pattern): Likewise.
9889         (vect_recog_mult_pattern): Likewise.
9890         (vect_recog_vector_vector_shift_pattern): Likewise.
9891         (vect_recog_divmod_pattern): Likewise.
9892         (vect_recog_mixed_size_cond_pattern): Likewise.
9893         (vect_recog_bool_pattern): Likewise.
9894         (vect_recog_mask_conversion_pattern): Likewise.
9895         (vect_try_gather_scatter_pattern): Likewise.
9896         (vect_recog_widen_mult_pattern): Likewise.  Check for a null vectype.
9897         (vect_recog_over_widening_pattern): Likewise.
9898         (vect_recog_gather_scatter_pattern): Likewise.
9899         (vect_recog_func_ptr): Move from tree-vectorizer.h
9900         (vect_vect_recog_func_ptrs): Move further down the file.
9901         (vect_recog_func): Likewise.  Remove the third argument.
9902         (NUM_PATTERNS): Define based on vect_vect_recog_func_ptrs.
9903         (vect_pattern_recog_1): Expect the pattern function to do any
9904         necessary target tests.  Also expect it to provide a vector type.
9905         Remove the type_in handling.
9907 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9909         * tree-vect-patterns.c (vect_pattern_detected): New function.
9910         (vect_recog_dot_prod_patternm, vect_recog_sad_pattern)
9911         (vect_recog_widen_mult_pattern, vect_recog_widen_sum_pattern)
9912         (vect_recog_over_widening_pattern, vect_recog_widen_shift_pattern
9913         (vect_recog_rotate_pattern, vect_recog_vector_vector_shift_pattern)
9914         (vect_recog_mult_pattern, vect_recog_divmod_pattern)
9915         (vect_recog_mixed_size_cond_pattern, vect_recog_bool_pattern)
9916         (vect_recog_mask_conversion_pattern)
9917         (vect_try_gather_scatter_pattern): Likewise.
9919 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9921         * tree-vect-patterns.c (vect_get_internal_def): New function.
9922         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern)
9923         (vect_recog_vector_vector_shift_pattern, check_bool_pattern)
9924         (search_type_for_mask_1): Use it.
9926 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9928         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
9929         redundant WIDEN_SUM_EXPR handling.
9930         (vect_recog_sad_pattern): Likewise.
9932 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9934         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Remove
9935         redundant check that the types of a PLUS_EXPR or MULT_EXPR agree.
9936         (vect_recog_sad_pattern): Likewise PLUS_EXPR, ABS_EXPR and MINUS_EXPR.
9937         (vect_recog_widen_mult_pattern): Likewise MULT_EXPR.
9938         (vect_recog_widen_sum_pattern): Likewise PLUS_EXPR.
9940 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9942         * tree-vect-stmts.c (vectorizable_call): Make sure that we
9943         use the stmt_vec_info of the original bb statement for the
9944         new zero assignment, even if the call is part of a pattern.
9946 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9948         * tree-vectorizer.h (_stmt_vec_info): Note above pattern_def_seq
9949         that the sequence is attached to the original statement rather
9950         than the pattern statement.
9951         * tree-vect-loop.c (vect_determine_vf_for_stmt): Take the
9952         PATTERN_DEF_SEQ from the original statement rather than
9953         the main pattern statement.
9954         * tree-vect-stmts.c (free_stmt_vec_info): Likewise.
9955         * tree-vect-patterns.c (vect_recog_dot_prod_pattern): Likewise.
9956         (vect_mark_pattern_stmts): Don't copy the PATTERN_DEF_SEQ.
9958 2018-06-20  Richard Sandiford  <richard.sandiford@arm.com>
9960         * tree-vect-stmts.c (vect_analyze_stmt): Move the handling of pattern
9961         definition statements before the early exit for statements that aren't
9962         live or relevant.
9963         * tree-vect-loop.c (vect_transform_loop_stmt): New function,
9964         split out from...
9965         (vect_transform_loop): ...here.  Process pattern definition
9966         statements without first checking whether the main pattern
9967         statement is live or relevant.
9969 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
9971         * tree-cfgcleanup.c (tree_forwarder_block_p): Do not return false at
9972         -O0 if the locus represent UNKNOWN_LOCATION but have different values.
9974 2018-06-19  Aaron Sawdey  <acsawdey@linux.ibm.com>
9976         * config/rs6000/rs6000-string.c (select_block_compare_mode): Check
9977         TARGET_EFFICIENT_OVERLAPPING_UNALIGNED here instead of in caller.
9978         (do_and3, do_and3_mask, do_compb3, do_rotl3): New functions.
9979         (expand_block_compare): Change select_block_compare_mode call.
9980         (expand_strncmp_align_check): Use new functions, fix comment.
9981         (emit_final_str_compare_gpr): New function.
9982         (expand_strn_compare): Refactor and clean up code.
9983         * config/rs6000/vsx.md (vsx_mov<mode>_64bit): Remove *.
9985 2018-06-19  Tony Reix  <tony.reix@atos.com>
9986             Damien Bergamini  <damien.bergamini@atos.com>
9987             David Edelsohn  <dje.gcc@gmail.com>
9989         * collect2.c (static_obj): New variable.
9990         (static_libs): New variable.
9991         (is_in_list): Uncomment declaration.
9992         (main): Track AIX libraries linked statically.
9993         (is_in_list): Uncomment definition.
9994         (scan_prog_file): Don't add AIX shared libraries initializer
9995         to constructor list if linking statically.
9997 2018-06-19  Max Filippov  <jcmvbkbc@gmail.com>
9999         * config/xtensa/xtensa.md (UNSPEC_FRAME_BLOCKAGE): New unspec
10000         constant.
10001         (allocate_stack, frame_blockage, *frame_blockage): New patterns.
10003 2018-06-19  Jan Hubicka  <hubicka@ucw.cz>
10005         * tree.c (find_decls_types_r): Remove all non-VAR_DECLs from
10006         blocks.
10008 2018-06-19  Martin Liska  <mliska@suse.cz>
10010         * config/i386/i386.c (ix86_can_inline_p): Do not use
10011         ipa_fn_summaries::get_create.
10012         * ipa-cp.c (ipcp_cloning_candidate_p): Replace get_create with
10013         get.
10014         (devirtualization_time_bonus): Likewise.
10015         (ipcp_propagate_stage): Likewise.
10016         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
10017         (edge_set_predicate): Likewise.
10018         (evaluate_conditions_for_known_args): Likewise.
10019         (evaluate_properties_for_edge): Likewise.
10020         (ipa_call_summary::reset): Tranform to ...
10021         (ipa_call_summary::~ipa_call_summary): ... this.
10022         (ipa_fn_summary::reset): Transform to ...
10023         (ipa_fn_summary::~ipa_fn_summary): ... this.
10024         (ipa_fn_summary_t::remove): Rename to ...
10025         (ipa_fn_summary_t::remove_callees): ... this.
10026         (ipa_fn_summary_t::duplicate): Use placement new
10027         instead of memory copy.
10028         (ipa_call_summary_t::duplicate): Likewise.
10029         (ipa_call_summary_t::remove): Remove.
10030         (dump_ipa_call_summary): Change get_create to get.
10031         (ipa_dump_fn_summary): Dump only when summary exists.
10032         (analyze_function_body): Use symbol_summary::get instead
10033         of get_create.
10034         (compute_fn_summary): Likewise.
10035         (estimate_edge_devirt_benefit): Likewise.
10036         (estimate_edge_size_and_time): Likewise.
10037         (inline_update_callee_summaries): Likewise.
10038         (remap_edge_change_prob): Likewise.
10039         (remap_edge_summaries): Likewise.
10040         (ipa_merge_fn_summary_after_inlining): Likewise.
10041         (write_ipa_call_summary): Likewise.
10042         (ipa_fn_summary_write): Likewise.
10043         (ipa_free_fn_summary): Likewise.
10044         * ipa-fnsummary.h (struct GTY): Add new ctor and copy ctor.
10045         (struct ipa_call_summary): Likewise.
10046         * ipa-icf.c (sem_function::merge): Use symbol_summary::get instead
10047         of get_create.
10048         * ipa-inline-analysis.c (do_estimate_edge_time): Likewise.
10049         (estimate_size_after_inlining): Likewise.
10050         (estimate_growth): Likewise.
10051         (growth_likely_positive): Likewise.
10052         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
10053         (inline_call): Likewise.
10054         * ipa-inline.c (caller_growth_limits): Likewise.
10055         (can_inline_edge_p): Likewise.
10056         (can_inline_edge_by_limits_p): Likewise.
10057         (compute_uninlined_call_time): Likewise.
10058         (compute_inlined_call_time): Likewise.
10059         (want_inline_small_function_p): Likewise.
10060         (edge_badness): Likewise.
10061         (update_caller_keys): Likewise.
10062         (update_callee_keys): Likewise.
10063         (inline_small_functions): Likewise.
10064         (inline_to_all_callers_1): Likewise.
10065         (dump_overall_stats): Likewise.
10066         (early_inline_small_functions): Likewise.
10067         (early_inliner): Likewise.
10068         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
10069         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
10070         * ipa-pure-const.c (malloc_candidate_p): Likewise.
10071         * ipa-split.c (execute_split_functions): Likewise.
10072         * symbol-summary.h: Likewise.
10073         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
10075 2018-06-19  Richard Biener  <rguenther@suse.de>
10077         * tree-vectorizer.c (try_vectorize_loop_1): Split out of ...
10078         (vectorize_loops): ... here.  Fix dbgcnt handling.
10079         (try_vectorize_loop): Wrap try_vectorize_loop_1.
10081 2018-06-19  Segher Boessenkool  <segher@kernel.crashing.org>
10083         PR target/86197
10084         * config/rs6000/rs6000.md (rs6000_discover_homogeneous_aggregate): An
10085         ieee128 argument takes up only one (vector) register, not two (floating
10086         point) registers.
10088 2018-06-19  Eric Botcazou  <ebotcazou@adacore.com>
10090         * gimplify.c (gimplify_init_constructor): Really never clear for an
10091         incomplete constructor if CONSTRUCTOR_NO_CLEARING is set.
10093 2018-06-19  Richard Biener  <rguenther@suse.de>
10095         PR tree-optimization/86179
10096         * tree-vect-patterns.c (vect_pattern_recog_1): Clean up
10097         after failed recognition.
10099 2018-06-18  Martin Sebor  <msebor@redhat.com>
10101         PR middle-end/85602
10102         * calls.c (maybe_warn_nonstring_arg): Handle strncat.
10103         * tree-ssa-strlen.c (is_strlen_related_p): Make extern.
10104         Handle integer subtraction.
10105         (maybe_diag_stxncpy_trunc): Handle nonstring source arguments.
10106         * tree-ssa-strlen.h (is_strlen_related_p): Declare.
10108 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
10110         * config/frv/frv-protos.h (frv_ifcvt_modify_insn): Strengthen 3rd
10111         param from rtx to rtx_insn *.
10112         * config/frv/frv.c (frv_ifcvt_add_insn): Likewise for "insn"
10113         param.
10114         (frv_ifcvt_modify_insn): Likwise.
10115         (frv_ifcvt_modify_final): Likwise for local "existing_insn",
10116         adding an as_a <rtx_insn *> cast.  Likewise for local "insn".
10117         * config/mips/mips.c (r10k_insert_cache_barriers): Add an
10118         as_a <rtx_insn *> cast to local "unprotected_region" once
10119         it's been established that it's not NULL or pc_rtx.
10120         * config/nds32/nds32-relax-opt.c (nds32_group_insns): Strengthen
10121         param "sethi" from rtx to rtx_insn *.
10122         (nds32_group_float_insns): Likewise for param "insn".
10123         * config/vax/vax-protos.h (vax_output_int_add): Likewise for 1st
10124         param.
10125         (vax_output_int_subtract): Likewise.
10126         * config/vax/vax.c (vax_output_int_add): Likewise for param
10127         "insn".
10128         (vax_output_int_subtract): Likewise.
10129         * emit-rtl.c (set_insn_deleted): Likewise, removing cast.
10130         (emit_pattern_after): Likewise for param "after".
10131         (emit_insn_after): Likewise.
10132         (emit_jump_insn_after): Likewise.
10133         (emit_call_insn_after): Likewise.
10134         (emit_debug_insn_after): Likewise.
10135         (emit_pattern_before): Likewise for param "before".
10136         (emit_insn_before): Likewise.
10137         (emit_jump_insn_before): Likewise.
10138         * final.c (get_insn_template): Likewise for param "insn", removing
10139         a cast.
10140         * output.h (get_insn_template): Likewise for 2nd param.
10141         * rtl.h (emit_insn_before): Likewise.
10142         (emit_jump_insn_before): Likewise.
10143         (emit_debug_insn_before_noloc): Likewise.
10144         (emit_insn_after): Likewise.
10145         (emit_jump_insn_after): Likewise.
10146         (emit_call_insn_after): Likewise.
10147         (emit_debug_insn_after): Likewise.
10148         (set_insn_deleted): Likewise for param.
10150 2018-06-18  Michael Meissner  <meissner@linux.ibm.com>
10152         PR target/85358
10153         * config/rs6000/rs6000-modes.def (toplevel): Rework the 128-bit
10154         floating point modes, so that IFmode is numerically greater than
10155         TFmode, which is greater than KFmode using FRACTIONAL_FLOAT_MODE
10156         to declare the ordering.  This prevents IFmode from being
10157         converted to TFmode when long double is IEEE 128-bit on an ISA 3.0
10158         machine.  Include rs6000-modes.h to share the fractional values
10159         between genmodes* and the rest of the compiler.
10160         (IFmode): Likewise.
10161         (KFmode): Likewise.
10162         (TFmode): Likewise.
10163         * config/rs6000/rs6000-modes.h: New file.
10164         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Change the
10165         meaning of rs6000_long_double_size so that 126..128 selects an
10166         appropriate 128-bit floating point type.
10167         (rs6000_option_override_internal): Likewise.
10168         * config/rs6000/rs6000.h (toplevel): Include rs6000-modes.h.
10169         (TARGET_LONG_DOUBLE_128): Change the meaning of
10170         rs6000_long_double_size so that 126..128 selects an appropriate
10171         128-bit floating point type.
10172         (LONG_DOUBLE_TYPE_SIZE): Update comment.
10173         * config/rs6000/rs6000.md (trunciftf2): Correct the modes of the
10174         source and destination to match the standard usage.
10175         (truncifkf2): Likewise.
10176         (copysign<mode>3, IEEE iterator): Rework copysign of float128 on
10177         ISA 2.07 to use an explicit clobber, instead of passing in a
10178         temporary.
10179         (copysign<mode>3_soft): Likewise.
10181 2018-06-18  David Malcolm  <dmalcolm@redhat.com>
10183         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
10184         Replace dump_printf_loc call with DUMP_VECT_SCOPE.
10185         (vect_slp_analyze_instance_dependence): Likewise.
10186         (vect_enhance_data_refs_alignment): Likewise.
10187         (vect_analyze_data_refs_alignment): Likewise.
10188         (vect_slp_analyze_and_verify_instance_alignment
10189         (vect_analyze_data_ref_accesses): Likewise.
10190         (vect_prune_runtime_alias_test_list): Likewise.
10191         (vect_analyze_data_refs): Likewise.
10192         * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
10193         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
10194         (vect_analyze_scalar_cycles_1): Likewise.
10195         (vect_get_loop_niters): Likewise.
10196         (vect_analyze_loop_form_1): Likewise.
10197         (vect_update_vf_for_slp): Likewise.
10198         (vect_analyze_loop_operations): Likewise.
10199         (vect_analyze_loop): Likewise.
10200         (vectorizable_induction): Likewise.
10201         (vect_transform_loop): Likewise.
10202         * tree-vect-patterns.c (vect_pattern_recog): Likewise.
10203         * tree-vect-slp.c (vect_analyze_slp): Likewise.
10204         (vect_make_slp_decision): Likewise.
10205         (vect_detect_hybrid_slp): Likewise.
10206         (vect_slp_analyze_operations): Likewise.
10207         (vect_slp_bb): Likewise.
10208         * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
10209         (vectorizable_bswap): Likewise.
10210         (vectorizable_call): Likewise.
10211         (vectorizable_simd_clone_call): Likewise.
10212         (vectorizable_conversion): Likewise.
10213         (vectorizable_assignment): Likewise.
10214         (vectorizable_shift): Likewise.
10215         (vectorizable_operation): Likewise.
10216         * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
10218 2018-06-18  Martin Sebor  <msebor@redhat.com>
10220         PR tree-optimization/81384
10221         * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
10222         * builtins.c (expand_builtin_strnlen): New function.
10223         (expand_builtin): Call it.
10224         (fold_builtin_n): Avoid setting TREE_NO_WARNING.
10225         * builtins.def (BUILT_IN_STRNLEN): New.
10226         * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
10227         Warn for bounds in excess of maximum object size.
10228         * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
10229         single-value ranges.  Handle strnlen.
10230         (handle_builtin_strlen): Handle strnlen.
10231         (strlen_check_and_optimize_stmt): Same.
10232         * doc/extend.texi (Other Builtins): Document strnlen.
10234 2018-06-18  Maya Rashish  <coypu@sdf.org>
10236         * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
10237         (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
10238         (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
10240         * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
10241         here to ...
10242         * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
10244 2018-06-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10246         * tree.c (escaped_string::escape): Replace cast to char * by
10247         const_cast<char *> (unescaped).
10249 2018-06-18  Nick Clifton  <nickc@redhat.com>
10251         PR 84195
10252         * tree.c (escaped_string): New class.  Converts an unescaped
10253         string into its escaped equivalent.
10254         (warn_deprecated_use): Use the new class to convert the
10255         deprecation message, if present.
10256         (test_escaped_strings): New self test.
10257         (test_c_tests): Add test_escaped_strings.
10258         * doc/extend.texi (deprecated): Add a note that the
10259         deprecation message is affected by the -fmessage-length
10260         option, and that control characters will be escaped.
10261         (#pragma GCC error): Document this pragma.
10262         (#pragma GCC warning): Likewise.
10263         * doc/invoke.texi (-fmessage-length): Document this option's
10264         effect on the #warning and #error preprocessor directives and
10265         the deprecated attribute.
10267 2018-06-18  Eric Botcazou  <ebotcazou@adacore.com>
10269         * tree.c (decl_value_expr_lookup): Revert latest change.
10270         (decl_value_expr_insert): Likewise.
10272 2018-06-17  Eric Botcazou  <ebotcazou@adacore.com>
10274         * gimplify.c (nonlocal_vlas): Delete.
10275         (nonlocal_vla_vars): Likewise.
10276         (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
10277         referenced VLAs.
10278         (gimplify_body): Do not create and destroy nonlocal_vlas.
10279         * tree-nested.c: Include diagnostic.h.
10280         (use_pointer_in_frame): Tweak.
10281         (lookup_field_for_decl): Add assertion and declare the transformation.
10282         (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
10283         internal error when the reference is in a wrong context.  Do not
10284         create a debug decl by default.
10285         (note_nonlocal_block_vlas): Delete.
10286         (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
10287         (convert_local_reference_op) <PARM_DECL>: Skip the frame decl.  Do not
10288         create a debug decl by default.
10289         (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
10290         call is in a wrong context.
10291         (fixup_vla_decls): New function.
10292         (finalize_nesting_tree_1): Adjust comment.  Call fixup_vla_decls if no
10293         debug variables were created.
10294         * tree.c (decl_value_expr_lookup): Add checking assertion.
10295         (decl_value_expr_insert): Likewise.
10297 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
10299         PR middle-end/82479
10300         * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
10301         * tree-scalar-evolution.c (interpret_expr): Likewise.
10302         (expression_expensive_p): Likewise.
10303         * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
10304         * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
10305         (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
10306         (ssa_defined_by_minus_one_stmt_p): New.
10308 2018-06-16  Kugan Vivekanandarajah  <kuganv@linaro.org>
10310         PR middle-end/64946
10311         * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
10312         * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
10313         * dojump.c (do_jump): Likewise.
10314         * expr.c (expand_expr_real_2): Check operand type's sign.
10315         * fold-const.c (const_unop): Handle ABSU_EXPR.
10316         (fold_abs_const): Likewise.
10317         * gimple-pretty-print.c (dump_unary_rhs): Likewise.
10318         * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
10319         (strip_sign_op_1): Likesise.
10320         * match.pd: Add new pattern to generate ABSU_EXPR.
10321         * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
10322         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
10323         * tree-eh.c (operation_could_trap_helper_p): Likewise.
10324         * tree-inline.c (estimate_operator_cost): Likewise.
10325         * tree-pretty-print.c (dump_generic_node): Likewise.
10326         * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
10327         * tree.def (ABSU_EXPR): New.
10329 2018-06-16  Jakub Jelinek  <jakub@redhat.com>
10331         PR middle-end/86095
10332         * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
10333         documented as preserved for backward compatibility only.
10334         * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
10336         PR rtl-optimization/86108
10337         * bb-reorder.c (create_forwarder_block): Renamed to ...
10338         (create_eh_forwarder_block): ... this.  Split OLD_BB after labels and
10339         jump from new landing pad to the second part.
10340         (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
10341         Adjust callers.
10343 2018-06-15  Jakub Jelinek  <jakub@redhat.com>
10345         PR middle-end/85878
10346         * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
10347         check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
10348         Only call store_expr for halves if the mode is the same.
10350         PR middle-end/86123
10351         * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
10352         Fix up comment formatting.
10354 2018-06-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
10356         * typed-splay-tree.h (typed_splay_tree::remove): New function.
10357         (typed_splay_tree::closure,
10358         typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
10359         (typed_splay_tree::typed_splay_tree,
10360         typed_splay_tree::operator =): Declared private.
10361         (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
10362         typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
10363         typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
10364         typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
10365         typed_splay_tree::splay_tree_splay,
10366         typed_splay_tree::splay_tree_foreach_helper,
10367         typed_splay_tree::splay_tree_insert,
10368         typed_splay_tree::splay_tree_remove,
10369         typed_splay_tree::splay_tree_lookup,
10370         typed_splay_tree::splay_tree_predecessor,
10371         typed_splay_tree::splay_tree_successor,
10372         typed_splay_tree::splay_tree_min,
10373         typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
10374         (typed_splay_tree::root, typed_splay_tree::comp,
10375         typed_splay_tree::delete_key,
10376         typed_splay_tree::delete_value): New data members.
10377         * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
10378         typed_splay_tree::remove.
10380 2018-06-15  Matthew Fortune  <matthew.fortune@mips.com>
10382         * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
10383         -mginv and -mno-ginv to the assembler.
10384         * config/mips/mips.opt (-mcrc): New option.
10385         (-mginv): Likewise.
10386         * doc/invoke.text (-mcrc): Document.
10387         (-mginv): Likewise.
10389 2018-06-15  Nick Clifton  <nickc@redhat.com>
10391         PR 84195
10392         * tree.c (escaped_string): New class.  Converts an unescaped
10393         string into its escaped equivalent.
10394         (warn_deprecated_use): Use the new class to convert the
10395         deprecation message, if present.
10396         (test_escaped_strings): New self test.
10397         (test_c_tests): Add test_escaped_strings.
10398         * doc/extend.texi (deprecated): Add a note that the
10399         deprecation message is affected by the -fmessage-length
10400         option, and that control characters will be escaped.
10401         (#pragma GCC error): Document this pragma.
10402         (#pragma GCC warning): Likewise.
10403         * doc/invoke.texi (-fmessage-length): Document this option's
10404         effect on the #warning and #error preprocessor directives and
10405         the deprecated attribute.
10407 2018-06-15  Richard Biener  <rguenther@suse.de>
10409         * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
10410         here, also noting vector size used.
10411         * tree-vectorizer.c (vectorize_loops): Adjust.  Note vector
10412         size used in MSG_OPTIMIZED_LOCATIONS dump.
10413         (pass_slp_vectorize::execute): Adjust.
10415 2018-06-15  Claudiu Zissulescu  <claziss@synopsys.com>
10417         PR target/85968
10418         * config/arc/arc.c (arc_return_address_register): Fix
10419         if-condition.
10421 2018-06-15  Richard Biener  <rguenther@suse.de>
10423         PR middle-end/86159
10424         * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
10425         leave useless conversion stripping to force_gimple_operand_gsi.
10426         (gimplify_build2): Likewise.
10427         (gimplify_build1): Likewise.
10429 2018-06-15  Richard Biener  <rguenther@suse.de>
10431         PR middle-end/86076
10432         * tree-cfg.c (move_stmt_op): unshare invariant addresses
10433         before adjusting their block.
10435 2018-06-15  Sebastian Huber  <sebastian.huber@embedded-brains.de>
10437         * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
10438         multilibs for *-*-rtems*.
10439         * config/riscv/t-rtems: New file.
10441 2018-06-14  Jakub Jelinek  <jakub@redhat.com>
10443         PR middle-end/86122
10444         * match.pd ((A +- CST1) +- CST2): Punt if last resort
10445         unsigned_type_for returns NULL.
10447         PR target/85945
10448         * lower-subreg.c (find_decomposable_subregs): Don't decompose float
10449         subregs of multi-word pseudos unless the float mode has word size.
10451 2018-06-14  Richard Biener  <rguenther@suse.de>
10453         PR middle-end/86139
10454         * tree-vect-generic.c (build_word_mode_vector_type): Remove
10455         duplicate and harmful type_hash_canon.
10456         * tree.c (type_hash_canon): Assert we didn't find ourselves.
10458 2018-06-14  Richard Biener  <rguenther@suse.de>
10460         PR ipa/86124
10461         * tree-ssa-struct-alias.c (create_variable_info_for): Handle
10462         NULL cgraph_node.
10464 2018-06-14  Sebastian Huber  <sebastian.huber@embedded-brains.de>
10466         * config/rtems.h (STDINT_LONG32): Define.
10468 2018-06-13  Matthew Fortune  <matthew.fortune@mips.com>
10469             Prachi Godbole  <prachi.godbole@imgtec.com>
10471         * config/mips/mips-cpus.def: Define P6600.
10472         * config/mips/mips-tables.opt: Regenerate.
10473         * config/mips/mips.c (mips_ucbranch_type): New enum.
10474         (mips_rtx_cost_data): Add support for P6600.
10475         (mips_issue_rate): Likewise.
10476         (mips_multipass_dfa_lookahead): Likewise.
10477         (mips_avoid_hazard): Likewise.
10478         (mips_reorg_process_insns): Likewise.
10479         (mips_classify_branch_p6600): New function.
10480         * config/mips/mips.h (TUNE_P6600): New define.
10481         (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
10482         (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
10483         * config/mips/mips.md: Include p6600.md.
10484         (processor): Add p6600.
10485         * config/mips/p6600.md: New file.
10486         * doc/invoke.texi: Add p6600 to supported architectures.
10488 2018-06-13  Martin Sebor  <msebor@redhat.com>
10490         PR tree-optimization/86114
10491         * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
10492         of integer types.
10493         * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
10495 2018-06-13  Richard Biener  <rguenther@suse.de>
10497         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
10498         Properly set vector type of the intermediate stmt.
10499         * tree-vect-stmts.c (vectorizable_operation): The destination
10500         var always has vectype_out type.
10502 2018-06-13  Jeff Law  <law@redhat.com>
10504         * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
10505         integer 0 for argument to print_rtl_with_bb.
10506         (rl78_reorg): Likewise.
10508 2018-06-13  David Malcolm  <dmalcolm@redhat.com>
10510         * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
10511         from rtx to rtx_insn *.
10512         * config/bfin/bfin.c (hwloop_optimize): Likewise for local
10513         "label".
10514         (add_sched_insns_for_speculation): Likewise for local "target",
10515         converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
10516         * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
10517         from rtx_insn ** to rtx_code_label **.
10518         (reorg_emit_nops): Likewise.
10519         (c6x_reorg): Likewise for local "call_labels".
10520         * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
10521         rtx to rtx_insn *.
10522         * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
10523         rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
10524         the loops over LABEL_REFS.
10525         (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
10526         braf_label.
10527         (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
10528         (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
10529         (split_branches): Strengthen local "olabel" from rtx to
10530         rtx_insn *, adding a safe_as_a cast.
10531         * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
10532         to "rtx_insn *".
10533         (add_insn_after): Likewise for first two params.
10534         (add_insn_before): Likewise.
10535         (remove_insn): Likewise for param.
10536         (emit_pattern_before_noloc): Likewise for second and third params.
10537         (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
10538         (emit_call_insn_before_noloc): Likewise.
10539         (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
10540         to "rtx_insn *".
10541         (emit_barrier_before): Likewise.
10542         (emit_label_before): Strengthen "label" param from "rtx" to
10543         "rtx_code_label *".  Strengthen "before" param from "rtx" to
10544         "rtx_insn *".
10545         (emit_insn_after_1): Strengthen "after" param from "rtx" to
10546         "rtx_insn *".
10547         (emit_pattern_after_noloc): Likewise.
10548         (emit_insn_after_noloc): Likewise.
10549         (emit_jump_insn_after_noloc): Likewise.
10550         (emit_call_insn_after_noloc): Likewise.
10551         (emit_debug_insn_after_noloc): Likewise.
10552         (emit_barrier_after): Likewise.
10553         (emit_label_after): Likewise for both params.
10554         (emit_pattern_after_setloc): Likewise for "after" param.  Convert
10555         "loc" param from "int" to "location_t".
10556         (emit_insn_after_setloc): Likewise.
10557         (emit_jump_insn_after_setloc): Likewise.
10558         (emit_call_insn_after_setloc): Likewise.
10559         (emit_debug_insn_after_setloc): Likewise.
10560         (emit_pattern_before_setloc): Likewise for "before" param.  Convert
10561         "loc" param from "int" to "location_t".
10562         (emit_pattern_before): Convert NULL_RTX to NULL.
10563         (emit_insn_before_setloc): Convert "loc" param from "int" to
10564         "location_t".
10565         (emit_jump_insn_before_setloc): Likewise.
10566         (emit_call_insn_before_setloc): Likewise.
10567         (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
10568         rtx_insn *.  Convert "loc" param from "int" to "location_t".
10569         * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
10570         emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
10571         Convert 3rd param from "int" to "location_t".
10572         (emit_barrier_before, emit_barrier_after, next_real_insn):
10573         Strengthen param from rtx to rtx_insn *.
10574         (emit_label_before): Strengthen 1st param from "rtx" to
10575         "rtx_code_label *".  Strengthen 2nd param from "rtx" to
10576         "rtx_insn *".
10577         (emit_insn_after_noloc, emit_jump_insn_after_noloc,
10578         emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
10579         Strengthen 2nd param from "rtx" to "rtx_insn *".
10580         (emit_insn_after_setloc, emit_jump_insn_after_setloc)
10581         emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
10582         Likewise. Convert 3rd param from "int" to "location_t".
10583         (emit_label_after): Strengthen 1st param from "rtx" to
10584         "rtx_code_label *".
10585         (next_real_insn, remove_insn): Strengthen param from "rtx" to
10586         "rtx_insn *".
10587         (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
10588         from "rtx" to "rtx_insn *".
10590 2018-06-13  Jan Hubicka  <hubicka@gcc.gnu.org>
10592         * cgraph.c (cgraph_node::get_untransformed_body): Dump function
10593         bodies streamed in with -Q.
10594         * dumpfile.c (dump_files): Add lto-stream-out dump file.
10595         * dumpfile.h (tree_dump_index): Add lto_stream_out.
10596         * gimple-streamer-out.c: Include gimple-pretty-print.h
10597         (output_bb): Dump stmts streamed.
10598         * lto-section-out.c: Include print-tree.h
10599         (lto_begin_section): Dump sections created.
10600         (lto_output_decl_index): Dump decl encoded.
10601         * lto-streamer-out.c: Include print-tree.h
10602         (create_output_block): Dump output block created.
10603         (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
10604         (output_function): Dump function output.
10605         (output_constructor): Dump constructor streamed.
10606         (write_global_stream): Output indexes encoded.
10607         (produce_asm_for_decls): Dump streams encoded.
10608         * lto-streamer.c (streamer_dump_file): New global var.
10609         * lto-streamer.h (streamer_dump_file): Declare.
10610         * passes.c (ipa_write_summaries): Initialize streamer dump.
10611         * varpool.c (varpool_node::get_constructor): Dump constructors streamed
10612         in.
10614 2018-06-13  Eric Botcazou  <ebotcazou@adacore.com>
10616         PR target/86048
10617         * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
10618         offsets for register save directives.  Emit a second batch of save
10619         directives, if need be, when the function accesses prior frames.
10621 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
10623         * config/arc/fpu.md (fmasf4): Force operand to register.
10624         (fnmasf4): Likewise.
10626 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
10628         * config/arc/arc-protos.h (arc_pad_return): Remove.
10629         * config/arc/arc.c (machine_function): Remove force_short_suffix
10630         and size_reason.
10631         (arc_print_operand): Adjust printing of '&'.
10632         (arc_verify_short): Remove conditional printing of short suffix.
10633         (arc_final_prescan_insn): Remove reference to size_reason.
10634         (pad_return): New function.
10635         (arc_reorg): Call pad_return.
10636         (arc_pad_return): Remove.
10637         (arc_init_machine_status): Remove reference to force_short_suffix.
10638         * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
10639         (attr length): When attribute iscompact is true force to 2
10640         regardless; in the case of maybe check if we want to force the
10641         instruction to have 4 bytes length.
10642         (nopv): Change it to generate 4 byte long nop as well.
10643         (blockage): New pattern.
10644         (simple_return): Remove call to arc_pad_return.
10645         (p_return_i): Likewise.
10647 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
10649         * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
10651 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
10653         * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
10654         ARC cores.
10656 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
10658         * config/arc/arc.c (atomic_exchangesi): EX instruction is default
10659         for ARC700 and ARCv2.
10661 2018-06-13  Chenghua Xu <paul.hua.gm@gmail.com>
10663         PR target/86076
10664         * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
10665         operands[2] instead of operands[1].
10668 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
10670         * lra-constraints.c (simplify_operand_subreg): In the paradoxical
10671         case, check whether the outer register overlaps an unallocatable
10672         register, not just whether it fits the required class.
10674 2018-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
10676         * poly-int.h (can_div_trunc_p): Add new overload in which all values
10677         are poly_ints.
10678         * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
10679         (memrefs_conflict_p): Likewise.
10680         (init_alias_analysis): Likewise.
10681         * cfgexpand.c (expand_debug_expr): Likewise.
10682         * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
10683         * cse.c (fold_rtx): Likewise.
10684         * explow.c (adjust_stack, anti_adjust_stack): Likewise.
10685         * expr.c (emit_block_move_hints): Likewise.
10686         (clear_storage_hints, push_block, emit_push_insn): Likewise.
10687         (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
10688         (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
10689         (emit_group_store): Likewise.
10690         (find_args_size_adjust): Use strip_offset.  Use rtx_to_poly_int64
10691         to read the PRE/POST_MODIFY increment.
10692         * calls.c (store_one_arg): Use strip_offset.
10693         * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
10694         poly_int_rtx_p.
10695         (set_noop_p): Use rtx_to_poly_int64 for the elements selected
10696         by a VEC_SELECT.
10697         * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
10698         (simplify_binary_operation_1): Extend CONST_INT handling to
10699         poly_int_rtx_p.
10700         * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
10701         than a HOST_WIDE_INT.
10702         (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
10703         poly_int64.
10704         (adjust_mems, add_stores): Update accodingly.
10705         (vt_canonicalize_addr): Track polynomial offsets.
10706         (emit_note_insn_var_location): Likewise.
10707         (vt_add_function_parameter): Likewise.
10708         (vt_initialize): Likewise.
10710 2018-06-12  Jeff Law  <law@redhat.com>
10712         * config.gcc (alpha*-*-freebsd*): Remove.
10713         * config/alpha/freebsd.h: Remove.
10715 2018-06-12  David Malcolm  <dmalcolm@redhat.com>
10717         PR other/69968
10718         * spellcheck-tree.c (levenshtein_distance): Rename to...
10719         (get_edit_distance): ...this, and update for underlying renaming.
10720         * spellcheck-tree.h (levenshtein_distance): Rename to...
10721         (get_edit_distance): ...this.
10722         * spellcheck.c (levenshtein_distance): Rename to...
10723         (get_edit_distance): ...this.  Convert from Levenshtein distance
10724         to Damerau-Levenshtein distance by supporting transpositions of
10725         adjacent characters.  Rename "v1" to "v_next" and "v0" to
10726         "v_one_ago".
10727         (selftest::levenshtein_distance_unit_test_oneway): Rename to...
10728         (selftest::test_edit_distance_unit_test_oneway): ...this, and
10729         update for underlying renaming.
10730         (selftest::levenshtein_distance_unit_test): Rename to...
10731         (selftest::test_get_edit_distance_unit): ...this, and update for
10732         underlying renaming.
10733         (selftest::test_find_closest_string): Add example from PR 69968
10734         where transposition helps
10735         (selftest::test_metric_conditions): Update for renaming.
10736         (selftest::test_metric_conditions): Likewise.
10737         (selftest::spellcheck_c_tests): Likewise.
10738         * spellcheck.h (levenshtein_distance): Rename both overloads to...
10739         (get_edit_distance): ...this.
10740         (best_match::consider): Update for renaming.
10742 2018-06-12  Martin Sebor  <msebor@redhat.com>
10744         PR tree-optimization/85259
10745         * builtins.c (compute_objsize): Handle constant offsets.
10746         * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
10747         true iff a warning has been issued.
10748         * gimple.h (gimple_nonartificial_location): New function.
10749         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
10750         gimple_nonartificial_location and handle -Wno-system-headers.
10751         (handle_builtin_stxncpy): Same.
10753 2018-06-12  Martin Sebor  <msebor@redhat.com>
10755         PR c/85931
10756         * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
10758 2018-06-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
10760         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
10761         BUILTIN_VEC_XST entries for pointer to double and long long.
10763 2018-06-12  H.J. Lu  <hongjiu.lu@intel.com>
10765         PR target/85990
10766         * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
10767         Update comments.
10768         * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
10769         Likewise.
10771 2018-06-12  Martin Liska  <mliska@suse.cz>
10773         * doc/options.texi: Document IntegerRange.
10775 2018-06-12  Martin Liska  <mliska@suse.cz>
10777         * config/i386/i386.opt: Make MPX-related options as Deprecated.
10778         * opt-functions.awk: Handle Deprecated flag.
10779         * opts-common.c (decode_cmdline_option): Handle cl_deprecated
10780         and report error.
10781         (read_cmdline_option): Report warning for a deprecated option.
10782         * opts.h (struct cl_option): Add new field cl_deprecated.
10783         (CL_ERR_DEPRECATED): New.
10785 2018-06-12  Martin Liska  <mliska@suse.cz>
10787         * doc/options.texi: Document Deprecated option flag.
10789 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
10791         * config/arc/arc-arch.h (arc_extras): New enum.
10792         (arc_cpu_t):Add field extra.
10793         (arc_cpu_types): Consider the extras.
10794         * config/arc/arc-cpus.def: Add extras info.
10795         * config/arc/arc-opts.h (processor_type): Consider extra field.
10796         * config/arc/arc.c (arc_override_options): Handle extra field.
10798 2018-06-12  Claudiu Zissulescu  <claziss@synopsys.com>
10800         * config/arc/arc-arch.h: Update ARC_OPTX macro.
10801         * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
10802         field.
10803         * config/arc/arc.c (arc_init): Update pic warning.
10804         (irq_range): Update irq range parsing warnings.
10805         (arc_override_options): Update various warning messages.
10806         (arc_handle_aux_attribute): Likewise.
10808 2018-06-12  Robert Suchanek  <robert.suchanek@mips.com>
10810         * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
10812 2018-06-12  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
10814         * doc/sourcebuild.texi: Document usage of line number 0 in verify
10815         compiler messages directives.
10817 2018-06-12  Matthew Fortune  <mfortune@gmail.com>
10819         * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
10820         * config/mips/mips-tables.opt: Regenerate.
10821         * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
10822         mips64r6.
10823         * doc/invoke.texi: Document -march=i6500.
10825 2018-06-12  Prachi Godbole  <prachi.godbole@imgtec.com>
10827         * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
10828         (i6400_gpmul): Add cpu_unit.
10829         (i6400_gpdiv): Likewise.
10830         (i6400_msa_add_d): Update reservations.
10831         (i6400_msa_int_add) Likewise.
10832         (i6400_msa_short_logic3) Likewise.
10833         (i6400_msa_short_logic2) Likewise.
10834         (i6400_msa_short_logic) Likewise.
10835         (i6400_msa_move) Likewise.
10836         (i6400_msa_cmp) Likewise.
10837         (i6400_msa_short_float2) Likewise.
10838         (i6400_msa_div_d) Likewise.
10839         (i6400_msa_long_logic1) Likewise.
10840         (i6400_msa_long_logic2) Likewise.
10841         (i6400_msa_mult) Likewise.
10842         (i6400_msa_long_float2) Likewise.
10843         (i6400_msa_long_float4) Likewise.
10844         (i6400_msa_long_float5) Likewise.
10845         (i6400_msa_long_float8) Likewise.
10846         (i6400_fpu_fadd): Include frint type.
10847         (i6400_fpu_store): New define_insn_reservation.
10848         (i6400_fpu_load): Likewise.
10849         (i6400_fpu_move): Likewise.
10850         (i6400_fpu_fcmp): Likewise.
10851         (i6400_fpu_fmadd): Likewise.
10852         (i6400_int_mult): Include imul3nc type and update reservation.
10853         (i6400_int_div): Include idiv3 type and update reservation.
10854         (i6400_int_load): Update to check type not move_type.
10855         (i6400_int_store): Likewise.
10856         (i6400_int_prefetch): Set zero latency.
10858 2018-06-12  Eric Botcazou  <ebotcazou@adacore.com>
10860         * gcc.c: Document new %@{...} sequence.
10861         (LINK_COMMAND_SPEC): Use it for the -L switches.
10862         (cpp_unique_options): Use it for the -I switches.
10863         (at_file_argbuf): New global variable.
10864         (in_at_file): Likewise.
10865         (alloc_args): Create at_file_argbuf.
10866         (clear_args): Truncate at_file_argbuf.
10867         (store_arg): If in_at_file, push the argument onto at_file_argbuf.
10868         (open_at_file): New function.
10869         (close_at_file): Likewise.
10870         (create_at_file): Delete.
10871         (do_spec_1) <'i'>: Use open_at_file/close_at_file.
10872         <'o'>: Likewise.
10873         <'@'>: New case.
10874         (validate_switches_from_spec): Deal with %@{...} sequence.
10875         (validate_switches): Likewise.
10876         (driver::finalize): Call clear_args.
10878 2018-06-11  Rasmus Villemoes  <rasmus.villemoes@prevas.dk>
10880         * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
10882 2018-06-11  Martin Sebor  <msebor@redhat.com>
10884         * doc/invoke.texi (-Wall): List -Wc++17-compat.
10885         (Wno-class-memaccess): Add @opindex.
10886         (Wno-templates, Wno-multiple-inheritance): Same.
10887         (Wno-virtual-inheritance, Wno-namespaces): Same.
10888         (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
10889         (Wno-format-overflow, Wno-format-truncation): Same.
10890         (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
10891         (Wno-alloc-size-larger-than, Wframe-larger-than): Same
10892         (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
10893         (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
10894         (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
10895         (Wno-misspelled-isr): Same.
10897 2018-06-11  Martin Sebor  <msebor@redhat.com>
10899         * PR tree-optimization/86083
10900         * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
10902 2018-06-11  Zhouyi Zhou <zhouzhouyi@gmail.com>
10904         * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
10906 2018-06-11  Segher Boessenkool  <segher@kernel.crashing.org>
10908         PR target/85755
10909         * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
10910         on the correct operand.
10911         (*movdi_internal64): Ditto.
10913 2018-06-11  Martin Liska  <mliska@suse.cz>
10915         PR tree-optimization/86089
10916         * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
10918 2018-06-11  Julia Koval  <julia.koval@intel.com>
10920         * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
10921         _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
10922         * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
10924 2018-06-11  Olivier Hainque  <hainque@adacore.com>
10926         * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
10927         for Ada with strict dwarf2.
10929 2018-06-08  Peter Bergner  <bergner@vnet.ibm.com>
10931         PR target/85755
10932         * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
10933         addresses.
10935 2018-06-08  Jan Hubicka  <hubicka@ucw.cz>
10937         * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
10939 2018-06-08  David Edelsohn  <dje.gcc@gmail.com>
10941         * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
10942         TARGET_ELF.
10944 2018-06-08  Martin Liska  <mliska@suse.cz>
10946         * tree-cfg.h (debug_function): Fix argument type to match
10947         implementation.
10949 2018-06-08  Martin Liska  <mliska@suse.cz>
10951         * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
10952         Remove usage of MPX-related (and removed) fields.
10953         * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
10955 2018-06-08  David Malcolm  <dmalcolm@redhat.com>
10957         * cfg.c (debug): Use TDF_NONE rather than 0.
10958         * cfghooks.c (debug): Likewise.
10959         * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
10960         (struct dump_option_value_info): Convert to...
10961         (struct kv_pair): ...this template type.
10962         (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
10963         rather than 0.
10964         (optinfo_verbosity_options): Likewise.
10965         (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
10966         OPTGROUP_NONE.
10967         (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
10968         than int for "optgroup_flags" param.
10969         (dump_generic_expr_loc): Use dump_flags_t rather than int for
10970         "dump_kind" param.
10971         (dump_dec): Likewise.
10972         (dump_finish): Use TDF_NONE rather than 0.
10973         (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
10974         rather than int for "optgroup_flags" param.  Use TDF_NONE rather
10975         than 0.  Update for change to option_ptr.
10976         (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
10977         to optgroup_flags_t *.  Use TDF_NONE and OPTGROUP_NONE rather than
10978         0.  Update for changes to optinfo_verbosity_options and
10979         optgroup_options.
10980         (opt_info_switch_p): Convert optgroup_flags from int to
10981         optgroup_flags_t.
10982         (dump_basic_block): Use dump_flags_t rather than int
10983         for "dump_kind" param.
10984         * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
10985         TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
10986         TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
10987         TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
10988         TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
10989         MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
10990         TDF_NONE): Convert from macros to...
10991         (enum dump_flag): ...this new enum.
10992         (dump_flags_t): Update to use enum.
10993         (operator|, operator&, operator~, operator|=, operator&=):
10994         Implement for dump_flags_t.
10995         (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
10996         OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
10997         Convert from macros to...
10998         (enum optgroup_flag): ...this new enum.
10999         (optgroup_flags_t): New typedef.
11000         (operator|, operator|=): Implement for optgroup_flags_t.
11001         (struct dump_file_info): Convert field "alt_flags" to
11002         dump_flags_t.  Convert field "optgroup_flags" to
11003         optgroup_flags_t.
11004         (dump_basic_block): Use dump_flags_t rather than int for param.
11005         (dump_generic_expr_loc): Likewise.
11006         (dump_dec): Likewise.
11007         (dump_register): Convert param "optgroup_flags" to
11008         optgroup_flags_t.
11009         (opt_info_enable_passes): Likewise.
11010         * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
11011         than 0.
11012         * gimple-pretty-print.c (debug): Likewise.
11013         * gimple-ssa-store-merging.c (bswap_replace): Likewise.
11014         (merged_store_group::apply_stores): Likewise.
11015         * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
11016         * gimple.c (verify_gimple_pp): Likewise.
11017         * graphite-poly.c (print_pbb_body): Likewise.
11018         * passes.c (pass_manager::register_one_dump_file): Convert
11019         local "optgroup_flags" to optgroup_flags_t.
11020         * print-tree.c (print_node): Use TDF_NONE rather than 0.
11021         (debug): Likewise.
11022         (debug_body): Likewise.
11023         * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
11024         to optgroup_flags_t.
11025         * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
11026         than 0.
11027         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
11028         (convert_mult_to_fma): Likewise.
11029         * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
11030         * tree-ssa-sccvn.c (vn_eliminate): Likewise.
11031         * tree-vect-data-refs.c (dump_lower_bound): Convert param
11032         "dump_kind" to dump_flags_t.
11034 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
11036         * config/rs6000/rs6000.c (min, max): Delete.
11038 2018-06-08  Segher Boessenkool  <segher@kernel.crashing.org>
11040         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
11041         -mabi=spe and -mabi=no-spe.
11043 2018-06-08  Martin Liska  <mliska@suse.cz>
11045         * ipa-pure-const.c (propagate_pure_const): Use ::get at places
11046         where we expect an existing summary.
11048 2018-06-08  Martin Liska  <mliska@suse.cz>
11050         * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
11051         * ipa-inline.h (estimate_edge_growth): Likewise.
11053 2018-06-08  Martin Liska  <mliska@suse.cz>
11055         * cgraph.c (function_version_hasher::hash): Use
11056         cgraph_node::get_uid ().
11057         (function_version_hasher::equal):
11058         * cgraph.h (cgraph_node::get_uid): New method.
11059         * ipa-inline.c (update_caller_keys): Use
11060         cgraph_node::get_uid ().
11061         (update_callee_keys): Likewise.
11062         * ipa-utils.c (searchc): Likewise.
11063         (ipa_reduced_postorder): Likewise.
11064         * lto-cgraph.c (input_node): Likewise.
11065         * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
11066         * symbol-summary.h (symtab_insertion): Likewise.
11067         (symtab_removal): Likewise.
11068         (symtab_duplication): Likewise.
11069         * tree-pretty-print.c (dump_function_header): Likewise.
11070         * tree-sra.c (convert_callers_for_node): Likewise.
11072 2018-06-08  Martin Liska  <mliska@suse.cz>
11074         * cgraph.c (symbol_table::create_edge): Always assign a new
11075         unique number.
11076         (symbol_table::free_edge): Do not recycle numbers.
11077         * cgraph.h (cgraph_edge::get): New method.
11078         * symbol-summary.h (symtab_removal): Use it.
11079         (symtab_duplication): Likewise.
11080         (call_summary::hashable_uid): Remove.
11082 2018-06-08  Martin Liska  <mliska@suse.cz>
11084         * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
11085         (initialize_growth_caches): Remove.
11086         (free_growth_caches): Likewise.
11087         (do_estimate_edge_time): Use edge_growth_cache.
11088         (do_estimate_edge_size): Likewise.
11089         (do_estimate_edge_hints): Likewise.
11090         * ipa-inline.c (reset_edge_caches): Likewise.
11091         (recursive_inlining): Likewise.
11092         (inline_small_functions): Likewise.
11093         * ipa-inline.h (initialize_growth_caches): Remove.
11094         (estimate_edge_size): Likewise.
11095         (estimate_edge_time): Likewise.
11096         (estimate_edge_hints): Likewise.
11097         (reset_edge_growth_cache): Likewise.
11098         * symbol-summary.h (call_summary::remove): New method.
11100 2018-06-08  Martin Liska  <mliska@suse.cz>
11102         * ipa-cp.c (class edge_clone_summary): New summary.
11103         (grow_edge_clone_vectors): Remove.
11104         (ipcp_edge_duplication_hook): Remove.
11105         (class edge_clone_summary_t): New call_summary class.
11106         (ipcp_edge_removal_hook): Remove.
11107         (edge_clone_summary_t::duplicate): New function.
11108         (get_next_cgraph_edge_clone): Use edge_clone_summaries.
11109         (create_specialized_node): Likewise.
11110         (ipcp_driver): Initialize edge_clone_summaries and do not
11111         register hooks.
11113 2018-06-08  Martin Liska  <mliska@suse.cz>
11115         * symbol-summary.h (get): New function.
11116         (call_summary::m_initialize_when_cloning): New class member.
11118 2018-06-08  Martin Liska  <mliska@suse.cz>
11120         * cgraph.c (cgraph_node::remove): Do not recycle uid.
11121         * cgraph.h (symbol_table::release_symbol): Do not pass uid.
11122         (symbol_table::allocate_cgraph_symbol): Do not set uid.
11123         * passes.c (uid_hash_t): Record removed_nodes by their uids.
11124         (remove_cgraph_node_from_order): Use the removed_nodes set.
11125         (do_per_function_toporder): Likwise.
11126         * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
11127         instead of summary_uid.
11128         (symtab_removal): Likewise.
11129         (symtab_duplication): Likewise.
11131 2018-06-08  Martin Liska  <mliska@suse.cz>
11133         * ipa-cp.c (ipcp_store_bits_results): Use
11134         ipcp_transformation_sum.
11135         (ipcp_store_vr_results): Likewise.
11136         * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
11137         to ...
11138         (ipcp_transformation_initialize): ... this.
11139         (ipa_set_node_agg_value_chain):
11140         (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
11141         (write_ipcp_transformation_info): Likewise.
11142         (read_ipcp_transformation_info): Likewise.
11143         (ipcp_update_bits): Likewise.
11144         (ipcp_update_vr): Likewise.
11145         (ipcp_transform_function): Likewise.
11146         * ipa-prop.h: Rename ipcp_transformation_summary to
11147         ipcp_transformation.
11148         (class ipcp_transformation_t): New function summary.
11149         (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
11150         (ipa_get_agg_replacements_for_node): Likewise.
11152 2018-06-08  Martin Liska  <mliska@suse.cz>
11154         * ipa-pure-const.c (struct funct_state_d): Do it class instead
11155         of struct.
11156         (class funct_state_summary_t): New function_summary class.
11157         (has_function_state): Remove.
11158         (get_function_state): Likewise.
11159         (set_function_state): Likewise.
11160         (add_new_function): Likewise.
11161         (funct_state_summary_t::insert): New function.
11162         (duplicate_node_data): Remove.
11163         (remove_node_data): Remove.
11164         (funct_state_summary_t::duplicate): New function.
11165         (register_hooks): Create new funct_state_summaries.
11166         (pure_const_generate_summary): Use it.
11167         (pure_const_write_summary): Likewise.
11168         (pure_const_read_summary): Likewise.
11169         (propagate_pure_const): Likewise.
11170         (propagate_nothrow): Likewise.
11171         (dump_malloc_lattice): Likewise.
11172         (propagate_malloc): Likewise.
11173         (execute): Do not register hooks, just remove summary
11174         instead.
11175         (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
11176         constructor.
11178 2018-06-08  Martin Liska  <mliska@suse.cz>
11180         * ipa-reference.c (remove_node_data): Remove.
11181         (duplicate_node_data): Likewise.
11182         (class ipa_ref_var_info_summary_t): New class.
11183         (class ipa_ref_opt_summary_t): Likewise.
11184         (get_reference_vars_info): Use ipa_ref_var_info_summaries.
11185         (get_reference_optimization_summary): Use
11186         ipa_ref_opt_sum_summaries.
11187         (set_reference_vars_info): Remove.
11188         (set_reference_optimization_summary): Likewise.
11189         (ipa_init): Create summaries.
11190         (init_function_info): Use function summary.
11191         (ipa_ref_opt_summary_t::duplicate): New function.
11192         (ipa_ref_opt_summary_t::remove): New function.
11193         (get_read_write_all_from_node): Fix GNU coding style.
11194         (propagate): Use function summary.
11195         (write_node_summary_p): Fix GNU coding style.
11196         (stream_out_bitmap): Likewise.
11197         (ipa_reference_read_optimization_summary): Use function summary.
11198         (ipa_reference_c_finalize): Do not release hooks.
11200 2018-06-08  Martin Liska  <mliska@suse.cz>
11202         * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
11203         (analyze_function_body): Extract multiple calls of get_create.
11204         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
11205         * ipa-inline.c (recursive_inlining): Use ::get method.
11206         * ipa-inline.h (estimate_edge_growth): Likewise.
11208 2018-06-08  Martin Liska  <mliska@suse.cz>
11210         * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
11211         HSA_INVALID.
11212         (hsa_function_summary::hsa_function_summary): Use the new enum
11213         value.
11214         (hsa_gpu_implementation_p): Use hsa_summaries::get.
11215         * hsa-gen.c (hsa_get_host_function): Likewise.
11216         (get_brig_function_name): Likewise.
11217         * ipa-hsa.c (process_hsa_functions): Likewise.
11218         (ipa_hsa_write_summary): Likewise.
11219         * symbol-summary.h (symtab_duplication): Use ::get function/
11220         (get): New function.
11222 2018-06-08  Martin Liska  <mliska@suse.cz>
11224         * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
11225         of get.
11226         * hsa-common.c (hsa_summary_t::link_functions): Likewise.
11227         (hsa_register_kernel): Likewise.
11228         * hsa-common.h (hsa_gpu_implementation_p): Likewise.
11229         * hsa-gen.c (hsa_get_host_function): Likewise.
11230         (get_brig_function_name): Likewise.
11231         (generate_hsa): Likewise.
11232         (pass_gen_hsail::execute): Likewise.
11233         * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
11234         (devirtualization_time_bonus): Likewise.
11235         (ipcp_propagate_stage): Likewise.
11236         * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
11237         (edge_set_predicate): Likewise.
11238         (evaluate_conditions_for_known_args): Likewise.
11239         (evaluate_properties_for_edge): Likewise.
11240         (ipa_fn_summary::reset): Likewise.
11241         (ipa_fn_summary_t::duplicate): Likewise.
11242         (dump_ipa_call_summary): Likewise.
11243         (ipa_dump_fn_summary): Likewise.
11244         (analyze_function_body): Likewise.
11245         (compute_fn_summary): Likewise.
11246         (estimate_edge_devirt_benefit): Likewise.
11247         (estimate_edge_size_and_time): Likewise.
11248         (estimate_calls_size_and_time): Likewise.
11249         (estimate_node_size_and_time): Likewise.
11250         (inline_update_callee_summaries): Likewise.
11251         (remap_edge_change_prob): Likewise.
11252         (remap_edge_summaries): Likewise.
11253         (ipa_merge_fn_summary_after_inlining): Likewise.
11254         (ipa_update_overall_fn_summary): Likewise.
11255         (read_ipa_call_summary): Likewise.
11256         (inline_read_section): Likewise.
11257         (write_ipa_call_summary): Likewise.
11258         (ipa_fn_summary_write): Likewise.
11259         (ipa_free_fn_summary): Likewise.
11260         * ipa-hsa.c (process_hsa_functions): Likewise.
11261         (ipa_hsa_write_summary): Likewise.
11262         (ipa_hsa_read_section): Likewise.
11263         * ipa-icf.c (sem_function::merge): Likewise.
11264         * ipa-inline-analysis.c (simple_edge_hints): Likewise.
11265         (do_estimate_edge_time): Likewise.
11266         (estimate_size_after_inlining): Likewise.
11267         (estimate_growth): Likewise.
11268         (growth_likely_positive): Likewise.
11269         * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
11270         (inline_call): Likewise.
11271         * ipa-inline.c (caller_growth_limits): Likewise.
11272         (can_inline_edge_p): Likewise.
11273         (can_inline_edge_by_limits_p): Likewise.
11274         (compute_uninlined_call_time): Likewise.
11275         (compute_inlined_call_time): Likewise.
11276         (want_inline_small_function_p): Likewise.
11277         (edge_badness): Likewise.
11278         (update_caller_keys): Likewise.
11279         (update_callee_keys): Likewise.
11280         (recursive_inlining): Likewise.
11281         (inline_small_functions): Likewise.
11282         (inline_to_all_callers_1): Likewise.
11283         (dump_overall_stats): Likewise.
11284         (early_inline_small_functions): Likewise.
11285         (early_inliner): Likewise.
11286         * ipa-inline.h (estimate_edge_growth): Likewise.
11287         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
11288         * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
11289         * ipa-prop.h (IPA_NODE_REF): Likewise.
11290         (IPA_EDGE_REF): Likewise.
11291         * ipa-pure-const.c (malloc_candidate_p): Likewise.
11292         (propagate_malloc): Likewise.
11293         * ipa-split.c (execute_split_functions): Likewise.
11294         * symbol-summary.h: Rename get to get_create.
11295         (get): Likewise.
11296         (get_create): Likewise.
11297         * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
11299 2018-06-08  Martin Liska  <mliska@suse.cz>
11301         * symbol-summary.h (release): Move definition out of class
11302         declaration.
11303         (symtab_removal): Likewise.
11304         (symtab_duplication): Likewise.
11306 2018-06-08  Martin Liska  <mliska@suse.cz>
11308         * symbol-summary.h (function_summary): Move constructor
11309         implementation out of class declaration.
11310         (release): Likewise.
11311         (symtab_insertion): Likewise.
11312         (symtab_removal): Likewise.
11313         (symtab_duplication): Likewise.
11314         (get): Likewise.
11316 2018-06-08  Martin Liska  <mliska@suse.cz>
11318         * Makefile.in: Remove support for MPX (macros, related functions,
11319         fields in cgraph_node, ...).
11320         * builtin-types.def (BT_BND): Likewise.
11321         (BT_FN_BND_CONST_PTR): Likewise.
11322         (BT_FN_CONST_PTR_BND): Likewise.
11323         (BT_FN_VOID_PTR_BND): Likewise.
11324         (BT_FN_BND_CONST_PTR_SIZE): Likewise.
11325         (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
11326         * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
11327         (expand_builtin_mempcpy_with_bounds): Likewise.
11328         (expand_builtin_memset_with_bounds): Likewise.
11329         (expand_builtin_memset_args): Likewise.
11330         (std_expand_builtin_va_start): Likewise.
11331         (expand_builtin): Likewise.
11332         (expand_builtin_with_bounds): Likewise.
11333         * builtins.def (DEF_BUILTIN_CHKP): Likewise.
11334         (DEF_LIB_BUILTIN_CHKP): Likewise.
11335         (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
11336         (DEF_CHKP_BUILTIN): Likewise.
11337         (BUILT_IN_MEMCPY): Likewise.
11338         (BUILT_IN_MEMMOVE): Likewise.
11339         (BUILT_IN_MEMPCPY): Likewise.
11340         (BUILT_IN_MEMSET): Likewise.
11341         (BUILT_IN_STPCPY): Likewise.
11342         (BUILT_IN_STRCAT): Likewise.
11343         (BUILT_IN_STRCHR): Likewise.
11344         (BUILT_IN_STRCPY): Likewise.
11345         (BUILT_IN_STRLEN): Likewise.
11346         (BUILT_IN_MEMCPY_CHK): Likewise.
11347         (BUILT_IN_MEMMOVE_CHK): Likewise.
11348         (BUILT_IN_MEMPCPY_CHK): Likewise.
11349         (BUILT_IN_MEMSET_CHK): Likewise.
11350         (BUILT_IN_STPCPY_CHK): Likewise.
11351         (BUILT_IN_STRCAT_CHK): Likewise.
11352         (BUILT_IN_STRCPY_CHK): Likewise.
11353         * calls.c (store_bounds): Likewise.
11354         (emit_call_1): Likewise.
11355         (special_function_p): Likewise.
11356         (maybe_warn_nonstring_arg): Likewise.
11357         (initialize_argument_information): Likewise.
11358         (finalize_must_preallocate): Likewise.
11359         (compute_argument_addresses): Likewise.
11360         (expand_call): Likewise.
11361         * cfgexpand.c (expand_call_stmt): Likewise.
11362         (expand_return): Likewise.
11363         (expand_gimple_stmt_1): Likewise.
11364         (pass_expand::execute): Likewise.
11365         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
11366         (cgraph_node::remove): Likewise.
11367         (cgraph_node::dump): Likewise.
11368         (cgraph_node::verify_node): Likewise.
11369         * cgraph.h (chkp_function_instrumented_p): Likewise.
11370         (symtab_node::get_alias_target): Likewise.
11371         (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
11372         (cgraph_local_p): Likewise.
11373         * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
11374         (cgraph_edge::rebuild_references): Likewise.
11375         * cgraphunit.c (varpool_node::finalize_decl): Likewise.
11376         (walk_polymorphic_call_targets): Likewise.
11377         (cgraph_node::expand_thunk): Likewise.
11378         (symbol_table::output_weakrefs): Likewise.
11379         * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
11380         (ix86_handle_option): Likewise.
11381         * config/i386/constraints.md: Likewise.
11382         * config/i386/i386-builtin-types.def (BND): Likewise.
11383         (VOID): Likewise.
11384         (PVOID): Likewise.
11385         (ULONG): Likewise.
11386         * config/i386/i386-builtin.def (BDESC_END): Likewise.
11387         (BDESC_FIRST): Likewise.
11388         (BDESC): Likewise.
11389         * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
11390         * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
11391         * config/i386/i386.c (enum reg_class): Likewise.
11392         (ix86_target_string): Likewise.
11393         (ix86_option_override_internal): Likewise.
11394         (ix86_conditional_register_usage): Likewise.
11395         (ix86_valid_target_attribute_inner_p): Likewise.
11396         (ix86_set_indirect_branch_type): Likewise.
11397         (ix86_set_current_function): Likewise.
11398         (ix86_function_arg_regno_p): Likewise.
11399         (init_cumulative_args): Likewise.
11400         (ix86_function_arg_advance): Likewise.
11401         (ix86_function_arg): Likewise.
11402         (ix86_pass_by_reference): Likewise.
11403         (ix86_function_value_regno_p): Likewise.
11404         (ix86_function_value_1): Likewise.
11405         (ix86_function_value_bounds): Likewise.
11406         (ix86_return_in_memory): Likewise.
11407         (ix86_setup_incoming_vararg_bounds): Likewise.
11408         (ix86_va_start): Likewise.
11409         (indirect_thunk_need_prefix): Likewise.
11410         (print_reg): Likewise.
11411         (ix86_print_operand): Likewise.
11412         (ix86_expand_call): Likewise.
11413         (ix86_output_function_return): Likewise.
11414         (reg_encoded_number): Likewise.
11415         (BDESC_VERIFYS): Likewise.
11416         (ix86_init_mpx_builtins): Likewise.
11417         (ix86_init_builtins): Likewise.
11418         (ix86_emit_cmove): Likewise.
11419         (ix86_emit_move_max): Likewise.
11420         (ix86_expand_builtin): Likewise.
11421         (ix86_builtin_mpx_function): Likewise.
11422         (ix86_get_arg_address_for_bt): Likewise.
11423         (ix86_load_bounds): Likewise.
11424         (ix86_store_bounds): Likewise.
11425         (ix86_load_returned_bounds): Likewise.
11426         (ix86_store_returned_bounds): Likewise.
11427         (ix86_class_likely_spilled_p): Likewise.
11428         (ix86_hard_regno_mode_ok): Likewise.
11429         (x86_order_regs_for_local_alloc): Likewise.
11430         (ix86_mitigate_rop): Likewise.
11431         (ix86_bnd_prefixed_insn_p): Likewise.
11432         (ix86_mpx_bound_mode): Likewise.
11433         (ix86_make_bounds_constant): Likewise.
11434         (ix86_initialize_bounds): Likewise.
11435         (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
11436         (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
11437         (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
11438         (TARGET_STORE_RETURNED_BOUNDS): Likewise.
11439         (TARGET_CHKP_BOUND_MODE): Likewise.
11440         (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
11441         (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
11442         (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
11443         (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
11444         * config/i386/i386.h (TARGET_MPX): Likewise.
11445         (TARGET_MPX_P): Likewise.
11446         (VALID_BND_REG_MODE): Likewise.
11447         (FIRST_BND_REG): Likewise.
11448         (LAST_BND_REG): Likewise.
11449         (enum reg_class): Likewise.
11450         (BND_REG_P): Likewise.
11451         (BND_REGNO_P): Likewise.
11452         (BNDmode): Likewise.
11453         (ADJUST_INSN_LENGTH): Likewise.
11454         * config/i386/i386.md: Likewise.
11455         * config/i386/i386.opt: Likewise.
11456         * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
11457         (defined): Likewise.
11458         (LINK_MPX): Likewise.
11459         (MPX_SPEC): Likewise.
11460         (LIBMPX_SPEC): Likewise.
11461         (LIBMPXWRAPPERS_SPEC): Likewise.
11462         (CHKP_SPEC): Likewise.
11463         * config/i386/predicates.md: Likewise.
11464         * dbxout.c (dbxout_type): Likewise.
11465         * doc/extend.texi: Likewise.
11466         * doc/invoke.texi: Likewise.
11467         * doc/md.texi: Likewise.
11468         * doc/tm.texi: Likewise.
11469         * doc/tm.texi.in: Likewise.
11470         * dwarf2out.c (is_base_type): Likewise.
11471         (gen_formal_types_die): Likewise.
11472         (gen_subprogram_die): Likewise.
11473         (gen_type_die_with_usage): Likewise.
11474         (gen_decl_die): Likewise.
11475         (dwarf2out_late_global_decl): Likewise.
11476         * expr.c (expand_assignment): Likewise.
11477         (emit_storent_insn): Likewise.
11478         (store_expr_with_bounds): Likewise.
11479         (store_expr): Likewise.
11480         (expand_expr_real_1): Likewise.
11481         * expr.h (store_expr_with_bounds): Likewise.
11482         * function.c (use_register_for_decl): Likewise.
11483         (struct bounds_parm_data): Likewise.
11484         (assign_parms_augmented_arg_list): Likewise.
11485         (assign_parm_find_entry_rtl): Likewise.
11486         (assign_parm_is_stack_parm): Likewise.
11487         (assign_parm_load_bounds): Likewise.
11488         (assign_bounds): Likewise.
11489         (assign_parms): Likewise.
11490         (expand_function_start): Likewise.
11491         * gcc.c (CHKP_SPEC): Likewise.
11492         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
11493         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
11494         (wrestrict_dom_walker::check_call): Likewise.
11495         * gimple.c (gimple_build_call_from_tree): Likewise.
11496         * gimple.h (enum gf_mask): Likewise.
11497         (gimple_call_with_bounds_p): Likewise.
11498         (gimple_call_set_with_bounds): Likewise.
11499         * gimplify.c (gimplify_init_constructor): Likewise.
11500         * ipa-cp.c (initialize_node_lattices): Likewise.
11501         (propagate_constants_across_call): Likewise.
11502         (find_more_scalar_values_for_callers_subset): Likewise.
11503         * ipa-hsa.c (process_hsa_functions): Likewise.
11504         * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
11505         * ipa-icf.c (sem_function::merge): Likewise.
11506         * ipa-inline.c (early_inliner): Likewise.
11507         * ipa-pure-const.c (warn_function_noreturn): Likewise.
11508         (warn_function_cold): Likewise.
11509         (propagate_pure_const): Likewise.
11510         * ipa-ref.h (enum GTY): Likewise.
11511         * ipa-split.c (find_retbnd): Likewise.
11512         (consider_split): Likewise.
11513         (split_function): Likewise.
11514         * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
11515         * ipa.c (walk_polymorphic_call_targets): Likewise.
11516         (symbol_table::remove_unreachable_nodes): Likewise.
11517         (process_references): Likewise.
11518         (cgraph_build_static_cdtor_1): Likewise.
11519         * lto-cgraph.c (lto_output_node): Likewise.
11520         (output_refs): Likewise.
11521         (compute_ltrans_boundary): Likewise.
11522         (input_overwrite_node): Likewise.
11523         (input_node): Likewise.
11524         (input_cgraph_1): Likewise.
11525         * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
11526         * passes.c (pass_manager::execute_early_local_passes): Likewise.
11527         (class pass_chkp_instrumentation_passes): Likewise.
11528         (make_pass_chkp_instrumentation_passes): Likewise.
11529         * passes.def: Likewise.
11530         * rtl.h (struct GTY): Likewise.
11531         (CALL_EXPR_WITH_BOUNDS_P): Likewise.
11532         * stor-layout.c (layout_type): Likewise.
11533         * symtab.c: Likewise.
11534         * target.def: Likewise.
11535         * targhooks.c (default_chkp_bound_type): Likewise.
11536         (default_chkp_bound_mode): Likewise.
11537         (default_builtin_chkp_function): Likewise.
11538         (default_chkp_function_value_bounds): Likewise.
11539         (default_chkp_make_bounds_constant): Likewise.
11540         (default_chkp_initialize_bounds): Likewise.
11541         * targhooks.h (default_chkp_bound_type): Likewise.
11542         (default_chkp_bound_mode): Likewise.
11543         (default_builtin_chkp_function): Likewise.
11544         (default_chkp_function_value_bounds): Likewise.
11545         (default_chkp_make_bounds_constant): Likewise.
11546         (default_chkp_initialize_bounds): Likewise.
11547         * toplev.c (compile_file): Likewise.
11548         (process_options): Likewise.
11549         * tree-core.h (DEF_BUILTIN): Likewise.
11550         (DEF_BUILTIN_CHKP): Likewise.
11551         * tree-inline.c (declare_return_variable): Likewise.
11552         (remap_gimple_stmt): Likewise.
11553         (copy_bb): Likewise.
11554         (initialize_inlined_parameters): Likewise.
11555         (expand_call_inline): Likewise.
11556         * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
11557         (make_pass_ipa_chkp_early_produce_thunks): Likewise.
11558         (make_pass_ipa_chkp_produce_thunks): Likewise.
11559         (make_pass_chkp): Likewise.
11560         (make_pass_chkp_opt): Likewise.
11561         (make_pass_chkp_instrumentation_passes): Likewise.
11562         * tree-pretty-print.c (dump_generic_node): Likewise.
11563         * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
11564         * tree-ssa-dce.c (propagate_necessity): Likewise.
11565         (eliminate_unnecessary_stmts): Likewise.
11566         * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
11567         * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
11568         * tree-ssa-sccvn.h: Likewise.
11569         * tree-ssa-strlen.c (get_string_length): Likewise.
11570         (valid_builtin_call): Likewise.
11571         (adjust_last_stmt): Likewise.
11572         (handle_builtin_strchr): Likewise.
11573         (handle_builtin_strcpy): Likewise.
11574         (handle_builtin_stxncpy): Likewise.
11575         (handle_builtin_memcpy): Likewise.
11576         (handle_builtin_strcat): Likewise.
11577         (strlen_check_and_optimize_stmt): Likewise.
11578         * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
11579         * tree-streamer-in.c: Likewise.
11580         * tree-streamer.c (record_common_node): Likewise.
11581         * tree.c (tree_code_size): Likewise.
11582         (wide_int_to_tree_1): Likewise.
11583         (type_contains_placeholder_1): Likewise.
11584         (build_common_tree_nodes): Likewise.
11585         * tree.def (POINTER_BOUNDS_TYPE): Likewise.
11586         * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
11587         (POINTER_BOUNDS_P): Likewise.
11588         (BOUNDED_TYPE_P): Likewise.
11589         (BOUNDED_P): Likewise.
11590         (CALL_WITH_BOUNDS_P): Likewise.
11591         (pointer_bounds_type_node): Likewise.
11592         * value-prof.c (gimple_ic): Likewise.
11593         * var-tracking.c (vt_add_function_parameters): Likewise.
11594         * varasm.c (make_decl_rtl): Likewise.
11595         (assemble_start_function): Likewise.
11596         (output_constant): Likewise.
11597         (maybe_assemble_visibility): Likewise.
11598         * varpool.c (ctor_for_folding): Likewise.
11599         * chkp-builtins.def: Remove.
11600         * ipa-chkp.c: Remove.
11601         * ipa-chkp.h: Remove.
11602         * rtl-chkp.c: Remove.
11603         * rtl-chkp.h: Remove.
11604         * tree-chkp-opt.c: Remove.
11605         * tree-chkp.c: Remove.
11606         * tree-chkp.h: Remove.
11608 2018-06-07  Carl Love  <cel@us.ibm.com>
11610         * config/rs6000/vsx.md (vextract_fp_from_shorth,
11611         vextract_fp_from_shortl): Add BE support.
11613 2018-06-07  Paul Koning  <ni1d@arrl.net>
11615         * compare-elim.c (try_merge_compare): Don't merge compare if
11616         address contains a side effect.
11617         (try_eliminate_compare): Likewise.
11619 2018-06-07  Olga Makhotina  <olga.makhotina@intel.com>
11621         * config.gcc: Support "tremont".
11622         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
11623         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
11624         PROCESSOR_TREMONT.
11625         * config/i386/i386.c (m_TREMONT): Define.
11626         (processor_target_table): Add "tremont".
11627         (PTA_TREMONT): Define.
11628         (ix86_lea_outperforms): Add TARGET_TREMONT.
11629         (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
11630         (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
11631         and M_INTEL_GOLDMONT_PLUS.
11632         (fold_builtin_cpu): Add "tremont".
11633         (ix86_add_stmt_cost): Add TARGET_TREMONT.
11634         (ix86_option_override_internal): Add "tremont".
11635         * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
11636         (processor_type): Add PROCESSOR_TREMONT.
11637         * config/i386/x86-tune.def: Add m_TREMONT.
11638         * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
11640 2018-06-07  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
11642         * config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
11643         symbol defined for msp430i* devices to be lower case.
11645 2018-06-07  Richard Biener  <rguenther@suse.de>
11647         * graphite-sese-to-poly.c (extract_affine): Avoid unneded
11648         wrapping.  Properly wrap the result of a BIT_NOT_EXPR.
11649         Properly wrap signed arithmetic if overflow wraps.
11651 2018-06-07  Jakub Jelinek  <jakub@redhat.com>
11653         PR tree-optimization/69615
11654         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
11655         of a cast from a same precision integral SSA_NAME in a bb dominated
11656         by first_bb, retry with rhs2 set to the rhs1 of the cast.  Don't emit
11657         cast to utype if rhs2 has already a compatible type.
11659 2018-06-07  Richard Biener  <rguenther@suse.de>
11661         PR tree-optimization/85935
11662         * graphite-scop-detection.c (find_params_in_bb): Analyze
11663         condition operands with respect to the correct loop.  Assert
11664         the analysis doesn't fail.
11666 2018-06-04  Carl Love  <cel@us.ibm.com>
11668         * config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
11669         using natural element order.  Use gen_lshrsi3 instead of gen_ashrsi3
11670         as it is slightly cheaper.
11671         (first_match_or_eos_index_<mode>):
11672         Calculate index using natural element order.
11673         (first_match_index_<mode>):
11674         Calculate index using natural element order.
11675         (first_match_or_eos_index_<mode>):
11676         Calculate index using natural order.
11677         (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
11678         for BE and LE modes.
11679         * config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
11680         P9V_BUILTIN_VCLZLSBB_V16QI.
11681         * config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
11682         specific.
11684 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11686         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
11687         indentation and line wrap for many prototypes.  Add missing
11688         @smallexample directives around block of prototypes for vec_xl and
11689         vec_xst.
11691 2018-06-05  Michael Meissner  <meissner@linux.ibm.com>
11693         * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
11694         track if we pass or return IEEE 128-bit floating point.
11695         (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
11696         C++ mangling that is compatible with GCC 8.1.
11697         (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
11698         (init_cumulative_args): Note if we pass or return IEEE 128-bit
11699         floating point types.
11700         (rs6000_function_arg_advance_1): Likewise.
11701         (rs6000_mangle_type): Optionally generate mangled names that match
11702         what GCC 8.1 generated for IEEE 128-bit floating point types.
11703         (rs6000_globalize_decl_name): If we have an external function that
11704         passes or returns IEEE 128-bit types, generate a weak reference
11705         from the mangled name used in GCC 8.1 to the current mangled
11706         name.
11707         (rs6000_init_builtins): Make __ibm128 use the long double type if
11708         long double is IBM extended double.  Make __float128 use the long
11709         double type if long double is IEEE 128-bit.
11711         PR target/85657
11712         * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
11713         macro for __ibm128 built-in functions.
11714         (PACK_IF): Add __ibm128 pack/unpack functions.
11715         (UNPACK_IF): Likewise.
11716         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
11717         enable long double built-in functions if long double is IEEE
11718         128-bit floating point.
11719         (rs6000_invalid_builtin): Update long double built-in function
11720         error message.
11721         (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
11722         functions, adjust the built-in function to use the long double
11723         built-in function if __ibm128 and long double are the same type.
11724         * doc/extend.texi (PowerPC builtins): Update documention for
11725         __builtin_{,un}pack_longdouble.  Add documentation for
11726         __builtin_{,un}pack_ibm128.
11728 2018-06-06  Jim Wilson  <jimw@sifive.com>
11730         * config/riscv/riscv.c (enum riscv_privilege_levels): New.
11731         (struct machine_function): New field interrupt_mode.
11732         (riscv_handle_type_attribute): New function.  Add forward declaration.
11733         (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
11734         (riscv_expand_epilogue): Check interrupt_mode field.
11735         (riscv_set_current_function): Check interrupt attribute args and
11736         set interrupt_mode field.
11737         * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
11738         (riscv_sret, riscv_uret): New.
11739         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
11740         new arguments to interrupt attribute.
11742 2018-06-06  Peter Bergner  <bergner@vnet.ibm.com>
11744         PR target/63177
11745         * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
11746         Don't handle -mcpu=power8 if -mpower9-vector is also used.
11748 2018-06-06  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11750         * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
11751         VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
11752         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
11753         several redundant entries.
11755 2018-06-06  David Malcolm  <dmalcolm@redhat.com>
11757         * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
11758         type from "rtx" to "rtx_insn *".
11759         * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
11760         for local "call_insn", removing cast.
11761         (ix86_expand_call): Likewise, introducing a "call_insn" local.
11763 2018-06-06  Eric Botcazou  <ebotcazou@adacore.com>
11765         PR tree-optimization/86066
11766         * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
11767         for BIT_INSERT_EXPR stores.
11769 2018-06-06  Richard Biener  <rguenther@suse.de>
11771         PR tree-optimization/86062
11772         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
11773         component refs ontop
11774         of to be offsetted base.
11776 2018-06-06  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
11778         * config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
11779         to be static and remove check on interrupt attribute name.
11781 2018-06-05  Kelvin Nilsen  <kelvin@gcc.gnu.org>
11783         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
11784         volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
11786 2018-06-05  Steve Ellcey  <sellcey@cavium.com>
11788         PR target/79924
11789         * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
11790         second argument.
11791         * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
11792         Remove second argument, change how error is called.
11793         (aarch64_layout_arg): Remove second argument from
11794         aarch64_err_no_fpadvsimd call.
11795         (aarch64_init_cumulative_args): Ditto.
11796         (aarch64_gimplify_va_arg_expr): Ditto.
11797         * config/aarch64/aarch64.md (mov<mode>): Ditto.
11799 2018-06-05  Uros Bizjak  <ubizjak@gmail.com>
11801         * config/i386/i386.md (simple_return_indirect_internal): New expander.
11802         (*simple_return_indirect_internal<mode>): Rename from
11803         simple_return_indirect_internal.  Use W mode iterator.
11804         (rstorssp): New expander.
11805         (*rstorssp<mode>): Rename from rstorssp.  Use P mode iterator.
11806         (clrssbsy): New expander.
11807         (*clrssbsy<mode>): Rename from clrssbsy.  Use P mode iterator.
11809 2018-06-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
11811         * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
11812         __typeof__.
11813         (cmse_check_pointed_object): Likewise.
11815 2018-06-05  Martin Liska  <mliska@suse.cz>
11817         PR gcov-profile/47618
11818         * doc/invoke.texi: Document how -fprofile-dir format
11819         is extended.
11821 2018-06-05  Richard Biener  <rguenther@suse.de>
11823         * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
11824         removal pretend DOM info isn't available so we do not update
11825         it and only remove edges, not dominated blocks.  Actually free
11826         DOM info in case we removed something.  Remove unreachable blocks.
11827         (mfb_keep_latches): Work with either DOM info or marked backedges.
11828         (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
11829         first.  Mark backedges if DOM info isn't available.
11830         (Re-)compute DOM info after cleanup_control_flow_pre.
11832 2018-06-05  Richard Biener  <rguenther@suse.de>
11834         * tree-cfg.c (struct locus_discrim_map): Store line, not location.
11835         (locus_discrim_hasher::hash): Adjust.
11836         (locus_discrim_hasher::equal): Likewise.
11837         (next_discriminator_for_locus): Work on line directly.
11838         (same_line_p): Pass in expanded locus1 as well.
11839         (assign_discriminators): Avoid redundant location expansions.
11841 2018-06-05  Richard Biener  <rguenther@suse.de>
11843         PR tree-optimization/86046
11844         * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
11845         if required after clearing TREE_ADDRESSABLE.
11847 2018-06-05  Richard Biener  <rguenther@suse.de>
11849         PR tree-optimization/86047
11850         * tree-ssa-loop.c (for_each_index): Glob handling of all
11851         decls and constants and really handle all of them.
11853 2018-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
11855         PR target/81497
11856         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
11857         qualifier_void_pointer and qualifier_const_void_pointer.
11858         (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
11859         (arm_init_builtins): Handle the above.
11860         * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
11861         __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
11862         __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
11863         void intrinsics.
11865 2018-06-05  Martin Liska  <mliska@suse.cz>
11867         * auto-profile.c (read_autofdo_file): Do not use
11868         gcov_ctr_summary struct.
11869         (afdo_callsite_hot_enough_for_early_inline): Likewise.
11870         * coverage.c (struct counts_entry): Likewise.
11871         (read_counts_file): Read just single summary entry.
11872         (get_coverage_counts): Use gcov_summary struct.
11873         * coverage.h (get_coverage_counts): Likewise.
11874         * gcov-dump.c (dump_working_sets): Likewise.
11875         (tag_summary): Dump just single summary.
11876         * gcov-io.c (gcov_write_summary): Write just histogram
11877         summary.
11878         (gcov_read_summary): Read just single summary.
11879         (compute_working_sets): Use gcov_summary struct.
11880         * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
11881         of GCOV_COUNTERS_SUMMABLE.
11882         (GCOV_COUNTERS_SUMMABLE): Remove.
11883         (GCOV_FIRST_VALUE_COUNTER): Replace with
11884         GCOV_COUNTER_V_INTERVAL.
11885         (struct gcov_ctr_summary): Remove.
11886         (struct gcov_summary): Directly use fields of former
11887         gcov_ctr_summary.
11888         (compute_working_sets): Use gcov_summary struct.
11889         * gcov.c (read_count_file): Do not use ctrs fields.
11890         * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
11891         struct.
11892         * lto-streamer.h (struct GTY): Make profile_info gcov_summary
11893         struct.
11894         * profile.c: Likewise.
11895         * profile.h: Likewise.
11897 2018-06-05  Martin Liska  <mliska@suse.cz>
11899         PR gcov-profile/84846
11900         * gcov.c (output_lines): Print working directory only
11901         in intermediate format.
11903 2018-06-05  Andreas Krebbel  <krebbel@linux.ibm.com>
11905         * config/s390/s390-builtin-types.def: Add void function type.
11906         * config/s390/s390-builtins.def: Use the function type for the
11907         tbeginc builtin.
11909 2018-06-04  Jim Wilson  <jimw@sifive.com>
11911         * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
11912         to int.
11913         * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
11914         and maybe_eh_return.  Change regno to unsigned int.  Use new args to
11915         handle EH_RETURN_DATA_REGNO registers properly.
11916         (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
11917         (riscv_expand_epilogue): Update comment.  Change argument name and
11918         type.  Update code to use new name and type.  Pass new args to
11919         riscv_for_each_saved_reg.  Only use EH_RETURN_STACKADJ_RTX when
11920         EXCEPTION_RETURN.
11921         * config/riscv/riscv.md (NORMAL_RETURN): New.
11922         (SIBCALL_RETURN, EXCEPTION_RETURN): New.
11923         (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
11924         (eh_return): Call gen_eh_return_internal and emit barrier.
11925         (eh_return_internal): Call riscv_expand_epilogue.
11927 2018-06-04  Eric Botcazou  <ebotcazou@adacore.com>
11929         * gimple-ssa-store-merging.c (struct merged_store_group): Move up
11930         bit_insertion field and declare can_be_merged_into method.
11931         (merged_store_group::can_be_merged_into): New method.
11932         (imm_store_chain_info::coalesce_immediate): Call it to decide whether
11933         consecutive non-overlapping stores can be merged.  Turn MEM_REF stores
11934         into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
11936 2018-06-04  Richard Biener  <rguenther@suse.de>
11938         PR tree-optimization/85955
11939         * builtins.c (fold_builtin_sincos): Convert pointers to
11940         destination to appropriate type before dereferencing.
11942 2018-06-04  Segher Boessenkool  <segher@kernel.crashing.org>
11944         * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
11946 2018-06-04  Richard Sandiford  <richard.sandiford@linaro.org>
11948         * expr.c (expand_expr_real_1): Force the operand into memory if
11949         its TYPE_MODE is BLKmode and if there is no integer mode for
11950         the number of bits being extracted.
11952 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
11954         PR target/85832
11955         PR target/86036
11956         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
11957         Use vptestnm rather than vptestm in (=Yc,v,C) variant.
11959 2018-06-04  Richard Biener  <rguenther@suse.de>
11961         * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
11962         (cleanup_tree_cfg_noloop): ... single caller.  Do
11963         start_recording_case_labels later.
11965 2018-06-04  Sebastian Peryt  <sebastian.peryt@intel.com>
11967         * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
11968         to _IMMINTRIN_H_INCLUDED.
11969         * config/i386/pconfigintrin.h: Ditto.
11970         * config/i386/waitpkgintrin.h: Ditto.
11971         * config/i386/immintrin.h: Add includes for sgxintrin.h,
11972         pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
11973         * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
11974         emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
11975         bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
11976         waitpkgintrin.h and cldemoteintrin.h.
11978 2018-06-04  Richard Biener  <rguenther@suse.de>
11980         PR tree-optimization/86038
11981         * tracer.c (find_best_successor): Check probability for
11982         being initialized, bail out if not.
11984 2018-06-04  Richard Earnshaw  <rearnsha@arm.com>
11986         PR target/86003
11987         * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
11988         of bits to ignore when comparing architectures.
11990 2018-06-04  Jakub Jelinek  <jakub@redhat.com>
11992         PR tree-optimization/69615
11993         * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
11994         maximum or minimum of the type, try to merge it also as if
11995         range1 is + [-, x - 1] or + [x + 1, -].
11997         PR c++/86025
11998         * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
12000 2018-06-03  Eric Botcazou  <ebotcazou@adacore.com>
12002         PR tree-optimization/86034
12003         * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
12004         the unsigned bitfield type in a bit insertion sequence if it does not
12005         have a larger precision than the bitfield size.
12006         (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
12008 2018-06-03  Kito Cheng  <kito.cheng@gmail.com>
12010         * config/nds32/nds32-peephole2.md: Add new patterns for code size.
12012 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
12014         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
12015         * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
12016         * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
12017         * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
12019 2018-06-03  Chung-Ju Wu  <jasonwucj@gmail.com>
12021         * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
12022         Disable -fdelete-null-pointer-checks for ELF toolchain.
12024 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
12025             Kito Cheng  <kito.cheng@gmail.com>
12027         * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
12028         (nds32le-*-*, nds32be-*-*): Integrate checking process.
12029         (nds32*-*-*): Add glibc and uclibc conditions.
12030         * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
12031         (TARGET_EXCEPT_UNWIND_INFO): Define.
12032         * config/nds32/elf.h: New file.
12033         * config/nds32/linux.h: New file.
12034         * config/nds32/nds32-elf.opt: New file.
12035         * config/nds32/nds32-linux.opt: New file.
12036         * config/nds32/nds32-fp-as-gp.c
12037         (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
12038         * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
12039         TARGET_LINUX_ABI.
12040         (nds32_asm_file_end): Ditto.
12041         (nds32_print_operand): Ditto.
12042         (nds32_insert_attributes): Ditto.
12043         (nds32_init_libfuncs): New function.
12044         (TARGET_HAVE_TLS): Define.
12045         (TARGET_INIT_LIBFUNCS): Define.
12046         * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
12047         spec content.
12048         (TARGET_ELF): Apply different mcmodel setting.
12049         (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
12050         been migrated into elf.h and linux.h files.
12051         * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
12052         * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
12053         (mcmodel): The content has been migrated into nds32-elf.opt and
12054         nds32-linux.opt files.
12055         * config/nds32/t-elf: New file.
12056         * config/nds32/t-linux: New file.
12058 2018-06-02  Chung-Ju Wu  <jasonwucj@gmail.com>
12059             Shiva Chen  <shiva0217@gmail.com>
12061         * config/nds32/constants.md (unspec_volatile_element): Add
12062         UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
12063         * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
12064         optimization.
12065         * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
12066         (make_pass_nds32_fp_as_gp): Declare.
12067         * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
12068         optmization pass.
12069         (nds32_asm_function_end_prologue): Remove unused asm output.
12070         (nds32_asm_function_begin_epilogue): Remove unused asm output.
12071         (nds32_asm_file_start): Output necessary fp_as_gp information.
12072         (nds32_option_override): Adjust register usage.
12073         (nds32_expand_prologue): Consider fp_as_gp situation.
12074         (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
12075         * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
12076         (epilogue): Ditto.
12077         (return): Ditto.
12078         (simple_return): Ditto.
12079         (omit_fp_begin): Output special directive for fp_as_gp.
12080         (omit_fp_end): Output special directive for fp_as_gp.
12081         * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
12082         mforbid-fp-as-gp): New options.
12084 2018-06-01  Mark Wielaard  <mark@klomp.org>
12086         * dwarf2out.c (dwarf2out_finish): Remove generation of
12087         DW_AT_loclists_base.
12089 2018-06-01  Eric Botcazou  <ebotcazou@adacore.com>
12091         * gimple-ssa-store-merging.c: Include gimple-fold.h.
12092         (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
12093         (struct merged_store_group): Add bit_insertion field.
12094         (dump_char_array): Use standard hexadecimal format.
12095         (merged_store_group::merged_store_group): Set bit_insertion to false.
12096         (merged_store_group::apply_stores): Use optimal buffer size.  Deal
12097         with BIT_INSERT_EXPR stores.  Move up code updating the mask and
12098         also print the mask in the dump file.
12099         (pass_store_merging::gate): Minor tweak.
12100         (imm_store_chain_info::coalesce_immediate): Fix wrong association
12101         of stores with groups in dump.  Allow coalescing of BIT_INSERT_EXPR
12102         stores with INTEGER_CST stores.
12103         (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
12104         (imm_store_chain_info::output_merged_store): Add try_bitpos variable
12105         and use it throughout.  Generate bit insertion sequences if need be.
12106         (pass_store_merging::process_store): Remove redundant condition.
12107         Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
12109 2018-06-01  Segher Boessenkool  <segher@kernel.crashing.org>
12111         * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
12112         the 128-bit floating point types.  Fix function comment.
12114 2018-06-01  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12116         * config/aarch64/aarch64-simd.md
12117         (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
12118         mnemonics.
12119         (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
12120         mnemonics.
12122 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
12124         PR tree-optimization/85989
12125         * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
12126         variable.
12127         (backprop::intersect_uses): Check it when deciding whether this
12128         is a backedge reference.
12129         (backprop::process_block): Add each phi to m_visited_phis
12130         after visiting it, then clear it at the end.
12132 2018-06-01  Richard Biener  <rguenther@suse.de>
12134         * tree-vectorizer.h (vect_dr_stmt): New function.
12135         (vect_get_load_cost): Adjust.
12136         (vect_get_store_cost): Likewise.
12137         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
12138         Use vect_dr_stmt instead of DR_SMTT.
12139         (vect_record_base_alignments): Likewise.
12140         (vect_calculate_target_alignment): Likewise.
12141         (vect_compute_data_ref_alignment): Likewise and make static.
12142         (vect_update_misalignment_for_peel): Likewise.
12143         (vect_verify_datarefs_alignment): Likewise.
12144         (vector_alignment_reachable_p): Likewise.
12145         (vect_get_data_access_cost): Likewise.  Pass down
12146         vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
12147         (vect_get_peeling_costs_all_drs): Likewise.
12148         (vect_peeling_hash_get_lowest_cost): Likewise.
12149         (vect_enhance_data_refs_alignment): Likewise.
12150         (vect_find_same_alignment_drs): Likewise.
12151         (vect_analyze_data_refs_alignment): Likewise.
12152         (vect_analyze_group_access_1): Likewise.
12153         (vect_analyze_group_access): Likewise.
12154         (vect_analyze_data_ref_access): Likewise.
12155         (vect_analyze_data_ref_accesses): Likewise.
12156         (vect_vfa_segment_size): Likewise.
12157         (vect_small_gap_p): Likewise.
12158         (vectorizable_with_step_bound_p): Likewise.
12159         (vect_prune_runtime_alias_test_list): Likewise.
12160         (vect_analyze_data_refs): Likewise.
12161         (vect_supportable_dr_alignment): Likewise.
12162         * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
12163         (vect_gen_prolog_loop_niters): Likewise.
12164         * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
12165         * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
12166         modify DR_STMT.
12167         (vect_recog_mask_conversion_pattern): Likewise.
12168         (vect_try_gather_scatter_pattern): Likewise.
12169         * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
12170         to vect_get_store_cost.
12171         (vect_get_store_cost): Get stmt_info instead of DR.
12172         (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
12173         (vect_get_load_cost): Get stmt_info instead of DR.
12175 2018-06-01  Richard Biener  <rguenther@suse.de>
12177         PR middle-end/86017
12178         * gimple-fold.c (var_decl_component_p): Also allow offsetted
12179         vars wrapped in MEM_REFs.
12181 2018-06-01  Richard Sandiford  <richard.sandiford@linaro.org>
12183         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
12184         Fix subreg tests so that we only return a choice between
12185         GENERAL_REGS and FP_REGS if the original classes included both.
12187 2018-06-01  Richard Biener  <rguenther@suse.de>
12189         PR ipa/85960
12190         * tree-ssa-structalias.c (get_function_part_constraint):
12191         Handle NULL fi->decl.
12192         (find_func_aliases_for_call): Properly handle indirect
12193         fi from direct call.
12194         (find_func_clobbers): Likewise.
12195         (ipa_pta_execute): Likewise.
12196         (create_variable_info_for): For functions that are ifunc_resolver
12197         resolve to a varinfo that contains the result of the resolver call.
12198         (associate_varinfo_to_alias): Do not treat ifunc resolvers as
12199         aliases.
12201 2018-05-31  Michael Collison  <michael.collison@arm.com>
12203         * config/aarch64/aarch64.md:
12204         (*fix_to_zero_extenddfdi2): New pattern.
12205         * gcc.target/aarch64/fix_extend1.c: New testcase.
12207 2018-05-31  Qing Zhao <qing.zhao@oracle.com>
12209         PR middle-end/78809
12210         PR middle-end/83026
12211         * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
12212         and BUILT_IN_STRNCMP_EQ.
12213         * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
12214         BUILT_IN_STRNCMP_EQ.
12215         * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
12216         handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
12217         (gimple_fold_builtin): Likewise.
12218         * tree-ssa-strlen.c (compute_string_length): New function.
12219         (determine_min_obsize): New function.
12220         (handle_builtin_string_cmp): New function to handle calls to
12221         string compare functions.
12222         (strlen_optimize_stmt): Add handling to builtin string compare
12223         calls.
12224         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
12225         Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
12226         * tree.c (build_common_builtin_nodes): Add new defines of
12227         BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
12229 2018-05-31  Jakub Jelinek  <jakub@redhat.com>
12231         PR target/85984
12232         * bb-reorder.c (pass_partition_blocks::gate): Return false for
12233         functions with naked attribute.
12235 2018-05-31  Uros Bizjak  <ubizjak@gmail.com>
12237         * config/i386/sse.md (avx_vec_concat<mode>):
12238         Substitute concat_tg_mode mode attribute with xtg_mode.
12239         (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
12240         (concat_tg_mode): Remove mode attribute.
12242 2018-05-31  Martin Sebor  <msebor@redhat.com>
12244         PR c/82063
12245         * calls.c (alloc_max_size): Correct a logic error/typo.
12246         Treat excessive arguments as infinite.  Warn for invalid arguments.
12247         * doc/invoke.texi (-Walloc-size-larger-than): Update.
12249 2018-05-31  H.J. Lu  <hongjiu.lu@intel.com>
12251         PR target/85829
12252         * config/i386/x86-tune.def: Re-enable partial_reg_dependency
12253         and movx for Haswell.
12255 2018-05-31  Chung-Lin Tang  <cltang@codesourcery.com>
12256             Cesar Philippidis  <cesar@codesourcery.com>
12258         PR middle-end/85879
12259         * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
12260         when emitting error on private/firstprivate reductions.
12261         * omp-low.c (lower_omp_target): Avoid reference-type processing
12262         on pointers for firstprivate clause.
12264 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
12266         * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
12267         (st1x2): Likewise.
12268         (st1x3): Likewise.
12269         * config/aarch64/aarch64-simd.md
12270         (aarch64_ld1x3<VALLDIF:mode>): New pattern.
12271         (aarch64_ld1_x3_<mode>): Likewise
12272         (aarch64_st1x2<VALLDIF:mode>): Likewise
12273         (aarch64_st1_x2_<mode>): Likewise
12274         (aarch64_st1x3<VALLDIF:mode>): Likewise
12275         (aarch64_st1_x3_<mode>): Likewise
12276         * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
12277         (vld1_s8_x3): Likewise.
12278         (vld1_u16_x3): Likewise.
12279         (vld1_s16_x3): Likewise.
12280         (vld1_u32_x3): Likewise.
12281         (vld1_s32_x3): Likewise.
12282         (vld1_u64_x3): Likewise.
12283         (vld1_s64_x3): Likewise.
12284         (vld1_f16_x3): Likewise.
12285         (vld1_f32_x3): Likewise.
12286         (vld1_f64_x3): Likewise.
12287         (vld1_p8_x3): Likewise.
12288         (vld1_p16_x3): Likewise.
12289         (vld1_p64_x3): Likewise.
12290         (vld1q_u8_x3): Likewise.
12291         (vld1q_s8_x3): Likewise.
12292         (vld1q_u16_x3): Likewise.
12293         (vld1q_s16_x3): Likewise.
12294         (vld1q_u32_x3): Likewise.
12295         (vld1q_s32_x3): Likewise.
12296         (vld1q_u64_x3): Likewise.
12297         (vld1q_s64_x3): Likewise.
12298         (vld1q_f16_x3): Likewise.
12299         (vld1q_f32_x3): Likewise.
12300         (vld1q_f64_x3): Likewise.
12301         (vld1q_p8_x3): Likewise.
12302         (vld1q_p16_x3): Likewise.
12303         (vld1q_p64_x3): Likewise.
12304         (vst1_s64_x2): Likewise.
12305         (vst1_u64_x2): Likewise.
12306         (vst1_f64_x2): Likewise.
12307         (vst1_s8_x2): Likewise.
12308         (vst1_p8_x2): Likewise.
12309         (vst1_s16_x2): Likewise.
12310         (vst1_p16_x2): Likewise.
12311         (vst1_s32_x2): Likewise.
12312         (vst1_u8_x2): Likewise.
12313         (vst1_u16_x2): Likewise.
12314         (vst1_u32_x2): Likewise.
12315         (vst1_f16_x2): Likewise.
12316         (vst1_f32_x2): Likewise.
12317         (vst1_p64_x2): Likewise.
12318         (vst1q_s8_x2): Likewise.
12319         (vst1q_p8_x2): Likewise.
12320         (vst1q_s16_x2): Likewise.
12321         (vst1q_p16_x2): Likewise.
12322         (vst1q_s32_x2): Likewise.
12323         (vst1q_s64_x2): Likewise.
12324         (vst1q_u8_x2): Likewise.
12325         (vst1q_u16_x2): Likewise.
12326         (vst1q_u32_x2): Likewise.
12327         (vst1q_u64_x2): Likewise.
12328         (vst1q_f16_x2): Likewise.
12329         (vst1q_f32_x2): Likewise.
12330         (vst1q_f64_x2): Likewise.
12331         (vst1q_p64_x2): Likewise.
12332         (vst1_s64_x3): Likewise.
12333         (vst1_u64_x3): Likewise.
12334         (vst1_f64_x3): Likewise.
12335         (vst1_s8_x3): Likewise.
12336         (vst1_p8_x3): Likewise.
12337         (vst1_s16_x3): Likewise.
12338         (vst1_p16_x3): Likewise.
12339         (vst1_s32_x3): Likewise.
12340         (vst1_u8_x3): Likewise.
12341         (vst1_u16_x3): Likewise.
12342         (vst1_u32_x3): Likewise.
12343         (vst1_f16_x3): Likewise.
12344         (vst1_f32_x3): Likewise.
12345         (vst1_p64_x3): Likewise.
12346         (vst1q_s8_x3): Likewise.
12347         (vst1q_p8_x3): Likewise.
12348         (vst1q_s16_x3): Likewise.
12349         (vst1q_p16_x3): Likewise.
12350         (vst1q_s32_x3): Likewise.
12351         (vst1q_s64_x3): Likewise.
12352         (vst1q_u8_x3): Likewise.
12353         (vst1q_u16_x3): Likewise.
12354         (vst1q_u32_x3): Likewise.
12355         (vst1q_u64_x3): Likewise.
12356         (vst1q_f16_x3): Likewise.
12357         (vst1q_f32_x3): Likewise.
12358         (vst1q_f64_x3): Likewise.
12359         (vst1q_p64_x3): Likewise.
12361 2018-05-30  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
12363         * config/msp430/msp430.c (msp430_output_labelref): Prepend
12364         user_label_prefix to name.
12366         * tree-core.h: Update comment about the format of NAME string
12367         passed to handler in attribute_spec.
12369         * config/msp430/msp430.md: Remove erroneous subreg expression from
12370         zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
12371         zero_extend{q,h}isi2.
12373 2018-05-30  Borislav Petkov  <bp@suse.de>
12375         * doc/extend.texi: Document some architecture specific
12376         constraints and sort entries.
12378 2018-05-30  Martin Sebor  <msebor@redhat.com>
12380         PR middle-end/85369
12381         * builtins.c (expand_builtin_stpcpy_1): New function.
12382         (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
12383         only if the former succeeds.
12385 2018-05-31  Sameera Deshpande <sameera.deshpande@linaro.org>
12387         * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
12388         in saphira.
12390 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
12392         * doc/invoke.texi (-flinker-output): Document
12394 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
12396         * passes.c (ipa_write_summaries): Only modify statements if body
12397         is in memory.
12398         * cgraphunit.c (ipa_passes): Also produce intermeidate code when
12399         incrementally linking.
12400         (ipa_passes): Likewise.
12401         * lto-cgraph.c (lto_output_node): When incrementally linking do not
12402         pass down resolution info.
12403         * common.opt (flag_incremental_link): Update info.
12404         * gcc.c (plugin specs): Turn flinker-output=* to
12405         -plugin-opt=-linker-output-known
12406         * toplev.c (compile_file): Also cut compilation when doing incremental
12407         link.
12408         * flag-types. (enum lto_partition_model): Add
12409         LTO_LINKER_OUTPUT_NOLTOREL.
12410         (invoke.texi): Add -flinker-output docs.
12411         * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
12412         link same way as WPA; do not stream in dead initializers.
12414         * dwarf2out.c (dwarf2out_die_ref_for_decl,
12415         darf2out_register_external_decl): Support incremental link.
12417 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
12419         * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
12421 2018-05-30  Jan Hubicka  <hubicka@ucw.cz>
12423         * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
12424         it down to simple_object_copy_lto_debug_sections.
12425         (run_gcc): Determine incremental LTO link time and configure
12426         lto1 into non-wpa mode, disable renaming of debug sections.
12428 2018-05-30  Kelvin Nilsen  <kelvin@gcc.gnu.org>
12430         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
12431         descriptions of various incorrectly documented functions.
12433 2018-05-30  Andre Vieira  <andre.simoesdiasvieira@arm.com>
12435         Revert:
12436         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
12437         address check not strict.
12439 2018-05-30  Richard Biener  <rguenther@suse.de>
12441         PR tree-optimization/85964
12442         * tracer.c (better_p): Drop initialized count check, we only
12443         call the function with initialized counts now.
12444         (find_best_successor): Do find a best edge if one
12445         has uninitialized count.
12446         (find_best_predecessor): Likewise.  Do BB frequency check only
12447         if count is initialized.
12449 2017-05-30  Jackson Woodruff  <jackson.woodruff@arm.com>
12451         * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
12452         (aarch64_ldrstr_offset_compare): New.
12453         (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
12454         load/store orderings.
12455         (aarch64_gen_adjusted_ldpstp): Likewise.
12457 2018-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
12459         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
12460         Check for subset of GENERAL_REGS and FP_REGS.
12461         * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
12462         r=w alternative.
12464 2018-05-30  Richard Sandiford  <richard.sandiford@linaro.org>
12466         * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
12467         and wi::to_poly_offset.  Add the current offset and then check
12468         whether the sum fits, rather than using an unchecked addition of
12469         a checked term.  Check for a shwi rather than a uhwi.
12470         * expr.c (get_bit_range): Use tree_to_poly_uint64.
12471         (store_constructor): Use poly_int_tree_p.
12472         (expand_expr_real_1): Likewise.
12473         * function.c (assign_temp): Likewise.
12474         * fold-const.c (const_binop): Use poly_int_tree_p and
12475         wi::to_poly_offset.
12476         (fold_indirect_ref_1): Likewise.  Use multiple_p to attempt an exact
12477         division.
12478         * ipa-icf-gimple.c (func_checker::compare_operand): Use
12479         to_poly_offset for MEM offsets.
12480         * ipa-icf.c (sem_variable::equals): Likewise.
12481         * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
12482         * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
12483         wi::to_poly_offset for BIT_FIELD_REF offsets.
12484         (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
12485         wi::to_poly_offset.
12486         * var-tracking.c (emit_note_insn_var_location): Use
12487         tree_to_poly_uint64.
12489 2018-05-29  Jim Wilson  <jimw@sifive.com>
12491         * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
12493 2018-05-29  Uros Bizjak  <ubizjak@gmail.com>
12495         PR target/85950
12496         * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
12497         Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
12498         sequence.
12499         (sse4_1_round<mode>2): Use nonimmediate_operand
12500         for operand 1 predicate.
12502 2018-05-29  Martin Sebor  <msebor@redhat.com>
12503             Richard Biener  <rguenther@suse.de>
12505         PR testsuite/85888
12506         * calls.c (get_size_range): Call determine_value_range instead
12507         of get_value_range..
12508         * tree-vrp.h (determine_value_range): Declared new function.
12509         * tree-vrp.c (determine_value_range_1, determine_value_range): New.
12511 2018-05-29  Richard Biener  <rguenther@suse.de>
12513         * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
12514         sure to use non-pattern stmts for get_earlier_stmt arguments.
12515         * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
12516         called on pattern stmts.
12517         (get_later_stmt): Likewise.
12519 2018-05-29  Martin Liska  <mliska@suse.cz>
12521         PR gcov-profile/85759
12522         * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
12523         env variables.
12525 2018-05-29  Jakub Jelinek  <jakub@redhat.com>
12527         * tree-cfg.c (verify_gimple_assign_unary): Add checking for
12528         VEC_UNPACK_*_EXPR.
12529         (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
12530         VEC_PACK_*_EXPR.
12532         PR target/85918
12533         * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
12534         VEC_PACK_FLOAT_EXPR): New tree codes.
12535         * tree-pretty-print.c (op_code_prio): Handle
12536         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
12537         (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
12538         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
12539         * tree-inline.c (estimate_operator_cost): Likewise.
12540         * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
12541         * fold-const.c (const_binop): Likewise.
12542         (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
12543         VEC_UNPACK_FIX_TRUNC_LO_EXPR.
12544         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
12545         (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
12546         * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
12547         VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
12548         * expr.c (expand_expr_real_2): Likewise.
12549         * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
12550         vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
12551         vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
12552         optabs.
12553         * optabs.c (expand_widen_pattern_expr): For
12554         VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
12555         sign from result type rather than operand's type.
12556         (expand_binop_directly): For vec_packu_float_optab and
12557         vec_packs_float_optab allow result type to be different from operand's
12558         type.
12559         * optabs-tree.c (optab_for_tree_code): Handle
12560         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
12561         VEC_PACK_FLOAT_EXPR.  Formatting fixes.
12562         * tree-vect-generic.c (expand_vector_operations_1):  Handle
12563         VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
12564         VEC_PACK_FLOAT_EXPR.
12565         * tree-vect-stmts.c (supportable_widening_operation): Handle
12566         FIX_TRUNC_EXPR.
12567         (supportable_narrowing_operation): Handle FLOAT_EXPR.
12568         * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
12569         * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
12570         (float<floatunssuffix>v2div2sf2): ... this.  Formatting fix.
12571         (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
12572         mode attributes.
12573         (vec_pack<floatprefix>_float_<mode>): New expander.
12574         (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
12575         attributes.
12576         (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
12577         vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
12578         * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
12579         vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
12580         vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
12581         Document.
12582         * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
12583         VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
12584         (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
12585         VEC_PACK_FLOAT_EXPR): Document.
12587 2018-05-29  Richard Biener  <rguenther@suse.de>
12589         * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
12590         member.
12591         (stmt_vec_info_vec): Make pointer.
12592         (init_stmt_vec_info_vec): Remove.
12593         (free_stmt_vec_info_vec): Likewise.
12594         (set_stmt_vec_info_vec): New function.
12595         (free_stmt_vec_infos): Likewise.
12596         (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
12597         (set_vinfo_for_stmt): Likewise.
12598         (get_earlier_stmt): Likewise.
12599         (get_later_stmt): Likewise.
12600         * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
12601         (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
12602         (vec_info::~vec_info): Free stmt_vec_infos.
12603         (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
12604         Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
12605         (pass_slp_vectorize::execute): Likewise.
12606         * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
12607         (free_stmt_vec_info_vec): Likewise.
12608         (set_stmt_vec_info_vec): New function.
12609         (free_stmt_vec_infos): Likewise.
12610         * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
12611         the global stmt_vec_info_vec.
12612         * tree-parloops.c (gather_scalar_reductions): Use
12613         set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
12614         vector.
12616 2018-05-29  Richard Biener  <rguenther@suse.de>
12618         * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
12620 2018-05-29  Martin Liska  <mliska@suse.cz>
12621             David Malcolm  <dmalcolm@redhat.com>
12623         * vec.c (test_reverse): New.
12624         (vec_c_tests): Add new test.
12625         * vec.h (vl_ptr>::reverse): New function.
12627 2018-05-29  Gerald Pfeifer  <gerald@pfeifer.com>
12629         * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
12631         * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
12632         and later.
12634 2018-05-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
12636         * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
12638 2018-05-28  Richard Biener  <rguenther@suse.de>
12640         PR tree-optimization/85933
12641         * tree-vect-data-refs.c (vect_record_base_alignments): Only
12642         look at stmts marked as vectorizable.
12644 2018-05-28  Richard Biener  <rguenther@suse.de>
12646         PR tree-optimization/85934
12647         * tree-vect-generic.c (expand_vector_operations_1): Hoist
12648         vector boolean check before scalar optimization.
12650 2018-05-28  Jakub Jelinek  <jakub@redhat.com>
12652         * doc/invoke.texi (ARM Options): Use @item instead of @itemx
12653         for armv5te.
12655 2018-05-28  Mark Wielaard  <mark@klomp.org>
12657         * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
12658         if it is an expression containing a minus sign.
12660 2018-05-27  John David Anglin  <danglin@gcc.gnu.org>
12662         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
12664 2018-05-27  Paul Koning  <ni1d@arrl.net>
12666         * config/pdp11/pdp11.md (truncsihi2): Remove.
12668 2018-05-27  Monk Chiang  <sh.chiang04@gmail.com>
12669             Chung-Ju Wu  <jasonwucj@gmail.com>
12671         * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
12672         implementation.
12673         (unaligned_store_dw): Ditto.
12674         * config/nds32/nds32-memory-manipulation.c
12675         (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
12676         (nds32_gen_dup_4_byte_to_word_value): Rename to ...
12677         (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
12678         (emit_setmem_word_loop): Rename to ...
12679         (emit_setmem_doubleword_loop): ... this.
12680         (nds32_gen_dup_4_byte_to_word_value): New function.
12681         (nds32_gen_dup_8_byte_to_double_word_value): New function.
12682         (nds32_expand_setmem_loop): Refine implementation.
12683         (nds32_expand_setmem_loop_v3m): Ditto.
12684         * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
12685         pattern.
12687 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
12689         * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
12691 2018-05-27  Chung-Ju Wu  <jasonwucj@gmail.com>
12693         * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
12694         (nds32_init_machine_status): Initialize machine->attr_naked_p and
12695         machine->attr_no_prologue_p.
12696         (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
12697         (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
12698         (nds32_expand_epilogue): Consider attr_naked_p.
12699         (nds32_expand_epilogue_v3pop): Likewise.
12700         (nds32_can_use_return_insn): Likewise.
12701         * config/nds32/nds32.h (machine_function): Add attr_naked_p and
12702         attr_no_prologue_p fields.
12703         * config/nds32/nds32.opt (mret-in-naked-func): New option.
12705 2018-05-27  Jakub Jelinek  <jakub@redhat.com>
12707         PR target/85918
12708         * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
12709         attributes.
12710         * config/i386/sse.md
12711         (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
12712         Rename to ...
12713         (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
12714         ... this.
12715         (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
12716         Rename to ...
12717         (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
12718         ... this.
12719         (*<floatsuffix>floatv2div2sf2): Rename to ...
12720         (*float<floatunssuffix>v2div2sf2): ... this.
12721         (<floatsuffix>floatv2div2sf2_mask): Rename to ...
12722         (float<floatunssuffix>v2div2sf2_mask): ... this.
12723         (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
12724         (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
12725         (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
12726         to ...
12727         (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
12728         ... this.
12729         (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
12730         Rename to ...
12731         (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
12732         ... this.
12733         (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
12734         Rename to ...
12735         (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
12736         ... this.
12737         (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
12738         (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
12739         (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
12740         gen_ufix_truncv8dfv8si2.
12741         * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
12742         __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
12743         __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
12744         __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
12745         __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
12746         __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
12747         __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
12748         Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
12750 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
12752         PR target/85900
12753         PR target/85345
12754         * varasm.c (assemble_alias): Lookup ifunc attribute on error.
12756 2018-05-25  Jim Wilson  <jimw@sifive.com>
12758         * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
12759         * config/riscv/riscv.c (struct machine_function): Add
12760         interrupt_handler_p and attribute_checked_p fields.
12761         (riscv_attribute_table): Add interrupt.
12762         (riscv_interrupt_type_p): New.
12763         (riscv_save_reg_p): Save extra regs for interrupt handler.
12764         (riscv_use_save_libcall): Return false  for interrupt handler.
12765         (riscv_first_stack_step): Add forward declaration.
12766         (riscv_compute_frame_info): New local interrupt_save_t1.  Set it
12767         for interrupt handler with large frame.  Use it for saved reg list.
12768         (riscv_expand_prologue): Move flag_stack_usage_info support to
12769         eliminate duplication.
12770         (riscv_expand_epilogue): Generate mret for interrupt handler.
12771         (riscv_epilogue_uses): New.
12772         (riscv_can_use_return_insn): Return false for interrupt handler.
12773         (riscv_function_ok_for_sibcall): Likewise.
12774         (riscv_set_current_function): Add interrupt handler support.
12775         * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
12776         * config/riscv/riscv.md (UNSPECV_MRET): New.
12777         (GP_REGNUM): New.
12778         (riscv_frflags, riscv_fsflags): Use tab after opcode.
12779         (riscv_mret): New.
12780         * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
12782 2018-05-25  Bill Schmidt  <wschmidt@linux.ibm.com>
12784         PR tree-optimization/85712
12785         * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
12786         this candidate has already been replaced in-situ by a copy.
12788 2018-05-25  Jason Merrill  <jason@redhat.com>
12790         PR c++/80485 - inline function non-zero address.
12791         * symtab.c (nonzero_address): Check DECL_COMDAT.
12793 2018-05-25  Uros Bizjak  <ubizjak@gmail.com>
12795         PR target/83628
12796         * config/alpha/alpha.md (ashlsi3): New insn pattern.
12797         (*ashlsi_se): Rename from *ashldi_se.  Define as sign
12798         extension of SImode operation.  Use const123_operand predicate.
12799         (*saddsi_1): Remove.
12800         (*saddl_se_1): Ditto.
12801         (*ssubsi_1): Ditto.
12802         (*ssubl_se_1): Ditto.
12803         * config/alpha/predicates.md (const123_operand): New predicate.
12804         * config/alpha/constraints.md (P): Use IN_RANGE.
12806 2018-05-25  Richard Biener  <rguenther@suse.de>
12808         * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
12809         defaulted to true.
12810         (ref_maybe_used_by_stmt_p): Likewise.
12811         (stmt_may_clobber_ref_p): Likewise.
12812         (stmt_may_clobber_ref_p_1): Likewise.
12813         * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
12814         and pass it along.
12815         (ref_maybe_used_by_stmt_p): Likewise.
12816         (stmt_may_clobber_ref_p): Likewise.
12817         (stmt_may_clobber_ref_p_1): Likewise.
12818         * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
12819         the alias oracle to disambiguate DRs with stmts DR analysis
12820         couldn't handle.
12821         (vect_analyze_data_refs): Do not give up on not analyzable
12822         DRs for BB vectorization.  Remove code truncating the dataref
12823         vector.
12825 2018-05-25  Jakub Jelinek  <jakub@redhat.com>
12827         PR target/85832
12828         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
12829         Add (=Yk,v,C) variant using vptestm insn.  Use TARGET_AVX512BW
12830         in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
12832 2018-05-25  Richard Biener  <rguenther@suse.de>
12834         * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
12835         function, combining stmt data ref gathering and fatal analysis
12836         parts.
12837         (vect_analyze_data_refs): Remove now redudnant code and simplify.
12838         * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
12839         vect_analyze_loop_2 and use vect_find_stmt_data_reference.
12840         * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
12841         * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
12843 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
12845         PR tree-optimization/85720
12846         * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
12847         SCC if all partitions are builtins.
12848         (version_loop_by_alias_check): New parameter.  Generate cancelable
12849         runtime alias check if all partitions are builtins.
12850         (distribute_loop): Update call to above function.
12852 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
12854         * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
12855         (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
12856         (parm_default_def_partition_arg): Ditto.
12857         (set_parm_default_def_partition): Ditto.
12858         (get_parm_default_def_partitions): Ditto and make it static.
12859         (get_undefined_value_partitions): Ditto and make it static.
12860         (remove_ssa_form): Refactor call to init_var_map here.
12861         * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
12862         computation for loop region.
12863         (coalesce_partitions, compute_optimized_partition_bases): Ditto.
12864         (register_default_def): Delete.
12865         (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
12866         (parm_default_def_partition_arg): Ditto.
12867         (set_parm_default_def_partition): Ditto.
12868         (get_parm_default_def_partitions): Ditto and make it static.
12869         (get_undefined_value_partitions): Ditto and make it static.
12870         (coalesce_with_default, coalesce_with_default): Update comment.
12871         (create_coalesce_list_for_region): New func factored out from
12872         create_outofssa_var_map.
12873         (populate_coalesce_list_for_outofssa): New func factored out from
12874         create_outofssa_var_map and coalesce_ssa_name.
12875         (create_outofssa_var_map): Delete.
12876         (coalesce_ssa_name): Refactor to support live range computation.
12877         * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
12878         (get_parm_default_def_partitions): Delete.
12879         (get_undefined_value_partitions): Ditto.
12880         * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
12881         computation for loop region.
12882         (new_tree_live_info, loe_visit_block): Ditto.
12883         (live_worklist, set_var_live_on_entry): Ditto.
12884         (calculate_live_on_exit, verify_live_on_entry): Ditto.
12885         * tree-ssa-live.h (struct _var_map): New fields.
12886         (init_var_map): Change decl.
12887         (region_contains_p): New.
12889 2018-05-25  Bin Cheng  <bin.cheng@arm.com>
12891         * tree-ssa-live.h (live_merge_and_clear): Delete.
12893 2018-05-25  Richard Biener  <rguenther@suse.de>
12895         PR c++/85912
12896         * tree-dump.c (dequeue_and_dump): Remove access to removed
12897         operand 2 of a SWITCH_EXPR.
12899 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
12901         * doc/sourcebuild.texi (vect_double_cond_arith): Include
12902         multiplication and division.
12903         * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
12904         (cond_udiv@var{m}, cond_umod@var{m}): Document.
12905         * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
12906         (cond_udiv_optab, cond_umod_optab): New optabs.
12907         * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
12908         (IFN_COND_RDIV): New internal functions.
12909         * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
12910         TRUNC_MOD_EXPR and RDIV_EXPR.
12911         * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
12912         * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
12913         New unspecs.
12914         (SVE_INT_BINARY): Include mult.
12915         (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
12916         (optab, sve_int_op): Handle mult.
12917         (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
12918         UNSPEC_COND_DIV.
12919         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
12920         for SVE_INT_BINARY_SD.
12922 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
12924         * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
12925         (optab, sve_int_op): Handle div and udiv.
12926         * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
12927         for SVE_INT_BINARY_SD.
12928         (*<optab><mode>3): New insn for the same.
12930 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
12932         * tree-vect-patterns.c: Include predict.h.
12933         (vect_recog_divmod_pattern): Restrict check for division support
12934         to when optimizing for size.
12936 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
12938         * doc/sourcebuild.texi (vect_double_cond_arith: Document.
12939         * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
12940         (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
12941         (gimple_match_op::set_op): Likewise.
12942         (gimple_resimplify4): Declare.
12943         * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
12944         (expr::gen_transform): Likewise.
12945         (decision_tree::gen): Generate a simplification routine for 4 operands.
12946         * gimple-match-head.c (gimple_simplify): Add an overload for
12947         4 operands.  In the top-level function, handle up to 4 call
12948         arguments and call gimple_resimplify4.
12949         (gimple_resimplify4): New function.
12950         (build_call_internal): Pass a fourth operand.
12951         (maybe_push_to_seq): Likewise.
12952         * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
12953         Fold VEC_COND_EXPRs of an operation and a default value into
12954         an IFN_COND_* function if possible.
12955         * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
12956         New unspecs.
12957         (SVE_COND_FP_BINARY): Include them.
12958         (optab, sve_fp_op): Handle them.
12959         (SVE_INT_BINARY_REV): New code iterator.
12960         (SVE_COND_FP_BINARY_REV): New int iterator.
12961         (commutative): New int attribute.
12962         * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
12963         Declare.
12964         * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
12965         function.
12966         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
12967         (*cond_<optab><mode>): New patterns for reversed operands.
12969 2018-05-25  Richard Biener  <rguenther@suse.de>
12971         * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
12972         (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
12973         (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
12974         (STMT_VINFO_GROUPED_ACCESS): Adjust.
12975         * tree-vect-data-refs.c (everywhere): Adjust users.
12976         * tree-vect-loop.c (everywhere): Likewise.
12977         * tree-vect-slp.c (everywhere): Likewise.
12978         * tree-vect-stmts.c (everywhere): Likewise.
12979         * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
12981 2018-05-25  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
12983         * configure.ac (gcc_cv_as_section_has_e): Move to common section.
12984         Rename to...
12985         (gcc_cv_as_section_exclude): ... this.
12986         Try Solaris as #exclude syntax.
12987         * configure: Regenerate.
12988         * config.in: Regenerate.
12989         * config/i386/i386.c (i386_solaris_elf_named_section): Handle
12990         SECTION_EXCLUDE.
12991         * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
12992         [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
12994         * varasm.c (default_elf_asm_named_section): Don't check if
12995         HAVE_GAS_SECTION_EXCLUDE is defined.
12997 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
12999         * doc/md.texi: Update the documentation of the cond_* optabs
13000         to mention the new final operand.  Fix GET_MODE_NUNITS call.
13001         Describe the scalar case too.
13002         * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
13003         * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
13004         instead of 2.
13005         (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
13006         (get_conditional_internal_fn): Update comment.
13007         * tree-vect-loop.c (vectorizable_reduction): Pass the original
13008         accumulator value as a final argument to conditional functions.
13009         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
13010         a define_expand and add an "else" operand.  Assert for now that
13011         the else operand is equal to operand 2.  Use SVE_INT_BINARY and
13012         SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
13013         (*cond_<optab><mode>): New patterns.
13014         * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
13015         (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13016         (UNSPEC_COND_EOR): Delete.
13017         (optab): Remove associated mappings.
13018         (SVE_INT_BINARY): New code iterator.
13019         (sve_int_op): Remove int attribute and add "minus" to the code
13020         attribute.
13021         (SVE_COND_INT_OP): Delete.
13022         (SVE_COND_FP_OP): Rename to...
13023         (SVE_COND_FP_BINARY): ...this.
13025 2018-05-25  Richard Sandiford  <richard.sandiford@linaro.org>
13027         * optabs.c (can_reuse_operands_p): New function.
13028         (maybe_legitimize_operands): Try to reuse the results for
13029         earlier operands.
13031 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
13033         * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
13034         Add {q} suffix to insn mnemonic.
13036 2018-05-23  Jozef Lawrynowicz  <jozef.l@mittosystems.com>
13038         * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
13039         (msp430_warn_func_return): New.
13041 2018-05-24  Roger Sayle  <roger@nextmovesoftware.com>
13043         * fold-const.c (tree_nonzero_bits): New function.
13044         * fold-const.h (tree_nonzero_bits): Likewise.
13045         * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
13046         friends.  POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
13048 2018-05-24  H.J. Lu  <hongjiu.lu@intel.com>
13050         PR target/85900
13051         PR target/85345
13052         * varasm.c (assemble_alias): Check ifunc_resolver only on
13053         FUNCTION_DECL.
13055 2018-05-24  Uros Bizjak  <ubizjak@gmail.com>
13057         PR target/85903
13058         * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
13059         when memory input operand is handled.
13061 2018-05-24  Luis Machado  <luis.machado@linaro.org>
13063         * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
13064         global.
13065         (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
13067 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
13069         * match.pd: Delay FMA folds until after vectorization.
13071 2018-05-24  Andre Vieira  <andre.simoesdiasvieira@arm.com>
13073         PR target/83009
13074         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
13075         address check not strict.
13077 2018-05-24  Richard Sandiford  <richard.sandiford@linaro.org>
13079         * gimple-match.h (gimple_match_op): New class.
13080         (mprts_hook): Replace parameters with a gimple_match_op *.
13081         (maybe_build_generic_op): Likewise.
13082         (gimple_simplified_result_is_gimple_val): Replace parameters with
13083         a const gimple_match_op *.
13084         (gimple_simplify): Replace code_helper * and tree * parameters with
13085         a gimple_match_op * parameter.
13086         (gimple_resimplify1): Replace code_helper *, tree and tree *
13087         parameters with a gimple_match_op * parameter.
13088         (gimple_resimplify2): Likewise.
13089         (gimple_resimplify3): Likewise.
13090         (maybe_push_res_to_seq): Replace code_helper, tree and tree *
13091         parameters with a gimple_match_op * parameter.
13092         * gimple-match-head.c (gimple_simplify): Change prototypes of
13093         auto-generated functions to take a gimple_match_op * instead of
13094         separate code_helper * and tree * parameters.  Make the same
13095         change in the top-level overload and update calls to the
13096         gimple_resimplify routines.  Update calls to the auto-generated
13097         functions and to maybe_push_res_to_seq in the publicly-facing
13098         operation-specific gimple_simplify overloads.
13099         (gimple_match_op::MAX_NUM_OPS): Define.
13100         (gimple_resimplify1): Replace rcode and ops with a single res_op
13101         parameter.  Update call to gimple_simplify.
13102         (gimple_resimplify2): Likewise.
13103         (gimple_resimplify3): Likewise.
13104         (mprts_hook): Replace parameters with a gimple_match_op *.
13105         (maybe_build_generic_op): Likewise.
13106         (build_call_internal): Replace type, nargs and ops with
13107         a gimple_match_op *.
13108         (maybe_push_res_to_seq): Replace res_code, type and ops parameters
13109         with a single gimple_match_op *.  Update calls to mprts_hook,
13110         build_call_internal and gimple_simplified_result_is_gimple_val.
13111         Factor out code that is common to the tree_code and combined_fn cases.
13112         * genmatch.c (expr::gen_transform): Replace tem_code and
13113         tem_ops with a gimple_match_op called tem_op.  Update calls
13114         to the gimple_resimplify functions and maybe_push_res_to_seq.
13115         (dt_simplify::gen_1): Manipulate res_op instead of res_code and
13116         res_ops.  Update call to the gimple_resimplify functions.
13117         (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
13118         (decision_tree::gen): Make the functions take a gimple_match_op *
13119         called res_op instead of separate res_code and res_ops parameters.
13120         Update call accordingly.
13121         * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
13122         and ops with a single res_op parameter.  Update calls to
13123         maybe_build_generic_op and maybe_push_res_to_seq.
13124         (fold_stmt_1): Update calls to gimple_simplify and
13125         replace_stmt_with_simplification.
13126         (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
13127         and gimple_simplified_result_is_gimple_val.
13128         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
13129         gimple_simplify.
13130         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
13131         with a gimple_match_op *.
13132         (vn_nary_build_or_lookup): Likewise.  Update call to
13133         vn_nary_build_or_lookup_1.
13134         (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
13135         gimple_match_op *.  Update calls to the gimple_resimplify routines
13136         and to gimple_simplified_result_is_gimple_val.
13137         (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
13138         Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
13139         (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
13140         (visit_nary_op): Likewise.
13141         (visit_reference_op_load): Likewise.
13143 2018-05-23  Luis Machado  <luis.machado@linaro.org>
13145         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
13146         modifier for printing the step amount.
13148 2018-05-23  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
13150         PR target/78849
13151         * tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
13152         types.
13154 2018-05-23  Segher Boessenkool  <segher@kernel.crashing.org>
13156         * doc/sourcebuild.texi (Endianness): New subsubsection.
13158 2018-05-23  Luis Machado  <luis.machado@linaro.org>
13160         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13161         <prefetch_dynamic_strides>: New const bool field.
13162         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13163         prefetch_dynamic_strides.
13164         (exynosm1_prefetch_tune): Likewise.
13165         (thunderxt88_prefetch_tune): Likewise.
13166         (thunderx_prefetch_tune): Likewise.
13167         (thunderx2t99_prefetch_tune): Likewise.
13168         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
13169         false.
13170         (aarch64_override_options_internal): Update to set
13171         PARAM_PREFETCH_DYNAMIC_STRIDES.
13172         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
13173         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
13174         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
13175         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
13176         prefetch-dynamic-strides setting.
13178 2018-05-23  Luis Machado  <luis.machado@linaro.org>
13180         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
13181         <minimum_stride>: New const int field.
13182         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
13183         minimum_stride field defaulting to -1.
13184         (exynosm1_prefetch_tune): Likewise.
13185         (thunderxt88_prefetch_tune): Likewise.
13186         (thunderx_prefetch_tune): Likewise.
13187         (thunderx2t99_prefetch_tune): Likewise.
13188         (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
13189         <default_opt_level>: Set to 3.
13190         (aarch64_override_options_internal): Update to set
13191         PARAM_PREFETCH_MINIMUM_STRIDE.
13192         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
13193         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
13194         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
13195         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
13196         stride is constant and is below the minimum stride threshold.
13198 2018-05-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13200         * config/arm/arm-cpus.in (mode26): Delete.
13201         (armv4): Delete mode26 reference.
13202         * config/arm/arm.c (arm_configure_build_target): Delete use of
13203         isa_bit_mode26.
13205 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
13207         * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
13208         New insn pattern.
13209         (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
13210         Rewrite expander pattern.  Emit gen_floatunssi<mode>2_i387_with_xmm
13211         for non-SSE modes.
13212         (floatunsdisf2): Rewrite expander pattern.  Hanlde TARGET_AVX512F.
13213         (floatunsdidf2): Ditto.
13215 2018-05-23  Uros Bizjak  <ubizjak@gmail.com>
13217         * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
13218         (fixuns_trunc<mode>si2_avx512f): Ditto.
13219         (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
13220         (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
13221         Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
13223 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
13225         PR rtl-optimization/79985
13226         * df-scan.c (df_insn_refs_collect): Remove special case for
13227         global registers and asm statements.
13229 2018-05-23  Alexander Monakov  <amonakov@ispras.ru>
13231         * extend.texi (Global Register Variables): Rewrite the bullet list.
13232         Note that the register is available for allocation. Note that access
13233         via inline asm must use constraints. Add note about async-signal
13234         handlers. Remove paragraph about automagic register selection.
13236 2018-05-23  Richard Biener  <rguenther@suse.de>
13238         * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
13239         of fixed offset from memset VN.
13241 2018-05-23  Bill Schmidt  <wschmidt@linux.ibm.com>
13243         * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
13244         first_interp field.
13245         (alloc_cand_and_find_basis): Initialize first_interp field.
13246         (slsr_process_mul): Modify first_interp field.
13247         (slsr_process_add): Likewise.
13248         (slsr_process_cast): Modify first_interp field for each new
13249         interpretation.
13250         (slsr_process_copy): Likewise.
13251         (dump_candidate): Dump first_interp field.
13252         (replace_mult_candidate): Process all interpretations, not just
13253         subsequent ones.
13254         (replace_rhs_if_not_dup): Likewise.
13255         (replace_one_candidate): Likewise.
13257 2018-05-23  Wilco Dijkstra  <wdijkstr@arm.com>
13259         * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
13260         Add new boolean.
13261         (aarch64_needs_frame_chain): New function.
13262         (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
13264 2018-05-23  Sudakshina Das  <sudi.das@arm.com>
13266         PR target/84882
13267         * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
13268         Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
13269         * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
13270         * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
13271         as true for strict-align.
13272         (aarch64_can_inline_p): Perform checks even when callee has no
13273         attributes to check for strict alignment.
13274         * doc/extend.texi (AArch64 Function Attributes): Document
13275         no-strict-align.
13276         * doc/invoke.texi: (AArch64 Options): Likewise.
13278 2018-05-23  Richard Sandiford  <richard.sandiford@linaro.org>
13280         PR tree-optimization/85853
13281         * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
13282         the handling of the root of the node to...
13283         (vect_slp_analyze_node_operations_1): ...this new function,
13284         and run the whole thing with the child nodes' def types
13285         set according to their SLP node's def type.
13287 2018-05-23  Richard Biener  <rguenther@suse.de>
13289         PR middle-end/85874
13290         * tree-data-ref.c (create_runtime_alias_checks): Defer
13291         and ignore overflow warnings.
13293 2018-05-23  Yury Gribov  <tetra2005@gmail.com>
13295         PR tree-optimization/85822
13296         * tree-vrp.c (is_masked_range_test): Fix handling of negative
13297         constants.
13299 2018-05-23  Richard Biener  <rguenther@suse.de>
13301         * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
13302         memset constants via native_interpret_expr.
13304 2018-05-22  H.J. Lu  <hongjiu.lu@intel.com>
13306         PR target/85345
13307         * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
13308         attribute.
13309         (cgraph_node::create_alias): Likewise.
13310         (cgraph_node::get_availability): Check ifunc_resolver instead
13311         of looking up ifunc attribute.
13312         * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
13313         * varasm.c (do_assemble_alias): Likewise.
13314         (assemble_alias): Likewise.
13315         (default_binds_local_p_3): Likewise.
13316         * cgraph.h (cgraph_node): Add ifunc_resolver.
13317         (cgraph_node::only_called_directly_or_aliased_p): Return false
13318         for IFUNC resolver.
13319         * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
13320         attribute.
13321         * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
13322         is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
13323         (symtab_node::binds_to_current_def_p): Check ifunc_resolver
13324         instead of looking up ifunc attribute.
13326 2018-05-22  Luis Machado  <luis.machado@linaro.org>
13328         * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
13330 2018-05-22  Martin Sebor  <msebor@redhat.com>
13332         PR middle-end/85359
13333         * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
13334         only when expasion succeeds.
13335         (expand_builtin_strcmp): Same.
13336         (expand_builtin_strncmp): Same.
13338 2018-05-22  Martin Sebor  <msebor@redhat.com>
13340         * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
13342 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
13343             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13345         * config/aarch64/aarch64-ldpstp.md: Replace uses of
13346         aarch64_mem_pair_operand with memory_operand and delete operand swapping
13347         code.
13348         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
13349         Add check for legitimate_address.
13350         (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
13351         (aarch64_swap_ldrstr_operands): New.
13352         * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
13353         Define prototype.
13355 2018-05-22  Jackson Woodruff  <jackson.woodruff@arm.com>
13356             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13358         * config/aarch64/aarch64.md: New patterns to generate stp
13359         and ldp.
13360         (store_pair_sw, store_pair_dw): New patterns to generate stp for
13361         single words and double words.
13362         (load_pair_sw, load_pair_dw): Likewise.
13363         (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
13364         Delete.
13365         (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
13366         Delete.
13367         * config/aarch64/aarch64-ldpstp.md: Modify peephole
13368         for different mode ldpstp and add peephole for merged zero stores.
13369         Likewise for loads.
13370         * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
13371         Add size check.
13372         (aarch64_gen_store_pair): Rename calls to match new patterns.
13373         (aarch64_gen_load_pair): Rename calls to match new patterns.
13374         * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
13375         (load_pair<DREG:mode><DREG2:mode>): ... This.
13376         (store_pair<mode>): Rename to...
13377         (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
13378         * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
13379         New mode iterators.
13380         (V_INT_EQUIV): Handle SImode.
13381         * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
13382         New predicate.
13384 2018-05-22  Martin Sebor  <msebor@redhat.com>
13386         PR c/85623
13387         * calls.c (maybe_warn_nonstring_arg): Use string length to set
13388         or ajust the presumed bound on an operation to avoid unnecessary
13389         warnings.
13391 2018-05-22  Martin Sebor  <msebor@redhat.com>
13393         PR tree-optimization/85826
13394         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
13395         assuming that a DECL necesarily has a constant size.
13397 2018-05-22  Richard Sandiford  <richard.sandiford@linaro.org>
13399         PR middle-end/85862
13400         * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
13402 2018-05-22  Richard Biener  <rguenther@suse.de>
13404         PR tree-optimization/85834
13405         * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
13406         non-constant and non-zero memset arguments.
13408 2018-05-22  Martin Liska  <mliska@suse.cz>
13410         PR ipa/85607
13411         * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
13413 2018-05-22  Richard Biener  <rguenther@suse.de>
13415         PR tree-optimization/85863
13416         * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
13417         comparisons when vectype is specified.
13418         (vectorizable_condition): Do not specify vectype for
13419         vect_is_simple_cond when SLP vectorizing.
13421 2018-05-21  Michael Meissner  <meissner@linux.ibm.com>
13423         PR target/85657
13424         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
13425         define __ibm128 as long double.
13426         * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
13427         as a distinct type when IEEE 128-bit support is enabled.
13428         (init_float128_ieee): Fix up conversions between IFmode and IEEE
13429         128-bit types to use the correct functions.
13430         (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
13431         convert between 128-bit floating point types that have different
13432         modes but the same representation, instead of using gen_lowpart to
13433         makean alias.
13434         * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
13435         KFmode.
13436         (IFKF_reg): New attributes to give the register constraints for
13437         IFmode and KFmode.
13438         (extend<mode>tf2_internal): New insns to mark an explicit
13439         conversion between 128-bit floating point types that have a
13440         different mode but share the same representation.
13442 2018-05-21  Richard Sandiford  <richard.sandiford@linaro.org>
13444         PR tree-optimization/85814
13445         * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
13446         a null return from get_strinfo when unsharing the next
13447         strinfo in the chain.
13449 2018-05-21  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
13451         PR gcc/84923
13452         * varasm.c (weak_finish): Clean up weak_decls.
13454 2018-05-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13456         * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
13457         UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
13458         UNSPEC_UADALP values.
13459         * config/aarch64/iterators.md (ABAL): New int iterator.
13460         (ABDL2): Likewise.
13461         (ADALP): Likewise.
13462         (sur): Add mappings for the above.
13463         * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
13464         New define_insn.
13465         (aarch64_<sur>abal<mode>_4): Likewise.
13466         (aarch64_<sur>adalp<mode>_3): Likewise.
13467         (<sur>sadv16qi): New define_expand.
13469 2018-05-21  Alexander Nesterovskiy  <alexander.nesterovskiy@intel.com>
13471         * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
13472         (*movdf_internal): Ditto.
13473         (*rcpsf2_sse): Ditto.
13474         (*rsqrtsf2_sse): Ditto.
13475         (*sqrt<mode>2_sse): Ditto.
13477 2018-05-21  Tamar Christina  <tamar.christina@arm.com>
13479         * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
13480         eor3q<mode>4.
13481         (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
13482         * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
13483         veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
13484         vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
13485         vbcaxq_s64): New.
13486         * config/aarch64/arm_neon.h: Likewise.
13487         * config/aarch64/iterators.md (VQ_I): New.
13489 2018-05-21  Alexey Brodkin <abrodkin@synopsys.com>
13491         * config.gcc: Add arc/t-multilib-linux to tmake_file for
13492         arc*-*-linux*.
13493         * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
13494         MULTILIB_DIRNAMES
13496 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
13498         * config/nds32/constraints.md (S): New constraint.
13499         * config/nds32/nds32.md (call_internal): Use constraint S.
13500         (call_value_internal): Likewise.
13501         (sibcall_internal): Likewise.
13502         (sibcall_value_internal): Likewise.
13504 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
13505             Chung-Ju Wu  <jasonwucj@gmail.com>
13507         * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
13508         into consideration.
13510 2018-05-20  Kito Cheng  <kito.cheng@gmail.com>
13511             Chung-Ju Wu  <jasonwucj@gmail.com>
13513         * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
13514         (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
13515         (nds32_rtx_costs_impl): Simplify.
13516         (nds32_address_cost_impl): Simplify.
13517         (nds32_init_rtx_costs): New function.
13518         (nds32_rtx_costs_speed_prefer): Likewise.
13519         (nds32_rtx_costs_size_prefer): Likewise.
13520         (nds32_address_cost_speed_prefer): Likewise.
13521         (nds32_address_cost_speed_fwprop): Likewise.
13522         (nds32_address_cost_size_prefer): Likewise.
13523         * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
13524         * config/nds32/nds32.c (nds32_option_override): Use
13525         nds32_init_rtx_costs function.
13527 2018-05-20  Chung-Ju Wu  <jasonwucj@gmail.com>
13529         * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
13530         * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
13531         (TARGET_PIPELINE_N8): Likewise.
13532         (TARGET_PIPELINE_N10): Likewise.
13533         (TARGET_PIPELINE_N13): Likewise.
13534         (TARGET_PIPELINE_GRAYWOLF): Likewise.
13536 2018-05-19  Monk Chiang  <sh.chiang04@gmail.com>
13538         * config/nds32/nds32-fpu.md: Update copyright year.
13540 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
13542         * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
13544 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
13546         * config/nds32/nds32.c
13547         (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
13548         * config/nds32/nds32.opt (minline-asm-r15): New option.
13550 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
13552         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
13553         MASK_HW_ABS.
13554         * config/nds32/nds32.md (abssi2): New pattern.
13556 2018-05-19  Uros Bizjak  <ubizjak@gmail.com>
13558         * config/i386/i386.md (rex64namesuffix): New mode attribute.
13559         * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
13560         Merge insn pattern from sse_cvtsi2ss<round_name> and
13561         sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
13562         (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
13563         from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
13564         using SWI48 mode iterator.
13565         (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
13566         sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
13567         (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
13568         pattern from sse_cvttss2si<round_saeonly_name>
13569         and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
13570         (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
13571         from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
13572         using SWI48 mode iterator.
13573         (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
13574         insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
13575         avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
13576         (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
13577         from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
13578         using SWI48 mode iterator.
13579         (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
13580         insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
13581         avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
13582         (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
13583         sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
13584         SWI48 mode iterator.
13585         (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
13586         sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
13587         (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
13588         pattern from sse_cvttsd2si<round_saeonly_name>
13589         and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
13591 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
13593         * config/nds32/nds32-md-auxiliary.c
13594         (nds32_valid_smw_lwm_base_p): Refine.
13595         (nds32_output_smw_single_word): Refine.
13596         (nds32_output_smw_double_word): New.
13597         * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
13599 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
13601         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
13602         (nds32_output_stack_pop): Refine.
13603         (nds32_expand_unaligned_load): Refine.
13604         (nds32_expand_unaligned_store): Refine.
13606 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
13607             Chung-Ju Wu  <jasonwucj@gmail.com>
13609         * config/nds32/constants.md: Add TP_REGNUM constant.
13610         (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
13611         UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
13612         UNSPEC_ADD32.
13613         * config/nds32/nds32-doubleword.md: Consider flag_pic.
13614         * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
13615         * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
13616         * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
13617         and PIC code generation.
13618         * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
13619         code generation.
13620         * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
13621         optimization.
13622         * config/nds32/nds32.md: Support TLS and PIC.
13623         * config/nds32/nds32.c: Support TLS and PIC.
13624         * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
13625         * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
13626         predicate.
13628 2018-05-19  Chung-Ju Wu  <jasonwucj@gmail.com>
13630         * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
13631         mode with E_ prefix.
13633 2018-05-19  Kuan-Lin Chen <kuanlinchentw@gmail.com>
13634             Chung-Ju Wu  <jasonwucj@gmail.com>
13636         * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
13637         * config/nds32/nds32-md-auxiliary.c
13638         (symbolic_reference_mentioned_p): New.
13639         (nds32_legitimize_ict_address): New.
13640         (nds32_expand_ict_move): New.
13641         (nds32_indirect_call_referenced_p): New.
13642         (nds32_symbol_binds_local_p): Delete.
13643         (nds32_long_call_p): Modify.
13644         * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
13645         * config/nds32/nds32-protos.h
13646         (symbolic_reference_mentioned_p): Declare.
13647         (nds32_legitimize_ict_address): Declare.
13648         (nds32_expand_ict_move): Declare.
13649         (nds32_indirect_call_referenced_p): Declare.
13650         * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
13651         (nds32_relax_group): Use nds32_ict_const_p as condition.
13652         * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
13653         (nds32_asm_file_start): Output ict_model directive in asm code.
13654         (nds32_legitimate_address_p): Consider indirect call.
13655         (nds32_print_operand): Consider indirect call.
13656         (nds32_print_operand_address): Consider indirect call.
13657         (nds32_insert_attributes): Handle "indirect_call" attribute.
13658         (TARGET_LEGITIMATE_ADDRESS_P): Define.
13659         (TARGET_LEGITIMATE_CONSTANT_P): Define.
13660         (TARGET_CANNOT_FORCE_CONST_MEM): Define.
13661         (TARGET_DELEGITIMIZE_ADDRESS): Define.
13662         (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
13663         * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
13664         (TARGET_ICT_MODEL_SMALL): Define.
13665         (TARGET_ICT_MODEL_LARGE): Define.
13666         * config/nds32/nds32.md (movsi): Consider ict model.
13667         (call, call_value): Consider ict model.
13668         (sibcall, sibcall_value): Consider ict model.
13669         * config/nds32/nds32.opt (mict-model): New option.
13670         * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
13671         model.
13673 2018-05-18  Kito Cheng  <kito.cheng@gmail.com>
13674             Monk Chiang  <sh.chiang04@gmail.com>
13675             Jim Wilson <jimw@sifive.com>
13677         * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
13678         Add support to parse rv32e*.  Clear MASK_RVE for rv32i and rv64i.
13679         * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
13680         * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
13681         __riscv_32e when TARGET_RVE.  Handle ABI_ILP32E as soft-float ABI.
13682         * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
13683         * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
13684         compute save_libcall_adjustment properly.
13685         (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
13686         (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
13687         * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
13688         (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
13689         (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
13690         (ABI_SPEC): Handle mabi=ilp32e.
13691         * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
13692         (RVE): Add RVE mask.
13693         * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
13694         <-march>: Add rv32e as an example.
13696 2018-05-18  Marc Glisse  <marc.glisse@inria.fr>
13698         PR c++/82899
13699         * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
13700         (intra_create_variable_infos): Handle C++ constructors.
13702 2018-05-18  Martin Liska  <mliska@suse.cz>
13704         * passes.def: Remove a redundant pass.
13706 2018-05-18  Eric Botcazou  <ebotcazou@adacore.com>
13708         PR bootstrap/85838
13709         * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
13711 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13713         * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
13714         (ARMv4): Update.
13715         (ARMv2, ARMv3, ARMv3m): Delete fgroups.
13716         (ARMv6m): Update.
13717         (armv2, armv2a, armv3, armv3m): Delete architectures.
13718         (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
13719         arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
13720         arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
13721         Delete cpus.
13722         * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
13723         (*mulsidi3adddi): Likewise.
13724         (mulsidi3): Likewise.
13725         (*mulsidi3_nov6): Likewise.
13726         (umulsidi3): Likewise.
13727         (umulsidi3_nov6): Likewise.
13728         (umaddsidi4): Likewise.
13729         (*umulsidi3adddi): Likewise.
13730         (smulsi3_highpart): Likewise.
13731         (*smulsi3_highpart_nov6): Likewise.
13732         (umulsi3_highpart): Likewise.
13733         (*umulsi3_highpart_nov6): Likewise.
13734         * config/arm/arm.h (arm_arch3m): Delete.
13735         * config/arm/arm.c (arm_arch3m): Delete.
13736         (arm_option_override_internal): Update armv3-related comment.
13737         (arm_configure_build_target): Delete use of isa_bit_mode32.
13738         (arm_option_reconfigure_globals): Delete set of arm_ach3m.
13739         (arm_rtx_costs_internal): Delete check of arm_arch3m.
13740         * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
13741         (mulsa3): Likewise.
13742         (mulusa3): Likewise.
13743         * config/arm/arm-protos.h (arm_arch3m): Delete.
13744         * config/arm/arm-tables.opt: Regenerate.
13745         * config/arm/arm-tune.md: Likewise.
13746         * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
13747         deleted architectures.
13749 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13751         * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
13752         (armv5t, armv5te): New features.
13753         (ARMv5, ARMv5e): Delete fgroups.
13754         (ARMv5t, ARMv5te): Adjust for above changes.
13755         (ARMv6m): Likewise.
13756         (armv5, armv5e): Delete arches.
13757         * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
13758         arm_arch5.
13759         (*call_reg_arm): Likewise.
13760         (*call_value_reg_armv5): Likewise.
13761         (*call_value_reg_arm): Likewise.
13762         (*call_symbol): Likewise.
13763         (*call_value_symbol): Likewise.
13764         (*sibcall_insn): Likewise.
13765         (*sibcall_value_insn): Likewise.
13766         (clzsi2): Likewise.
13767         (prefetch): Likewise.
13768         (define_split and define_peephole2 dependent on arm_arch5):
13769         Likewise.
13770         * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
13771         arm_arch5e.
13772         (TARGET_ARM_QBIT): Likewise.
13773         (TARGET_DSP_MULTIPLY): Likewise.
13774         (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
13775         (arm_arch5, arm_arch5e): Delete.
13776         (arm_arch5t, arm_arch5te): Declare.
13777         * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
13778         (arm_arch5t): Declare.
13779         (arm_option_reconfigure_globals): Update for the above.
13780         (arm_options_perform_arch_sanity_checks): Update comment, replace
13781         use of arm_arch5 with arm_arch5t.
13782         (use_return_insn): Likewise.
13783         (arm_emit_call_insn): Likewise.
13784         (output_return_instruction): Likewise.
13785         (arm_final_prescan_insn): Likewise.
13786         (arm_coproc_builtin_available): Likewise.
13787         * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
13788         arm_arch5e with arm_arch5t and arm_arch5te.
13789         * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
13790         (arm_arch5t, arm_arch5te): Declare.
13791         * config/arm/arm-tables.opt: Regenerate.
13792         * config/arm/t-arm-elf: Remove references to armv5, armv5e.
13793         * config/arm/t-multilib: Likewise.
13794         * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
13795         instead of arm_arch5.
13796         (*call_reg_thumb1): Likewise.
13797         (*call_value_reg_thumb1_v5): Likewise.
13798         (*call_value_reg_thumb1): Likewise.
13799         * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
13800         unreachable path.
13801         * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
13803 2018-05-18  Martin Liska  <mliska@suse.cz>
13805         PR gcov-profile/84846
13806         * doc/gcov.texi: Document -t option of gcov tool.
13808 2018-05-18  Martin Liska  <mliska@suse.cz>
13810         PR gcov-profile/84846
13811         * gcov.c (print_usage): Add new -t option.
13812         (process_args): Handle the option.
13813         (generate_results): Use stdout as output when requested by
13814         the option.
13816 2018-05-18  Martin Liska  <mliska@suse.cz>
13818         PR gcov-profile/84846
13819         * coverage.c (coverage_init): Write PWD to .gcno file.
13820         * doc/gcov.texi: Document how working directory is printed.
13821         * gcov-dump.c (dump_gcov_file): Print PWD.
13822         * gcov.c (output_intermediate_file): Likewise.
13823         (read_graph_file): Read PWD string.
13824         (output_lines): Print PWD.
13826 2018-05-18  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
13828         PR middle-end/85817
13829         * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
13830         for retval and return false if all args to phi are zero.
13832 2018-05-18  Richard Biener  <rguenther@suse.de>
13834         * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
13835         method.
13836         (evrp_dom_walker::before_dom_children): Call it.
13838 2018-05-18  Richard Biener  <rguenther@suse.de>
13840         * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
13841         results when processing array refs with variable index.
13843 2018-05-18  Toon Moene  <toon@moene.org>
13845         * doc/invoke.texi: Move -floop-unroll-and-jam documentation
13846         directly after that of -floop-interchange. Indicate that both
13847         options are enabled by default when specifying -O3.
13849 2018-05-18  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
13851         * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
13852         iterator.  Delete separate integer-mode vec_set<mode> expander.
13853         (aarch64_simd_vec_setv2di): Delete.
13854         (vec_setv2di): Delete.
13855         (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
13856         Use VALL_F16 mode iterator.  Add LD1 alternative and use vwcore for
13857         the "w, r" alternative.
13859 2018-05-18  Martin Liska  <mliska@suse.cz>
13861         * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
13862         * tree-pass.h (make_pass_lower_switch_O0): New function.
13863         * tree-switch-conversion.c (node_has_low_bound): Remove.
13864         (node_has_high_bound): Likewise.
13865         (node_is_bounded): Likewise.
13866         (class pass_lower_switch): Make it a template type and create
13867         two instances.
13868         (pass_lower_switch::execute): Add template argument.
13869         (make_pass_lower_switch): New function.
13870         (make_pass_lower_switch_O0): New function.
13871         (do_jump_if_equal): Remove.
13872         (emit_case_nodes): Simplify to just handle all 3 cases and leave
13873         all the hard work to tree optimization passes.
13875 2018-05-18  Martin Liska  <mliska@suse.cz>
13877         * dbgcnt.c (limit_low): Renamed from limit.
13878         (limit_high): New variable.
13879         (dbg_cnt_is_enabled): Check for upper limit.
13880         (dbg_cnt): Adjust dumping.
13881         (dbg_cnt_set_limit_by_index): Add new argument for high
13882         value.
13883         (dbg_cnt_set_limit_by_name): Likewise.
13884         (dbg_cnt_process_single_pair): Parse new format.
13885         (dbg_cnt_process_opt): Use strtok.
13886         (dbg_cnt_list_all_counters): Remove 'value' and add
13887         'limit_high'.
13888         * doc/invoke.texi: Document changes.
13890 2018-05-18  Richard Sandiford  <richard.sandiford@linaro.org>
13892         * doc/sourcebuild.texi (scalar_all_fma): Document.
13893         * tree.def (FMA_EXPR): Delete.
13894         * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
13895         * internal-fn.c (ternary_direct): New macro.
13896         (expand_ternary_optab_fn): Likewise.
13897         (direct_ternary_optab_supported_p): Likewise.
13898         * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
13899         * builtins.c (fold_builtin_fma): Delete.
13900         (fold_builtin_3): Don't call it.
13901         * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
13902         * expr.c (expand_expr_real_2): Likewise.
13903         * fold-const.c (operand_equal_p): Likewise.
13904         (fold_ternary_loc): Likewise.
13905         * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
13906         * gimple.c (DEFTREECODE): Likewise.
13907         * gimplify.c (gimplify_expr): Likewise.
13908         * optabs-tree.c (optab_for_tree_code): Likewise.
13909         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
13910         * tree-eh.c (operation_could_trap_p): Likewise.
13911         (stmt_could_throw_1_p): Likewise.
13912         * tree-inline.c (estimate_operator_cost): Likewise.
13913         * tree-pretty-print.c (dump_generic_node): Likewise.
13914         (op_code_prio): Likewise.
13915         * tree-ssa-loop-im.c (stmt_cost): Likewise.
13916         * tree-ssa-operands.c (get_expr_operands): Likewise.
13917         * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
13918         * fold-const-call.h (fold_fma): Delete.
13919         * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
13920         CFN_FNMA and CFN_FNMS.
13921         (fold_fma): Delete.
13922         * genmatch.c (combined_fn): New enum.
13923         (commutative_ternary_tree_code): Remove FMA_EXPR handling.
13924         (commutative_op): New function.
13925         (commutate): Use it.  Handle more than 2 operands.
13926         (dt_operand::gen_gimple_expr): Use commutative_op.
13927         (parser::parse_expr): Allow :c to be used with non-binary
13928         operators if the commutative operand is known.
13929         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
13930         CFN_FMS, CFN_FNMA and CFN_FNMS.
13931         (backprop::process_assign_use): Remove FMA_EXPR handling.
13932         * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
13933         (gen_hsa_fma): New function.
13934         (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
13935         IFN_FNMA and IFN_FNMS.
13936         * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
13937         * gimple-fold.h (follow_all_ssa_edges): Declare.
13938         * gimple-fold.c (follow_all_ssa_edges): New function.
13939         * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
13940         gimple_build interface and use follow_all_ssa_edges to fold the result.
13941         (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
13942         instead of checking for optabs directly.
13943         * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
13944         rather than FMA_EXPRs.
13945         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
13946         call to IFN_FMA instead of an FMA_EXPR.
13948 2018-05-17  Jim Wilson  <jimw@sifive.com>
13950         * expr.c (do_tablejump): When converting index to Pmode, if we have a
13951         sign extended promoted subreg, and the range does not have the sign bit
13952         set, then do a sign extend.
13954         * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
13955         test, check for sign extended subreg and/or constant operands, and
13956         do a sign extend in that case.
13958 2018-05-17  Steve Ellcey  <sellcey@cavium.com>
13960         * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
13961         (thunderx2t99_multiple): Delete psuedo-units from used cpus.
13962         Add untyped.
13963         (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
13964         Change logics_shift_reg to logics_shift_imm.
13965         (thunderx2t99_fp_loadpair_basic): Delete.
13966         (thunderx2t99_fp_storepair_basic): Delete.
13967         (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
13968         (thunderx2t99_asimd_polynomial): Delete.
13969         (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
13970         and neon_fp_mul_d_scalar_q.
13971         (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
13972         (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
13973         (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
13974         (thunderx2t99_asimd_lut): Add missing tbl types.
13975         (thunderx2t99_asimd_ext): Delete.
13976         (thunderx2t99_asimd_load1_1_mult): Delete.
13977         (thunderx2t99_asimd_load1_2_mult): Delete.
13978         (thunderx2t99_asimd_load1_ldp): New.
13979         (thunderx2t99_asimd_load1): New.
13980         (thunderx2t99_asimd_load2): Add missing *load2* types.
13981         (thunderx2t99_asimd_load3): New.
13982         (thunderx2t99_asimd_load4): New.
13983         (thunderx2t99_asimd_store1_1_mult): Delete.
13984         (thunderx2t99_asimd_store1_2_mult): Delete.
13985         (thunderx2t99_asimd_store2_mult): Delete.
13986         (thunderx2t99_asimd_store2_onelane): Delete.
13987         (thunderx2t99_asimd_store_stp): New.
13988         (thunderx2t99_asimd_store1): New.
13989         (thunderx2t99_asimd_store2): New.
13990         (thunderx2t99_asimd_store3): New.
13991         (thunderx2t99_asimd_store4): New.
13993 2018-05-17  Jerome Lambourg  <lambourg@adacore.com>
13995         * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
13996         #include <stdint.h>.  Replace intptr_t with __INTPTR_TYPE__.
13998 2018-05-17  Pat Haugen  <pthaugen@us.ibm.com>
13999             Segher Boessenkool  <segher@kernel.crashing.org>
14001         PR target/85698
14002         * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
14003         operand.
14005 2018-05-17  Richard Biener  <rguenther@suse.de>
14007         * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
14008         for pruning loop and prune defs feeding only already visited PHIs.
14010 2018-05-17  Richard Biener  <rguenther@suse.de>
14012         * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
14014 2018-05-17  Bin Cheng  <bin.cheng@arm.com>
14015             Richard Biener  <rguenther@suse.de>
14017         PR tree-optimization/85793
14018         * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
14019         for VMAT_ELEMENTWISE.
14021 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
14023         * internal-fn.h (lookup_internal_fn): Declare
14024         * internal-fn.c (lookup_internal_fn): New function.
14025         * gimple.c (gimple_build_call_from_tree): Handle calls to
14026         internal functions.
14027         * gimple-pretty-print.c (dump_gimple_call): Print "." before
14028         internal function names.
14029         * tree-pretty-print.c (dump_generic_node): Likewise.
14030         * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
14032 2018-05-17  Richard Sandiford  <richard.sandiford@linaro.org>
14034         * gimple-fold.h (gimple_build): Make the function forms take
14035         combined_fn rather than built_in_function.
14036         (gimple_simplify): Likewise.
14037         * gimple-match-head.c (gimple_simplify): Likewise.
14038         * gimple-fold.c (gimple_build): Likewise.
14039         * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
14040         rather than gimple_build_call_internal.
14041         (get_initial_defs_for_reduction): Likewise.
14042         (vect_create_epilog_for_reduction): Likewise.
14043         (vectorizable_live_operation): Likewise.
14045 2018-05-17  Martin Liska  <mliska@suse.cz>
14047         * gimple-ssa-sprintf.c (format_directive): Do not use
14048         space in between 'G_' and '('.
14050 2018-05-17  Jakub Jelinek  <jakub@redhat.com>
14052         PR target/85323
14053         * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
14054         even if the mask is not all ones.
14056         PR target/85323
14057         * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
14058         vector.
14059         (ix86_gimple_fold_builtin): Likewise.
14061         PR target/85323
14062         * config/i386/i386.c: Include tree-vector-builder.h.
14063         (ix86_vector_shift_count): New function.
14064         (ix86_fold_builtin): Fold shift builtins by scalar count.
14065         (ix86_gimple_fold_builtin): Likewise.
14067         * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
14068         _mm512_setzero): New intrinsics.
14070 2018-05-17  James Greenhalgh  <james.greenhalgh@arm.com>
14071             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14073         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
14074         code generation for cases where splatting a value is not useful.
14075         * simplify-rtx.c (simplify_ternary_operation): Simplify
14076         vec_merge across a vec_duplicate and a paradoxical subreg forming
14077         a vector mode to a vec_concat.
14079 2018-05-17  Olga Makhotina  <olga.makhotina@intel.com>
14081         * config.gcc: Support "goldmont-plus".
14082         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
14083         "goldmont-plus".
14084         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14085         PROCESSOR_GOLDMONT_PLUS.
14086         * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
14087         (processor_target_table): Add "goldmont-plus".
14088         (PTA_GOLDMONT_PLUS): Define.
14089         (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
14090         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
14091         (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
14092         (fold_builtin_cpu): Add "goldmont-plus".
14093         (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
14094         (ix86_option_override_internal): Add "goldmont-plus".
14095         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
14096         (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
14097         * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
14098         * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
14100 2018-05-17  Richard Biener  <rguenther@suse.de>
14102         PR tree-optimization/85757
14103         * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
14104         remove defs that only feed that PHI from further processing.
14106 2018-05-16  Jim Wilson  <jimw@sifive.com>
14108         * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
14109         asterisk to name.
14110         (<optab>di3_mask, <optab>di3_mask_1): Likewise.
14112 2018-05-16  Mark Wielaard  <mark@klomp.org>
14114         * dwarf2out.c (count_index_strings): New function.
14115         (output_indirect_strings): Call count_index_strings and generate
14116         header for dwarf_version >= 5.
14118 2018-05-16  Mark Wielaard  <mark@klomp.org>
14120         * dwarf2out.c (dwarf_FORM): New function.
14121         (set_indirect_string): Use dwarf_FORM.
14122         (reset_indirect_string): Likewise.
14123         (size_of_die): Likewise.
14124         (value_format): Likewise.
14125         (output_die): Likewise.
14126         (add_skeleton_AT_string): Likewise.
14127         (output_macinfo_op): Likewise.
14128         (index_string): Likewise.
14129         (output_index_string_offset): Likewise.
14130         (output_index_string): Likewise.
14131         (count_index_strings): Likewise.
14133 2018-05-16  Carl Love  <cel@us.ibm.com>
14135         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
14136         dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
14138 2018-05-16  Martin Jambor  <mjambor@suse.cz>
14140         * ipa-prop.c (ipa_free_all_edge_args): Remove.
14141         * ipa-prop.h (ipa_free_all_edge_args): Likewise.
14143 2018-05-16  Wilco Dijkstra  <wdijkstr@arm.com>
14145         * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
14146         (fnma<mode>4): Likewise.
14147         (fms<mode>4): Likewise.
14148         (fnms<mode>4): Likewise.
14149         (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
14150         (aarch64_fnma<mode>4): Likewise.
14151         (aarch64_fms<mode>4): Likewise.
14152         (aarch64_fnms<mode>4): Likewise.
14153         (aarch64_fnmadd<mode>4): Likewise.
14155 2018-05-16  Jason Merrill  <jason@redhat.com>
14157         * tree.c (warn_deprecated_use): Return bool.  Simplify logic.
14159 2018-05-16  Richard Biener  <rguenther@suse.de>
14161         * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
14162         (dump_stmt_cost): Declare.
14163         (add_stmt_cost): Dump cost we add.
14164         (add_stmt_costs): New function.
14165         (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
14166         No longer exported.
14167         (vect_analyze_stmt): Adjust prototype.
14168         (vectorizable_condition): Likewise.
14169         (vectorizable_live_operation): Likewise.
14170         (vectorizable_reduction): Likewise.
14171         (vectorizable_induction): Likewise.
14172         * tree-vect-loop.c (vect_analyze_loop_operations): Create local
14173         cost vector to pass to vectorizable_ and record afterwards.
14174         (vect_model_reduction_cost): Take cost vector argument and adjust.
14175         (vect_model_induction_cost): Likewise.
14176         (vectorizable_reduction): Likewise.
14177         (vectorizable_induction): Likewise.
14178         (vectorizable_live_operation): Likewise.
14179         * tree-vect-slp.c (vect_create_new_slp_node): Initialize
14180         SLP_TREE_NUMBER_OF_VEC_STMTS.
14181         (vect_analyze_slp_cost_1): Remove.
14182         (vect_analyze_slp_cost): Likewise.
14183         (vect_slp_analyze_node_operations): Take visited args and
14184         a target cost vector.  Avoid processing already visited stmt sets.
14185         (vect_slp_analyze_operations): Use a local cost vector to gather
14186         costs and register those of non-discarded instances.
14187         (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
14188         (vect_schedule_slp_instance): Remove copying of
14189         SLP_TREE_NUMBER_OF_VEC_STMTS.  Instead assert that it is not
14190         zero.
14191         * tree-vect-stmts.c (record_stmt_cost): Remove path directly
14192         adding cost.  Record cost entry location.
14193         (vect_prologue_cost_for_slp_op): Function to compute cost of
14194         a constant or invariant generated for SLP vect in the prologue,
14195         split out from vect_analyze_slp_cost_1.
14196         (vect_model_simple_cost): Make static.  Adjust for SLP costing.
14197         (vect_model_promotion_demotion_cost): Likewise.
14198         (vect_model_store_cost): Likewise, make static.
14199         (vect_model_load_cost): Likewise.
14200         (vectorizable_bswap): Add cost vector arg and adjust.
14201         (vectorizable_call): Likewise.
14202         (vectorizable_simd_clone_call): Likewise.
14203         (vectorizable_conversion): Likewise.
14204         (vectorizable_assignment): Likewise.
14205         (vectorizable_shift): Likewise.
14206         (vectorizable_operation): Likewise.
14207         (vectorizable_store): Likewise.
14208         (vectorizable_load): Likewise.
14209         (vectorizable_condition): Likewise.
14210         (vectorizable_comparison): Likewise.
14211         (can_vectorize_live_stmts): Likewise.
14212         (vect_analyze_stmt): Likewise.
14213         (vect_transform_stmt): Adjust calls to vectorizable_*.
14214         * tree-vectorizer.c: Include gimple-pretty-print.h.
14215         (dump_stmt_cost): New function.
14217 2018-05-16  Richard Biener  <rguenther@suse.de>
14219         * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
14220         * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
14221         * tree-ssa-dse.c: Include tree-ssa-loop.h.
14222         (check_name): New callback.
14223         (dse_classify_store): Track cycles via a visited bitmap of PHI
14224         defs and simplify handling of in-loop and across loop dead stores
14225         and properly fail for loop-variant refs.  Handle byte-tracking with
14226         multiple defs.  Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
14227         limiting the walk.
14229 2018-05-16  Richard Sandiford  <richard.sandiford@linaro.org>
14231         * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
14232         (vect_get_mask_type_for_stmt): Likewise.
14233         * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
14234         split out from...
14235         (vect_build_slp_tree_1): ...here.  Use vect_get_vector_types_for_stmt
14236         to determine the statement's vector type and the vector type that
14237         should be used for calculating nunits.  Deal with cases in which
14238         the type has to be deferred.
14239         (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
14240         and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
14241         * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
14242         (vect_determine_vf_for_stmt): New functions, split out from...
14243         (vect_determine_vectorization_factor): ...here.
14244         * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
14245         (vect_get_mask_type_for_stmt): New functions, split out from
14246         vect_determine_vectorization_factor.
14248 2018-05-16  Richard Biener  <rguenther@suse.de>
14250         * tree-cfg.c (verify_gimple_assign_ternary): Properly
14251         verify the [VEC_]COND_EXPR embedded comparison.
14253 2018-05-15  Martin Sebor  <msebor@redhat.com>
14255         PR tree-optimization/85753
14256         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
14257         RECORD_TYPE in addition to ARRAY_TYPE.
14259 2018-05-15  Martin Sebor  <msebor@redhat.com>
14261         PR middle-end/85643
14262         * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
14264 2018-05-15  Richard Biener  <rguenther@suse.de>
14266         * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
14267         add by_clobber_p one.  Change algorithm to collect all defs
14268         representing uses we need to walk and try reducing them to
14269         a single one before failing.
14270         (dse_dom_walker::dse_optimize_stmt): Adjust.
14272 2018-05-13  Mark Wielaard  <mark@klomp.org>
14274         * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
14275         (size_of_loc_descr): Likewise.
14276         (output_loc_operands): Likewise.
14277         (output_loc_operands_raw): Likewise.
14278         (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
14279         (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
14280         (hash_loc_operands): Likewise.
14281         (compare_loc_operands): Likewise.
14283 2018-05-14  Mark Wielaard  <mark@klomp.org>
14285         * dwarf2out.c (count_index_addrs): New function.
14286         (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
14288 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14290         PR tree-optimization/83648
14291         * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
14292         return value as malloc candidate.
14294 2018-05-15  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14296         PR ipa/85734
14297         * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
14298         param as true in call to suggest_attribute.
14300 2018-05-14  Segher Boessenkool  <segher@kernel.crashing.org>
14302         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
14303         -mreadonly-in-sdata.
14305 2018-05-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
14307         * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
14308         New pattern.
14309         (aarch64_crypto_aesd_fused): Likewise.
14311 2018-05-14  Wilco Dijkstra  <wdijkstr@arm.com>
14313         * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
14314         (movsi_aarch64): Likewise.
14315         (load_pairsi): Likewise.
14316         (load_pairdi): Likewise.
14317         (store_pairsi): Likewise.
14318         (store_pairdi): Likewise.
14319         (load_pairsf): Likewise.
14320         (load_pairdf): Likewise.
14321         (store_pairsf): Likewise.
14322         (store_pairdf): Likewise.
14323         (zero_extend): Likewise.
14324         (trunc): Swap alternatives.
14325         (fcvt_target): Add '?' to prefer w over r.
14327 2018-05-14  Jakub Jelinek  <jakub@redhat.com>
14329         PR target/85756
14330         * config/i386/i386.md: Disallow non-commutative arithmetics in
14331         last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
14332         optimization.  Use COMMUTATIVE_ARITH_P test rather than != MINUS
14333         in the peephole2 before it.
14335 2018-05-14  Sebastian Peryt  <sebastian.peryt@intel.com>
14337         * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
14338         OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
14339         (ix86_handle_option): Handle -mcldemote.
14340         * config.gcc: New header.
14341         * config/i386/cldemoteintrin.h: New file.
14342         * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
14343         * config/i386/driver-i386.c (host_detect_local_cpu): Detect
14344         -mcldemote.
14345         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14346         OPTION_MASK_ISA_CLDEMOTE.
14347         * config/i386/i386.c (ix86_target_string): Add -mcldemote.
14348         (ix86_valid_target_attribute_inner_p): Ditto.
14349         (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
14350         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
14351         (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
14352         * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
14353         * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
14354         (cldemote): New.
14355         * config/i386/i386.opt: Add -mcldemote.
14356         * config/i386/x86intrin.h: New header.
14357         * doc/invoke.texi: Add -mcldemote.
14359 2018-05-14  Richard Biener  <rguenther@suse.de>
14361         * doc/match-and-simplify.texi: Adjust :s documentation.
14363 2018-05-14  Alexander Monakov  <amonakov@ispras.ru>
14365         * sort.cc (REORDER_23): Pass the type for the temporaries instead of
14366         intended memcpy size.
14367         (REORDER_45): Likewise.
14369 2018-05-13  Alexander Monakov  <amonakov@ispras.ru>
14371         * sort.cc: New file.
14372         * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
14373         * vec.c (qsort_chk): Use gcc_qsort.
14374         * Makefile.in (OBJS-libcommon): Add sort.o.
14375         (build/sort.o): New target.  Use it...
14376         (BUILD_RTL): ... here, and...
14377         (build/gencfn-macros): ... here, and...
14378         (build/genmatch): ... here.
14380 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
14381             Chung-Ju Wu  <jasonwucj@gmail.com>
14383         * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
14384         * config/nds32/nds32-graywolf.md: New file.
14385         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
14386         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
14387         pipeline.
14388         * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
14389         * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
14390         * config/nds32/nds32.md (pipeline_model): Add graywolf.
14391         * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
14392         * config/nds32/pipelines.md: Include n15 settings.
14394 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
14395             Chung-Ju Wu  <jasonwucj@gmail.com>
14397         * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
14398         * config/nds32/nds32-n13.md: New file.
14399         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
14400         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
14401         pipeline.
14402         * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
14403         * config/nds32/nds32.md (pipeline_model): Add n13.
14404         * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
14405         * config/nds32/pipelines.md: Include n13 settings.
14407 2018-05-13  Kito Cheng  <kito.cheng@gmail.com>
14408             Chung-Ju Wu  <jasonwucj@gmail.com>
14410         * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
14411         * config/nds32/nds32-n10.md: New file.
14412         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
14413         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
14414         pipeline.
14415         * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
14416         * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
14417         * config/nds32/nds32.md (pipeline_model): Add n10.
14418         * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
14419         * config/nds32/pipelines.md: Include n10 settings.
14421 2018-05-13  Monk Chiang  <sh.chiang04@gmail.com>
14422             Kito Cheng  <kito.cheng@gmail.com>
14423             Chung-Ju Wu  <jasonwucj@gmail.com>
14425         * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
14426         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
14427         Add enum values for DSP extension instructions.
14428         * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
14429         New constraints.
14430         * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
14431         sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
14432         New code iterators.
14433         (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
14434         * config/nds32/nds32-dspext.md: New file for DSP implementation.
14435         * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
14436         * config/nds32/nds32-intrinsic.md: Likewise.
14437         * config/nds32/nds32_intrinsic.h: Likewise.
14438         * config/nds32/nds32-md-auxiliary.c: Likewise.
14439         * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
14440         * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
14441         (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
14442         (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
14443         * config/nds32/nds32-protos.h: New declarations for DSP extension.
14444         * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
14445         TYPE_DMAC in switch statement.
14446         * config/nds32/nds32.c: New checking and implementation for DSP
14447         extension instructions.
14448         * config/nds32/nds32.h: Likewise.
14449         * config/nds32/nds32.md: Likewise.
14450         * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
14451         * config/nds32/predicates.md: Implement new predicates for DSP
14452         extension.
14454 2018-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
14456         * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
14457         Reformat alternatives and attributes so it is easier to identify
14458         which constraints/attributes go with which instruction.
14459         (mov<mode>_hardfloat32, FMOVE64): Likewise.
14460         (mov<mode>_softfloat32, FMOVE64): Likewise.
14461         (mov<mode>_hardfloat64, FMOVE64): Likewise.
14462         (mov<mode>_softfloat64, FMOVE64): Likewise.
14464 2018-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14466         * doc/extend.texi (PowerPC Built-in Functions): Rename this
14467         subsection.
14468         (Basic PowerPC Built-in Functions): The new name of the
14469         subsection previously known as "PowerPC Built-in Functions".
14470         (Basic PowerPC Built-in Functions Available on all Configurations):
14471         New subsubsection.
14472         (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
14473         (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
14474         (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
14475         (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
14477 2018-05-11  Martin Jambor  <mjambor@suse.cz>
14479         PR ipa/85655
14480         * ipa-cp.c (intersect_with_plats): Check that the lattice contains
14481         single const.
14483 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
14485         PR target/85733
14486         * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
14488 2018-05-11  Sebastian Peryt  <sebastian.peryt@intel.com>
14490         * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
14491         OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
14492         (ix86_handle_option): Handle -mwaitpkg.
14493         * config.gcc: New header.
14494         * config/i386/cpuid.h (bit_WAITPKG): New bit.
14495         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
14496         * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
14497         function type.
14498         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14499         OPTION_MASK_ISA_WAITPKG.
14500         * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
14501         (ix86_option_override_internal): Add PTA_WAITPKG.
14502         (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
14503         (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
14504         IX86_BUILTIN_TPAUSE.
14505         (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
14506         __builtin_ia32_umwait and __builtin_ia32_tpause.
14507         (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
14508         IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
14509         * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
14510         * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
14511         UNSPECV_TPAUSE): New.
14512         (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
14513         * config/i386/i386.opt: Add -mwaitpkg.
14514         * config/i386/waitpkgintrin.h: New file.
14515         * config/i386/x86intrin.h: New header.
14516         * doc/invoke.texi: Add -mwaitpkg.
14518 2018-05-11  Richard Earnshaw  <rearnsha@arm.com>
14520         PR target/85606
14521         * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
14522         equivalent.
14523         (cortex-m0): Use armv6s-m isa.
14524         (cortex-m0plus): Likewise.
14525         (cortex-m1): Likewise.
14526         (cortex-m0.small-multiply): Likewise.
14527         (cortex-m0plus.small-multiply): Likewise.
14528         (cortex-m1.small-multiply): Likewise.
14530 2018-05-11  Allan Sandfeld Jensen  <allan.jensen@qt.io>
14531             Jakub Jelinek  <jakub@redhat.com>
14533         PR tree-optimization/85692
14534         * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
14535         source permute as well.
14537 2018-05-11  Martin Liska  <mliska@suse.cz>
14539         PR sanitizer/85556
14540         * doc/extend.texi: Document LLVM style format for no_sanitize
14541         attribute.
14543 2018-05-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
14545         * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
14546         mode_supports_vsx_dform_quad to mode_supports_dq_form.
14547         (mode_supports_vsx_dform_quad): Likewise.
14548         (mode_supports_vmx_dform): Move these functions to be next to the
14549         other mode_supports functions.
14550         (mode_supports_dq_form): Likewise.
14551         (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
14552         mode_supports_dq_form.
14553         (reg_offset_addressing_ok_p): Likewise.
14554         (offsettable_ok_by_alignment): Likewise.
14555         (rs6000_legitimate_offset_address_p): Likewise.
14556         (legitimate_lo_sum_address_p): Likewise.
14557         (rs6000_legitimize_address): Likewise.
14558         (rs6000_legitimize_reload_address): Likewise.
14559         (rs6000_secondary_reload_inner): Likewise.
14560         (rs6000_preferred_reload_class): Likewise.
14561         (rs6000_output_move_128bit): Likewise.
14563 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
14565         * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
14566         Generate SImode target register for null target.
14567         <case IX86_BUILTIN_XGETBV>: Ditto.
14568         <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
14569         * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
14571 2018-05-10  Carl Love  <cel@us.ibm.com>
14573         * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
14574         dcbtt and dcbtstt if operands[2] is 0.
14576 2018-05-10  Uros Bizjak  <ubizjak@gmail.com>
14578         PR target/85693
14579         * config/i386/sse.md (usadv64qi): New expander.
14581 2018-05-10  Segher Boessenkool  <segher@kernel.crashing.org>
14583         * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
14584         altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
14585         -maltivec=be support.
14586         (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
14587         vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
14588         vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
14589         vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
14590         vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
14591         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
14592         altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
14593         altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
14594         altivec_vsumsws): Adjust.
14595         (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
14596         *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
14597         altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
14598         support.
14599         (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
14600         altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
14601         altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
14602         (altivec_lve<VI_char>x): Delete expand.
14603         (*altivec_lve<VI_char>x_internal): Rename to...
14604         (altivec_lve<VI_char>x): ... this.
14605         (altivec_lvxl_<mode>): Delete expand.
14606         (*altivec_lvxl_<mode>_internal): Rename to ...
14607         (altivec_lvxl_<mode>): ... this.
14608         (altivec_stvxl_<mode>): Delete expand.
14609         (*altivec_stvxl_<mode>_internal): Rename to ...
14610         (altivec_stvxl_<mode>): ... this.
14611         (altivec_stve<VI_char>x): Delete expand.
14612         (*altivec_stve<VI_char>x_internal): Rename to ...
14613         (altivec_stve<VI_char>x): ... this.
14614         (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
14615         doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
14616         reduc_plus_scal_<mode>): Adjust.
14617         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
14618         comment.
14619         (rs6000_cpu_cpp_builtins): Adjust.
14620         (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
14621         * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
14622         altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
14623         * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
14624         -maltivec=be support.
14625         (rs6000_split_vec_extract_var): Adjust.
14626         (rs6000_split_v4si_init): Adjust.
14627         (swap_selector_for_mode): Delete.
14628         (altivec_expand_lvx_be, altivec_expand_stvx_be,
14629         altivec_expand_stvex_be): Delete.
14630         (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
14631         -maltivec=be support.
14632         (rs6000_gimple_fold_builtin): Ditto.
14633         (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
14634         Adjust.
14635         * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
14636         (TARGET_DIRECT_MOVE_64BIT): Adjust.
14637         * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
14638         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
14639         * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
14640         unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
14641         vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
14642         *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
14643         *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
14644         *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
14645         anonymous split): Adjust.
14646         (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
14647         (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
14649 2018-05-10  Eric Botcazou  <ebotcazou@adacore.com>
14651         * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
14652         when --with-gxx-include-dir is also specified.
14653         * configure: Regenerate.
14655 2018-05-09  Jim Wilson  <jimw@sifive.com>
14657         PR target/84797
14658         * config.gcc (riscv*-*-*): Handle --with-multilib-list.
14659         * config/riscv/t-withmultilib: New.
14660         * config/riscv/withmultilib.h: New.
14661         * doc/install.texi: Document RISC-V --with-multilib-list support.
14663 2018-05-09  Richard Biener  <rguenther@suse.de>
14665         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
14666         vector.
14667         (vect_bb_vectorization_profitable_p): Adjust.  Compute
14668         actual scalar cost using the cost vector and the add_stmt_cost
14669         machinery.
14671 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
14673         PR rtl-optimization/85645
14674         * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
14675         in the REG_CFA_REGISTER note for LR, don't leave it empty.
14677 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
14679         PR rtl-optimization/85645
14680         * shrink-wrap.c (spread_components): Return a boolean saying if
14681         anything was changed.
14682         (try_shrink_wrapping_separate): Iterate spread_components until
14683         nothing changes anymore.
14685 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
14687         PR rtl-optimization/85645
14688         * regrename.c (build_def_use): Also kill the chains that include the
14689         destination of a REG_CFA_REGISTER note.
14691 2018-05-09  Segher Boessenkool  <segher@kernel.crashing.org>
14693         PR rtl-optimization/85645
14694         *  regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
14695         insn that has a REG_CFA_REGISTER note.
14697 2018-05-09  Richard Sandiford  <richard.sandiford@linaro.org>
14699         * cfgexpand.c (expand_clobber): New function.
14700         (expand_gimple_stmt_1): Use it.
14701         * tree-vect-stmts.c (vect_clobber_variable): New function,
14702         split out from...
14703         (vectorizable_simd_clone_call): ...here.
14704         (vectorizable_store): Emit a clobber either side of an
14705         IFN_STORE_LANES sequence.
14706         (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
14708 2018-05-09  Tom de Vries  <tom@codesourcery.com>
14710         PR target/85626
14711         * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
14712         (define_insn "trap_if_false"): Add exit after trap.
14714 2018-05-09  Eric Botcazou  <ebotcazou@adacore.com>
14716         PR rtl-optimization/85638
14717         * bb-reorder.c: Include common/common-target.h.
14718         (create_forwarder_block): New function extracted from...
14719         (fix_up_crossing_landing_pad): ...here.  Rename into...
14720         (dw2_fix_up_crossing_landing_pad): ...this.
14721         (sjlj_fix_up_crossing_landing_pad): New function.
14722         (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
14723         call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
14724         from both partitions and exit the loop after one iteration.
14726 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14728         Revert:
14729         * doc/extend.texi (PowerPC Built-in Functions): Rename this
14730         subsection.
14731         (Basic PowerPC Built-in Functions): The new name of the
14732         subsection previously known as "PowerPC Built-in Functions".
14733         (Basic PowerPC Built-in Functions Available on all Configurations):
14734         New subsubsection.
14735         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
14736         subsubsection.
14737         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
14738         subsubsection.
14739         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
14740         subsubsection.
14741         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
14742         subsubsection.
14744 2018-05-08  Jim Wilson  <jimw@sifive.com>
14746         * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
14747         (LD_EMUL_SUFFIX): New.
14748         (LINK_SPEC): Use it.
14750 2018-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14752         * doc/extend.texi (PowerPC Built-in Functions): Rename this
14753         subsection.
14754         (Basic PowerPC Built-in Functions): The new name of the
14755         subsection previously known as "PowerPC Built-in Functions".
14756         (Basic PowerPC Built-in Functions Available on all Configurations):
14757         New subsubsection.
14758         (Basic PowerPC Built-in Functions Available on ISA 2.05): New
14759         subsubsection.
14760         (Basic PowerPC Built-in Functions Available on ISA 2.06): New
14761         subsubsection.
14762         (Basic PowerPC Built-in Functions Available on ISA 2.07): New
14763         subsubsection.
14764         (Basic PowerPC Built-in Functions Available on ISA 3.0): New
14765         subsubsection.
14767 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
14769         PR target/85683
14770         * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
14771         after cmpelim optimization.
14773 2018-05-08  Olga Makhotina  <olga.makhotina@intel.com>
14775         * config.gcc: Support "goldmont".
14776         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
14777         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14778         PROCESSOR_GOLDMONT.
14779         * config/i386/i386.c (m_GOLDMONT): Define.
14780         (processor_target_table): Add "goldmont".
14781         (PTA_GOLDMONT): Define.
14782         (ix86_lea_outperforms): Add TARGET_GOLDMONT.
14783         (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
14784         (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
14785         (fold_builtin_cpu): Add "goldmont".
14786         (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
14787         (ix86_option_override_internal): Add "goldmont".
14788         * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
14789         (processor_type): Add PROCESSOR_GOLDMONT.
14790         * config/i386/i386.md: Add CPU "glm".
14791         * config/i386/glm.md: New file.
14792         * config/i386/x86-tune.def: Add m_GOLDMONT.
14793         * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
14795 2018-05-08  Jakub Jelinek  <jakub@redhat.com>
14797         PR target/85572
14798         * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
14799         E_V4DImode.
14800         * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
14801         VI1248_AVX512VL_AVX512BW.  Handle V2DImode and V4DImode if not
14802         TARGET_AVX512VL using ix86_expand_sse2_abs.  Formatting fixes.
14804         PR target/85317
14805         * config/i386/i386.c (ix86_fold_builtin): Handle
14806         IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
14808         PR target/85480
14809         * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
14810         (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
14812 2018-05-08  Richard Earnshaw  <rearnsha@arm.com>
14814         PR target/85658
14815         * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
14816         (check_arch): Likewise.
14817         (check_fpu): Return the result rather than printing it.
14818         (end arch): Fix operator precedence.
14819         (end cpu): Likewise.
14820         (END): Print the result from check_fpu.
14822 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
14823             Alan Hayward  <alan.hayward@arm.com>
14824             David Sherwood  <david.sherwood@arm.com>
14826         * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
14827         (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
14828         (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
14829         (*fcmuo<mode>_and): New patterns.
14831 2018-05-08  Richard Sandiford  <richard.sandiford@linaro.org>
14833         * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
14834         (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
14835         (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
14836         (cmp_op, sve_imm_con): New code attributes.
14837         (SVE_COND_INT_CMP, imm_con): Delete.
14838         (cmp_op): Remove above unspecs from int attribute.
14839         * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
14840         to...
14841         (*cmp<cmp_op><mode>): ...this.  Use UNSPEC_MERGE_PTRUE instead of
14842         comparison-specific unspecs.
14843         (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
14844         (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
14845         (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
14846         (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
14847         (*vec_fcm<cmp_op><mode>): Rename to...
14848         (*fcm<cmp_op><mode>): ...this and adjust likewise.
14849         (*vec_fcmuo<mode>): Rename to...
14850         (*fcmuo<mode>): ...this and adjust likewise.
14851         (*pred_fcm<cmp_op><mode>): New pattern.
14852         * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
14853         (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
14854         functions.
14855         (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
14856         and UNORDERED.
14857         (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
14858         (aarch64_emit_sve_predicated_cond): New function.
14859         (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
14860         (aarch64_emit_unspec_cond_or): Replace with...
14861         (aarch64_emit_sve_or_conds): ...this new function.  Use
14862         aarch64_emit_sve_ptrue_op for the individual comparisons and
14863         aarch64_emit_binop to OR them together.
14864         (aarch64_emit_inverted_unspec_cond): Replace with...
14865         (aarch64_emit_sve_inverted_cond): ...this new function.  Use
14866         aarch64_emit_sve_ptrue_op for the comparison and
14867         aarch64_emit_unop to invert the result.
14868         (aarch64_expand_sve_vec_cmp_float): Update after the above
14869         changes.  Use aarch64_emit_sve_ptrue_op for native comparisons.
14871 2018-05-07  Nathan Sidwell  <nathan@acm.org>
14873         * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
14874         * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
14875         (Backwards Compatibility): Likewise.
14877 2018-05-07  Luis Machado  <luis.machado@linaro.org>
14879         PR bootstrap/85681
14880         Revert:
14881         2018-05-07  Luis Machado  <luis.machado@linaro.org>
14883         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14884         <prefetch_dynamic_strides>: New const bool field.
14885         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14886         prefetch_dynamic_strides.
14887         (exynosm1_prefetch_tune): Likewise.
14888         (thunderxt88_prefetch_tune): Likewise.
14889         (thunderx_prefetch_tune): Likewise.
14890         (thunderx2t99_prefetch_tune): Likewise.
14891         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
14892         to false.
14893         (aarch64_override_options_internal): Update to set
14894         PARAM_PREFETCH_DYNAMIC_STRIDES.
14895         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14896         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14897         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14898         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14899         prefetch-dynamic-strides setting.
14901         2018-05-07  Luis Machado  <luis.machado@linaro.org>
14903         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14904         <minimum_stride>: New const int field.
14905         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14906         minimum_stride field.
14907         (exynosm1_prefetch_tune): Likewise.
14908         (thunderxt88_prefetch_tune): Likewise.
14909         (thunderx_prefetch_tune): Likewise.
14910         (thunderx2t99_prefetch_tune): Likewise.
14911         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
14912         (aarch64_override_options_internal): Update to set
14913         PARAM_PREFETCH_MINIMUM_STRIDE.
14914         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14915         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14916         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14917         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14918         stride is constant and is below the minimum stride threshold.
14920 2018-05-07  Luis Machado  <luis.machado@linaro.org>
14922         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
14923         to 512.
14925 2018-05-07  Luis Machado  <luis.machado@linaro.org>
14927         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14928         <prefetch_dynamic_strides>: New const bool field.
14929         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14930         prefetch_dynamic_strides.
14931         (exynosm1_prefetch_tune): Likewise.
14932         (thunderxt88_prefetch_tune): Likewise.
14933         (thunderx_prefetch_tune): Likewise.
14934         (thunderx2t99_prefetch_tune): Likewise.
14935         (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
14936         to false.
14937         (aarch64_override_options_internal): Update to set
14938         PARAM_PREFETCH_DYNAMIC_STRIDES.
14939         * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
14940         * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
14941         * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
14942         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
14943         prefetch-dynamic-strides setting.
14945 2018-05-07  Luis Machado  <luis.machado@linaro.org>
14947         * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
14948         <minimum_stride>: New const int field.
14949         * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
14950         minimum_stride field.
14951         (exynosm1_prefetch_tune): Likewise.
14952         (thunderxt88_prefetch_tune): Likewise.
14953         (thunderx_prefetch_tune): Likewise.
14954         (thunderx2t99_prefetch_tune): Likewise.
14955         (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
14956         (aarch64_override_options_internal): Update to set
14957         PARAM_PREFETCH_MINIMUM_STRIDE.
14958         * doc/invoke.texi (prefetch-minimum-stride): Document new option.
14959         * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
14960         * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
14961         * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
14962         stride is constant and is below the minimum stride threshold.
14964 2018-05-06  Jakub Jelinek  <jakub@redhat.com>
14966         PR c++/85659
14967         * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
14968         the type is addressable.  Don't force op into register if it has
14969         BLKmode.
14971 2018-05-05  Roland McGrath  <mcgrathr@google.com>
14973         PR other/77609
14974         * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
14975         any section for which we don't know a specific type it should have,
14976         regardless of name.  Previously this was done only for the exact
14977         names ".init_array", ".fini_array", and ".preinit_array".
14978         (default_elf_asm_named_section): Add comment about
14979         relationship with default_section_type_flags and SECTION_NOTYPE.
14980         (get_section): Don't consider it a type conflict if one side has
14981         SECTION_NOTYPE and the other doesn't, as long as neither has the
14982         SECTION_BSS et al used in the default_section_type_flags logic.
14984 2018-05-05  Tom de Vries  <tom@codesourcery.com>
14986         PR target/85653
14987         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
14988         (workaround_barsyncs): New function.
14989         (nvptx_reorg): Use workaround_barsyncs.
14990         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
14991         (define_expand "nvptx_membar_cta"): New define_expand.
14992         (define_insn "*nvptx_membar_cta"): New insn.
14994 2018-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
14996         * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
14997         To improve optimization opportunities.
14998         * builtin-types.def: The new needed builtin types for the above.
15000 2018-05-04  Richard Biener  <rguenther@suse.de>
15002         * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
15003         * gimple-ssa-store-merging.c
15004         (imm_store_chain_info::output_merged_store): Remove redundant create,
15005         release split_store vector contents on failure.
15006         * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
15007         scalar stmt vector on cache hit.
15009 2018-05-04  Segher Boessenkool  <segher@kernel.crashing.org>
15011         * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
15012         Xilinx FP support.
15013         * config.gcc (powerpc-xilinx-eabi*): Remove.
15014         * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
15015         support.
15016         (fusion_addis_mem_combo_load): Ditto.
15017         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
15018         FP support.
15019         (rs6000_cpu_cpp_builtins): Ditto.
15020         * config/rs6000/rs6000-linux.c
15021         (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
15022         * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
15023         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
15024         support.
15025         (rs6000_setup_reg_addr_masks): Ditto.
15026         (rs6000_init_hard_regno_mode_ok): Ditto.
15027         (rs6000_option_override_internal): Ditto.
15028         (legitimate_lo_sum_address_p): Ditto.
15029         (rs6000_legitimize_address): Ditto.
15030         (rs6000_legitimize_reload_address): Ditto.
15031         (rs6000_legitimate_address_p): Ditto.
15032         (abi_v4_pass_in_fpr): Ditto.
15033         (setup_incoming_varargs): Ditto.
15034         (rs6000_gimplify_va_arg): Ditto.
15035         (rs6000_split_multireg_move): Ditto.
15036         (rs6000_savres_strategy): Ditto.
15037         (rs6000_emit_prologue_components): Ditto.
15038         (rs6000_emit_epilogue_components): Ditto.
15039         (rs6000_emit_prologue): Ditto.
15040         (rs6000_emit_epilogue): Ditto.
15041         (rs6000_elf_file_end): Ditto.
15042         (rs6000_function_value): Ditto.
15043         (rs6000_libcall_value): Ditto.
15044         * config/rs6000/rs6000.h: Ditto.
15045         (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
15046         (TARGET_MINMAX): ... this.  New.
15047         (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
15048         * config/rs6000/rs6000.md: Remove Xilinx FP support.
15049         (*movsi_internal1_single): Delete.
15050         * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
15051         mfpu=, mxilinx-fpu): Delete.
15052         * config/rs6000/singlefp.h: Delete.
15053         * config/rs6000/sysv4.h: Remove Xilinx FP support.
15054         * config/rs6000/t-rs6000: Ditto.
15055         * config/rs6000/t-xilinx: Delete.
15056         * config/rs6000/titan.md: Adjust for fp_type removal.
15057         * config/rs6000/vsx.md: Remove Xilinx FP support.
15058         (VStype_simple): Delete.
15059         (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
15060         * config/rs6000/xfpu.h: Delete.
15061         * config/rs6000/xfpu.md: Delete.
15062         * config/rs6000/xilinx.h: Delete.
15063         * config/rs6000/xilinx.opt: Delete.
15064         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove
15065         -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
15067 2018-05-04  Tom de Vries  <tom@codesourcery.com>
15069         PR libgomp/85639
15070         * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
15071         if ignore == 0.
15073 2018-05-04  Richard Biener  <rguenther@suse.de>
15075         PR middle-end/85627
15076         * tree-complex.c (update_complex_assignment): We are always in SSA form.
15077         (expand_complex_div_wide): Likewise.
15078         (expand_complex_operations_1): Likewise.
15079         (expand_complex_libcall): Preserve EH info of the original stmt.
15080         (tree_lower_complex): Handle removed blocks.
15081         * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
15082         on complex multiplication and division libcall builtins.
15084 2018-05-04  Richard Biener  <rguenther@suse.de>
15086         PR middle-end/85574
15087         * fold-const.c (negate_expr_p): Restrict negation of operand
15088         zero of a division to when we know that can happen without
15089         overflow.
15090         (fold_negate_expr_1): Likewise.
15092 2018-05-04  Jakub Jelinek  <jakub@redhat.com>
15094         PR libstdc++/85466
15095         * real.h (real_nextafter): Declare.
15096         * real.c (real_nextafter): New function.
15097         * fold-const-call.c (fold_const_nextafter): New function.
15098         (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
15099         CASE_CFN_NEXTTOWARD.
15100         (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
15101         even when arg1_mode is different from arg0_mode.
15103 2018-05-03  Nathan Sidwell  <nathan@acm.org>
15105         * doc/extend.texi (Deprecated Features): Remove
15106         -ffriend-injection.
15107         (Backwards Compatibility): Likewise.
15108         * doc/invoke.texi (C++ Language Options): Likewise.
15109         (C++ Dialect Options): Likewise.
15111 2018-05-03  Jakub Jelinek  <jakub@redhat.com>
15113         PR target/85530
15114         * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
15115         _mm512_mask_mullox_epi64): New intrinsics.
15117 2018-05-03  Tom de Vries  <tom@codesourcery.com>
15119         PR testsuite/85106
15120         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15121         dump files): Add offload-tree.
15123 2018-05-03  Richard Biener  <rguenther@suse.de>
15125         PR tree-optimization/85615
15126         * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
15127         to loops not nested in BBs loop father to avoid creating multi-entry
15128         loops.
15130 2018-05-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15132         PR tree-optimization/70291
15133         * tree-complex.c (expand_complex_libcall): Add type, inplace_p
15134         arguments.  Change return type to tree.  Emit libcall as a new
15135         statement rather than replacing existing one when inplace_p is true.
15136         (expand_complex_multiplication_components): New function.
15137         (expand_complex_multiplication): Expand floating-point complex
15138         multiplication using the above.
15139         (expand_complex_division): Rename inner_type parameter to type.
15140         Update expand_complex_libcall call-site.
15141         (expand_complex_operations_1): Update expand_complex_multiplication
15142         and expand_complex_division call-sites.
15144 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
15146         PR target/85582
15147         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
15148         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
15149         *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
15150         the highest significant bit of the shift count mask is clear.  In
15151         check whether and[sq]i3 is needed verify that all significant bits
15152         of the shift count other than the highest are set.
15154 2018-05-02  Tom de Vries  <tom@codesourcery.com>
15156         PR libgomp/82428
15157         * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
15158         * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
15159         (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
15160         * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
15161         (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
15162         * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
15163         __builtin_goacc_parlevel_size.
15165 2018-05-02  Richard Biener  <rguenther@suse.de>
15167         PR tree-optimization/85597
15168         * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
15169         do not use split vect_get_vec_defs call but call vect_get_slp_defs
15170         directly.
15172 2018-05-02  Tom de Vries  <tom@codesourcery.com>
15174         PR testsuite/85106
15175         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15176         dump files): Add ltrans-tree.
15178 2018-05-02  Tom de Vries  <tom@codesourcery.com>
15180         PR testsuite/85106
15181         * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
15182         dump files): Add wpa-ipa.
15184 2018-05-02  Segher Boessenkool  <segher@kernel.crashing.org>
15186         * config.gcc (powerpc*-*-*): Remove paired.h.  Unsupport the
15187         powerpc*-*-linux*paired* target.
15188         * config/rs6000/750cl.h: Delete.
15189         * config/rs6000/paired.h: Delete.
15190         * config/rs6000/paired.md: Delete.
15191         * config/rs6000/predicates.md (easy_vector_constant): Remove paired
15192         float support.
15193         * config/rs6000/rs6000-builtin.def: Remove paired float support.
15194         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
15195         comment.  Remove paired float support.
15196         * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
15197         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
15198         VECTOR_PAIRED.
15199         * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
15200         paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
15201         declarations.
15202         * config/rs6000/rs6000.c: Remove paired float support.
15203         (paired_expand_vector_init, paired_expand_vector_move,
15204         paired_emit_vector_compare, paired_emit_vector_cond_expr,
15205         (paired_expand_lv_builtin, paired_expand_stv_builtin,
15206         paired_expand_builtin, paired_expand_predicate_builtin,
15207         paired_init_builtins): Delete.
15208         * config/rs6000/rs6000.h: Remove paired float support.
15209         * config/rs6000/rs6000.md: Remove paired float support.
15210         (move_from_CR_ov_bit): Delete.
15211         * config/rs6000/rs6000.opt (mpaired): Delete.
15212         * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
15213         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
15215 2018-05-02  Richard Biener  <rguenther@suse.de>
15217         PR middle-end/85567
15218         * gimplify.c (gimplify_save_expr): When in SSA form allow
15219         SAVE_EXPRs to compute to SSA vars.
15221 2018-05-02  Jakub Jelinek  <jakub@redhat.com>
15223         PR target/85582
15224         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
15225         *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
15226         *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
15227         clobber operands[2], instead use a new pseudo.  Formatting fixes.
15229 2018-05-02  Richard Sandiford  <richard.sandiford@linaro.org>
15231         PR tree-optimization/85586
15232         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
15233         exit early for statements in the same group if the accesses are
15234         not strided.
15236 2018-05-02  Tom de Vries  <tom@codesourcery.com>
15238         PR lto/85451
15239         * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
15240         error message.
15242 2018-05-01  Marc Glisse  <marc.glisse@inria.fr>
15244         PR tree-optimization/85143
15245         * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
15247 2018-05-01  Tom de Vries  <tom@codesourcery.com>
15249         PR lto/85451
15250         * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
15251         not found" error message.
15253 2018-05-01  Tom de Vries  <tom@codesourcery.com>
15255         PR other/83786
15256         * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
15257         * vec.c (test_ordered_remove_if): New function.
15258         (vec_c_tests): Call test_ordered_remove_if.
15259         * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
15260         * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
15261         * tree-vect-patterns.c (vect_pattern_recog_1): Use
15262         VEC_ORDERED_REMOVE_IF.
15264 2018-05-01  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
15266         PR tree-optimization/82665
15267         * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
15268         pointer subtraction where arguments come from a memchr call.
15270 2018-05-01  Jakub Jelinek  <jakub@redhat.com>
15272         * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
15273         --push-state --as-needed and --pop-state instead of --as-needed and
15274         --no-as-needed if ld supports it.
15275         * configure: Regenerated.
15277         PR web/85578
15278         * doc/install.texi2html: Replace _002d with - and _002a with * in
15279         generated html files using sed.
15281 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
15283         PR c++/85523
15284         * gcc-rich-location.c (blank_line_before_p): New function.
15285         (use_new_line): New function.
15286         (gcc_rich_location::add_fixit_insert_formatted): New function.
15287         * gcc-rich-location.h
15288         (gcc_rich_location::add_fixit_insert_formatted): New function.
15290 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
15292         * selftest.c (assert_streq): Rename "expected" and "actual" to
15293         "val1" and "val2".  Extend NULL-handling to cover both inputs
15294         symmetrically, while still requiring both to be non-NULL for a pass.
15295         * selftest.h (assert_streq): Rename "expected" and "actual" to
15296         "val1" and "val2".
15297         (ASSERT_EQ): Likewise.
15298         (ASSERT_EQ_AT): Likewise.
15299         (ASSERT_KNOWN_EQ): Likewise.
15300         (ASSERT_KNOWN_EQ_AT): Likewise.
15301         (ASSERT_NE): Likewise.
15302         (ASSERT_MAYBE_NE): Likewise.
15303         (ASSERT_MAYBE_NE_AT): Likewise.
15304         (ASSERT_STREQ): Likewise.  Clarify that both must be non-NULL for
15305         the assertion to pass.
15306         (ASSERT_STREQ_AT): Likewise.
15308 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
15310         * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
15311         interaction with -pie.
15313 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
15315         * selftest.h: Fix alphabetization of per-source-file selftest
15316         declarations.
15318 2018-04-30  Jason Merrill  <jason@redhat.com>
15320         PR c++/61982 - dead stores to destroyed objects.
15321         * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
15322         of clobber.
15324 2018-04-30  Jason Merrill  <jason@redhat.com>
15326         * tree.c (build_clobber): New.
15327         * tree.h: Declare it.
15328         * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
15330 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
15332         * diagnostic-show-locus.c (layout::layout): Update for
15333         location_get_source_line returning a char_span.
15334         (struct char_span): Move to input.h.
15335         (struct correction): Update for fields in char_span becoming
15336         private.
15337         (struct source_line): Update for location_get_source_line
15338         returning a char_span.
15339         (layout::print_line): Likewise.
15340         * edit-context.c (edited_file::print_content): Likewise.
15341         (edited_file::print_diff_hunk): Likewise.
15342         (edited_file::print_run_of_changed_lines): Likewise.
15343         (edited_file::get_num_lines): Likewise.
15344         (edited_line::edited_line): Likewise.
15345         * final.c (asm_show_source): Likewise.
15346         * input.c (location_get_source_line): Convert return type
15347         from const char * to char_span, losing the final "line_len"
15348         param.
15349         (dump_location_info): Update for the above.
15350         (get_substring_ranges_for_loc): Likewise.  Use a char_span
15351         when handling the literal within the line.
15352         (test_reading_source_line): Update for location_get_source_line
15353         returning a char_span.
15354         * input.h (class char_span): Move here from
15355         diagnostic-show-locus.c, converting from a struct to a class.
15356         Make data members private.
15357         (char_span::operator bool): New.
15358         (char_span::length): New.
15359         (char_span::get_buffer): New.
15360         (char_span::operator[]): New.
15361         (char_span::subspan): Make const.
15362         (char_span::xstrdup): New.
15363         (location_get_source_line): Convert return type from const char *
15364         to char_span, losing the final "line_size" param.
15366 2018-04-30  Jan Hubicka  <jh@suse.cz>
15368         * lto-wrapper.c (ltrans_priorities): New static var.
15369         (cmp_priority): New.
15370         (run_gcc): Read priorities and if doing parallel build order
15371         the Makefile by them.
15373 2018-04-30  David Malcolm  <dmalcolm@redhat.com>
15375         * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
15377 2018-04-30  Richard Biener  <rguenther@suse.de>
15379         * tree-cfg.c (verify_address): Remove base argument, add
15380         flag whether to check TREE_ADDRESSABLE and do that.
15381         (verify_expr): Remove.
15382         (verify_types_in_gimple_reference): Add pieces from verify_expr.
15383         (verify_gimple_assign_single): Likewise.
15384         (verify_gimple_switch): Likewise.
15385         (verify_expr_location_1): Dereference tp once.  Add (disabled)
15386         piece from verify_expr.
15387         (verify_gimple_in_cfg): Do not call verify_expr on all ops.
15389 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
15391         * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
15393 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
15395         * config/arc/arc-protos.h (prepare_extend_operands): Remove.
15396         (small_data_pattern): Likewise.
15397         (arc_rewrite_small_data): Likewise.
15398         * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
15399         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
15400         (get_symbol_alignment): New function.
15401         (legitimate_small_data_address_p): Likewise.
15402         (legitimate_scaled_address): Update, call
15403         legitimate_small_data_address_p.
15404         (output_sdata): New static variable.
15405         (arc_print_operand): Update how we handle small data operands.
15406         (arc_print_operand_address): Likewise.
15407         (arc_legitimate_address_p): Update, use
15408         legitimate_small_data_address_p.
15409         (arc_rewrite_small_data_p): Remove.
15410         (arc_rewrite_small_data_1): Likewise.
15411         (arc_rewrite_small_data): Likewise.
15412         (small_data_pattern): Likewise.
15413         (compact_sda_memory_operand): Update to use
15414         legitimate_small_data_address_p and get_symbol_alignment.
15415         (prepare_move_operands): Don't rewite sdata pattern.
15416         (prepare_extend_operands): Remove.
15417         * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
15418         pattern.
15419         (zero_extendqisi2): Likewise.
15420         (zero_extendhisi2): Likewise.
15421         (extendqihi2): Likewise.
15422         (extendqisi2): Likewise.
15423         (extendhisi2): Likewise.
15424         (addsi3): Likewise.
15425         (subsi3): Likewise.
15426         (andsi3): Likewise.
15427         * config/arc/constraints.md (Usd): Change it to memory constraint.
15429 2018-04-30  Claudiu Zissulescu  <claziss@synopsys.com>
15431         * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
15432         as source of std instructions.
15433         * config/arc/arc.md (movsi_insn): Update pattern predicate to
15434         allow 6-bit constants as source for store instructions.
15435         (movdi_insn): Update instruction pattern to allow 6-bit constants
15436         as source for store instructions.
15438 2018-04-30  Jonathan Wakely  <jwakely@redhat.com>
15440         * doc/invoke.texi (-fdebug-types-section): Fix grammar.
15442 2018-04-30  Nathan Sidwell  <nathan@acm.org>
15443             Sandra Loosemore <sandra@codesourcery.com>
15445         * dumpfile.c (dump_open): Allow '-' for stdout.
15446         * doc/invoke.texi (Developer Options): Document dump filename
15447         determination early.  Document stdin/stdout selection.
15449 2018-04-30  Andrew Sadek  <andrew.sadek.se@gmail.com>
15451         Microblaze Target: PIC data text relative
15453         * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
15454         * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
15455         Add declaration.
15456         * config/microblaze/microblaze.h (microblaze_constant_address_p):
15457         CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
15458         * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
15459         New addressing mode for data-text relative position indepenedent code.
15460         (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
15461         'ADDRESS_SYMBOLIC_TXT_REL'.
15462         (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
15463         (microblaze_legitimate_pic_operand): Exclude function calls from
15464         pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
15465         (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
15466         addresses cases.
15467         (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
15468         (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
15469         (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
15470         for 'address + offset'.
15471         (microblaze_expand_prologue): Add new function prologue call for
15472         'r20' assignation.
15473         (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
15474         'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
15475         table in case of TARGET_PIC_DATA_TEXT_REL.
15476         (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
15477         * config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
15478         Add new macros 'UNSPEC_TEXT',
15479         'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
15480         + exclude function calls from 'UNSPEC_PLT' in case of data text
15481         relative mode.
15482         * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
15483         new target hook for generating address diff vector tables in case of
15484         flag_pic.
15485         * doc/tm.texi : Regenerate.
15486         * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
15487         'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
15488         of addr diff vector generation.
15489         * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
15490         target hook definition.
15491         * targhooks.h, targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
15492         Add default function for generate_pic_addr_diff_vec -> flag_pic.
15493         * doc/invoke.texi (Add new pic option): Add new microblaze pic
15494         option for data text relative.
15496 2018-04-30  Richard Biener  <rguenther@suse.de>
15498         * tree-chrec.h (evolution_function_is_constant_p): Remove
15499         redundant check.
15500         * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
15502 2018-04-30  Richard Biener  <rguenther@suse.de>
15504         PR bootstrap/85571
15505         * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
15507 2018-04-30  Richard Biener  <rguenther@suse.de>
15509         PR tree-optimization/28364
15510         PR tree-optimization/85275
15511         * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
15512         copying first exit test.
15514 2018-04-28  Mark Wielaard  <mark@klomp.org>
15516         * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
15517         dwarf_version >= 5.
15518         (dwarf_AT): Handle DW_AT_addr_base.
15519         (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
15521 2018-04-28  Uros Bizjak  <ubizjak@gmail.com>
15523         PR target/84431
15524         * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
15525         (*ashl<dwi>3_doubleword_mask_1): Ditto.
15526         (*<shift_insn><dwi>3_doubleword_mask): Ditto.
15527         (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
15529 2018-04-28  Richard Biener  <rguenther@suse.de>
15531         * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
15532         (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
15533         to reflect use.  Only add interesting stmts.
15535 2018-04-27  Martin Jambor  <mjambor@suse.cz>
15537         PR ipa/85549
15538         * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
15539         the jump function allows for passing through aggregate values.
15541 2018-04-27  David Malcolm  <dmalcolm@redhat.com>
15543         * input.h (in_system_header_at): Convert from macro to inline
15544         function.
15545         (from_macro_expansion_at): Likewise.
15546         (from_macro_definition_at): Likewise.
15548 2018-04-27  Jeff Law  <law@redhat.com>
15550         * config.gcc: Mark tile* targets as deprecated/obsolete.
15552 2018-04-27  Richard Biener  <rguenther@suse.de>
15554         * config/aarch64/aarch64.c: Simplify ap.__stack advance and
15555         fix for ILP32.
15557 2018-04-27  Richard Biener  <rguenther@suse.de>
15559         * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
15561 2018-04-27  Uros Bizjak  <ubizjak@gmail.com>
15563         * config/i386/i386.md (*movti_internal): Substitute Ye constraint
15564         with Yd constraint. Set "preferred_for_speed" attribute from
15565         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
15566         with Yd constraint.
15567         (*movdi_internal): Ditto.
15568         (movti_interunit splitters): Remove
15569         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
15570         (movdi_interunit splitters): Ditto.
15571         * config/i386/constraints.md (Ye): Remove.
15572         (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
15574 2018-04-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15576         PR target/85512
15577         * config/aarch64/constraints.md (Usg): Limit to 31.
15578         (Usj): Limit to 63.
15580 2018-04-27  Jakub Jelinek  <jakub@redhat.com>
15582         PR tree-optimization/85529
15583         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
15584         argument.  Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
15585         rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
15586         zero extension or masking of the MSB bit.
15587         (optimize_range_tests): Add FIRST_BB argument, pass it through
15588         to optimize_range_tests_var_bound.
15589         (maybe_optimize_range_tests, reassociate_bb): Adjust
15590         optimize_range_tests callers.
15592 2018-04-26  Richard Biener  <rguenther@suse.de>
15593             Jakub Jelinek  <jakub@redhat.com>
15595         * cgraph.h (symbol_table): Just declare debug method here.
15596         * symtab.c (symbol_table::debug): Define.
15598 2018-04-26  Eric Botcazou  <ebotcazou@adacore.com>
15600         * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
15602 2018-04-26  Uros Bizjak  <ubizjak@gmail.com>
15604         * config/i386/i386.md ("isa" attribute): Add x64_sse2.
15605         ("enabled" attribute): Handle x64_sse2 "isa" attribute.
15606         (*movdi_internal): Substitute Yi and Yj constraint with x
15607         and Ym and Yn constraint with y constraint.  Update "isa"
15608         attribute and set "preferred_for_speed" attribute from
15609         TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
15610         (*movsi_internal): Ditto.
15611         (*movdf_internal): Ditto.
15612         (*movsf_internal): Ditto.
15613         (*zero_extendsidi2): Ditto.
15614         * config/i386/sse.md (vec_set<mode>_0): Ditto.
15615         (sse2_loadld): Ditto.
15616         (*vec_extract<ssevecmodelower>_0): Ditto.
15617         (*vec_extractv4si_0_zext_sse4): Ditto.
15618         (vec_concatv2di): Ditto.
15619         (*vec_dup<mode>): Ditto.
15620         * config/i386/mmx.md (*mov<mode>_internal): Ditto.
15621         * config/i386/constraints.md (Yi): Remove.
15622         (Yj): Remove.
15623         (Ym): Remove.
15624         (Yn): Remove.
15626 2018-04-26  Nathan Sidwell  <nathan@acm.org>
15628         * dumpfile.c (dump_open): New.
15629         (dump_open_alternate_stream, dump_start, dump_begin): Call it.
15630         (dump_finish): Detect stdio/stderr by value not name.
15632 2018-04-26  Jonathan Wakely  <jwakely@redhat.com>
15634         * doc/invoke.texi (-Wreturn-type): Document default status for C++.
15636 2018-04-26  Tom de Vries  <tom@codesourcery.com>
15638         PR target/84952
15639         * config/nvptx/nvptx.c (verify_neutering_jumps)
15640         (verify_neutering_labels): New function
15641         (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
15643 2018-04-26  Tom de Vries  <tom@codesourcery.com>
15645         PR target/84025
15646         * config/nvptx/nvptx.c (needs_neutering_p): New function.
15647         (nvptx_single): Use needs_neutering_p to skip over insns that do not
15648         need neutering.
15650 2018-04-26  Richard Biener <rguenther@suse.de>
15651             Tom de Vries  <tom@codesourcery.com>
15653         PR lto/85422
15654         * lto-streamer-out.c (output_function): Fixup loops if required to match
15655         discovery done in the reader.
15657 2018-04-26  Richard Biener  <rguenther@suse.de>
15659         PR tree-optimization/85116
15660         * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
15661         have a loop exit from the single latch predecessor.  Remove
15662         case of header with just condition.
15663         (ch_base::copy_headers): Exclude infinite loops from any
15664         processing.
15665         (pass_ch::execute): Record exits.
15667 2018-04-26  Richard Biener  <rguenther@suse.de>
15669         * tree-vect-data-refs.c (vect_get_data_access_cost): Get
15670         prologue cost vector and pass it to vect_get_load_cost.
15671         (vect_get_peeling_costs_all_drs): Likewise.
15672         (vect_peeling_hash_get_lowest_cost): Likewise.
15673         (vect_enhance_data_refs_alignment): Likewise.
15675 2018-04-26  Richard Biener  <rguenther@suse.de>
15677         PR middle-end/85450
15678         * tree-cfg.c (verify_gimple_assign_unary): Restore proper
15679         checking of integer<->pointer conversions.
15680         * omp-expand.c (expand_omp_for_static_nochunk): Avoid
15681         sign-/zero-extending pointer types.
15682         (expand_omp_for_static_chunk): Likewise.
15684 2018-03-22  Hans-Peter Nilsson  <hp@axis.com>
15685             Jean Lee  <xiaoyur347@gmail.com>
15687         * config/mips/mips.c (mips_asan_shadow_offset): New function.
15688         (TARGET_ASAN_SHADOW_OFFSET): Define.
15689         * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
15690         true for -fsanitize=address.
15692 2018-04-25  Mark Wielaard  <mark@klomp.org>
15694         * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
15695         shorter ones.
15697 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
15699         * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
15700         than "alu", remove explicit "memory" and "imm_disp" attributes.
15701         (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
15703         PR middle-end/85414
15704         * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
15705         case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
15706         gen_lowpart_no_emit.
15708 2018-04-25  Sebastian Peryt  <sebastian.peryt@intel.com>
15710         PR target/85473
15711         * config/i386/i386.c (ix86_expand_builtin): Change memory
15712         operand to XI, extend p0 to Pmode.
15713         * config/i386/i386.md: Change unspec volatile and operand
15714         1 mode to XI, change operand 0 mode to P.
15716 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
15718         * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
15719         GET_MODE_MASK before any checking.
15720         (nds32_can_use_bset_p): Likewise.
15721         (nds32_can_use_btgl_p): Likewise.
15723 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
15725         * config/nds32/nds32-doubleword.md: New define_split pattern for
15726         illegal register number.
15728 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
15730         * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
15732 2018-04-25  Chung-Ju Wu  <jasonwucj@gmail.com>
15734         * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
15736 2018-04-25  Richard Biener  <rguenther@suse.de>
15738         * lto-streamer.h (LTO_major_version): Bump to 8.
15740 2018-04-25  Jakub Jelinek  <jakub@redhat.com>
15742         * BASE-VER: Set to 9.0.0.
15744 2018-04-24  Segher Boessenkool  <segher@kernel.crashing.org>
15746         * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
15747         in __abskf2 and __powikf2.
15749 2018-04-24  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
15751         PR target/85512
15752         * config/aarch64/constraints.md (Usg, Usj): New constraints.
15753         * config/aarch64/iterators.md (cmode_simd): New mode attribute.
15754         * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
15755         Use the above on operand 2.  Reindent.
15756         (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
15758 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
15760         PR target/85485
15761         * common/config/i386/i386-common.c (ix86_handle_option): Don't
15762         handle OPT_mcet.
15763         * config/i386/i386.opt (mcet): Removed.
15764         * doc/install.texi: Remove -mcet documentation.
15765         * doc/invoke.texi: Likewise.
15767 2018-04-24  H.J. Lu  <hongjiu.lu@intel.com>
15769         PR target/85485
15770         * doc/install.texi: Remove -mcet from bootstrap-cet.
15772 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
15774         PR target/85511
15775         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
15776         __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
15777         if TARGET_64BIT.
15779         PR target/85503
15780         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
15781         const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
15782         containing a CONST_VECTOR.
15784 2018-04-24  Cesar Philippidis  <cesar@codesourcery.com>
15786         * doc/install.texi: Update newlib dependency for nvptx.
15788 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
15790         PR target/85508
15791         * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
15792         instead of INTVAL when shifting x left.
15794 2018-04-24  Andreas Krebbel  <krebbel@linux.ibm.com>
15796         PR tree-optimization/85478
15797         * tree-vect-loop.c (vect_analyze_loop_2): Do not call
15798         vect_grouped_store_supported for single element vectors.
15800 2018-04-24  Richard Biener  <rguenther@suse.de>
15802         PR target/85491
15803         * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
15804         load cost increase to the case of non-constant step.
15806 2018-04-24  Jakub Jelinek  <jakub@redhat.com>
15808         PR target/84828
15809         * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
15810         destination if any_malformed_asm.
15812 2018-04-23  Eric Botcazou  <ebotcazou@adacore.com>
15814         PR middle-end/85496
15815         * expr.c (store_field): In the bitfield case, if the value comes from
15816         a function call and is returned in registers by means of a PARALLEL,
15817         do not change the mode of the temporary unless BLKmode and VOIDmode.
15819 2018-04-23  Andrey Belevantsev  <abel@ispras.ru>
15821         PR rtl-optimization/85423
15822         * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
15823         dependencies to debug insns when the previous insn is non-debug.
15825 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
15827         * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
15828         enums into a single definition.
15829         (fls): Fix predicates and printing.
15830         (seti): Likewise.
15832 2018-04-23  Claudiu Zissulescu  <claziss@synopsys.com>
15834         * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
15835         * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
15836         and short u6 immediate.
15837         (check_if_valid_sleep_operand): Remove.
15838         * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
15840 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
15842         * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
15843         flag_always_save_lp condition.
15844         * config/nds32/nds32.opt (malways-save-lp): New option.
15846 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
15848         * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
15849         * config/nds32/nds32.c (nds32_use_load_post_increment): New.
15850         * config/nds32/nds32.h
15851         (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
15852         (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
15854 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
15856         * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
15857         * config/nds32/nds32.c (nds32_ls_333_p): Remove.
15859 2018-04-22  Shiva Chen  <shiva0217@gmail.com>
15860             Chung-Ju Wu  <jasonwucj@gmail.com>
15862         * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
15863         Declare.
15864         * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
15865         * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
15867 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
15869         * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
15871 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
15873         * config/nds32/nds32-protos.h (nds32_data_alignment,
15874         nds32_local_alignment): Declare.
15875         * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
15876         nds32_local_alignment): New functions.
15877         (TARGET_CONSTANT_ALIGNMENT): Define.
15878         * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
15880 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
15882         * config/nds32/nds32.c
15883         (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
15884         (TARGET_MODES_TIEABLE_P): Likewise.
15886 2018-04-22  Chung-Ju Wu  <jasonwucj@gmail.com>
15888         * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
15889         level Ofast and Og.
15891 2018-04-22  Monk Chiang  <sh.chiang04@gmail.com>
15892             Chung-Ju Wu  <jasonwucj@gmail.com>
15894         * config/nds32/constants.md (unspec_volatile_element): Add enum values
15895         for unaligned access.
15896         * config/nds32/nds32-intrinsic.c: Implementation of expanding
15897         unaligned access.
15898         * config/nds32/nds32-intrinsic.md: Likewise.
15899         * config/nds32/nds32_intrinsic.h: Likewise.
15900         * config/nds32/nds32.h (nds32_builtins): Likewise.
15901         * config/nds32/nds32.opt (munaligned-access): New option.
15902         * config/nds32/nds32.c (nds32_asm_file_start): Display
15903         flag_unaligned_access status.
15905 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
15907         * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
15908         -mno-relax is present.
15909         * config/riscv/linux.h (LINK_SPEC): Ditto.
15911 2018-04-20  Martin Sebor  <msebor@redhat.com>
15913         PR c/85365
15914         * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
15915         for null pointers.
15916         (gimple_fold_builtin_stxcpy_chk): Same.
15917         * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
15919 2018-04-20  Michael Meissner  <meissner@linux.ibm.com>
15921         PR target/85456
15922         * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
15923         __powikf2 when long double is IEEE 128-bit.
15925 2018-04-20  Kito Cheng  <kito.cheng@gmail.com>
15927         * config/riscv/riscv.c (riscv_first_stack_step): Round up min
15928         step to make sure stack always aligned.
15930 2018-04-20  Carl Love  <cel@us.ibm.com>
15932         PR target/83402
15933         * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
15934         size check for arg0.
15936 2018-04-20  Nathan Sidwell  <nathan@codesourcery.com>
15937             Tom de Vries  <tom@codesourcery.com>
15939         PR target/85445
15940         * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
15941         Emit insns for calls too.
15942         (nvptx_find_par): Always look for worker-level predecessor insn.
15943         (nvptx_propagate): Add is_call parm, return bool.  Copy frame for
15944         calls.
15945         (nvptx_vpropagate, nvptx_wpropagate): Adjust.
15946         (nvptx_process_pars): Propagate frames for calls.
15948 2018-04-20  H.J. Lu  <hongjiu.lu@intel.com>
15950         PR target/85469
15951         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
15952         Removed.
15953         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
15954         (ix86_handle_option): Don't handle OPT_mibt.
15955         * config/i386/cet.h: Check __CET__ instead of __IBT__ and
15956         __SHSTK__.
15957         * config/i386/driver-i386.c (host_detect_local_cpu): Remove
15958         has_ibt and ibt.
15959         * config/i386/i386-c.c (ix86_target_macros_internal): Don't
15960         check OPTION_MASK_ISA_IBT nor flag_cf_protection.
15961         (ix86_target_macros): Define __CET__ with flag_cf_protection
15962         for -fcf-protection.
15963         * config/i386/i386.c (isa2_opts): Remove -mibt.
15964         * config/i386/i386.h (TARGET_IBT): Removed.
15965         (TARGET_IBT_P): Likewise.
15966         (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
15967         * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
15968         * config/i386/i386.opt (mcet): Update help message.
15969         (mshstk): Likewise.
15970         (mibt): Removed.
15971         * doc/invoke.texi: Remove -mibt.  Document __CET__.  Document
15972         -mcet as an alias for -mshstk.
15974 2018-04-20  Richard Biener <rguenther@suse.de>
15976         PR middle-end/85475
15977         * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
15978         complexity by forcing a single use of the multiply operand.
15980 2018-04-20  Martin Jambor  <mjambor@suse.cz>
15982         ipa/85449
15983         * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
15984         recursion dependency to only apply to non-clones.
15986 2018-04-20  Martin Jambor  <mjambor@suse.cz>
15988         ipa/85447
15989         * ipa-cp.c (create_specialized_node): Check that clones of
15990         self-recursive edges exist during IPA-CP.
15992 2018-04-19  Toon Moene  <toon@moene.org>
15994         * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
15995         by -O3.
15997 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
15999         PR tree-optimization/85467
16000         * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
16001         VECTOR_TYPE_P macro.  If type is vector type, VIEW_CONVERT_EXPR the
16002         VECTOR_CST element to type.
16004 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
16006         PR target/85397
16007         * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
16008         * config/i386/i386.md (builtin_setjmp_setup): Removed.
16009         (builtin_longjmp): Likewise.
16010         (save_stack_nonlocal): New pattern.
16011         (restore_stack_nonlocal): Likewise.
16013 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
16015         PR target/85404
16016         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16017         Replace ASM_OUTPUT_LABEL with fprintf.
16019 2018-04-19  H.J. Lu  <hongjiu.lu@intel.com>
16021         PR target/85417
16022         * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
16023         Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
16024         * config/i386/i386-c.c (ix86_target_macros_internal): Also
16025         define __IBT__ and __SHSTK__ for -fcf-protection.
16026         * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
16027         TARGET_IBT.
16028         (ix86_trampoline_init): Likewise.
16029         (x86_output_mi_thunk): Likewise.
16030         (ix86_notrack_prefixed_insn_p): Likewise.
16031         (ix86_option_override_internal): Don't disallow -fcf-protection.
16032         * config/i386/i386.md (rdssp<mode>): Also enable for
16033         -fcf-protection.
16034         (incssp<mode>): Likewise.
16035         (nop_endbr): Likewise.
16036         * config/i386/i386.opt (mcet): Change help message to built-in
16037         functions only.
16038         (mibt): Likewise.
16039         (mshstk): Likewise.
16040         * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
16041         on -fcf-protection.  Change -mcet, -mibt and -mshstk to only
16042         enable CET built-in functions.
16044 2018-04-19  Sebastian Peryt  <sebastian.peryt@intel.com>
16046         * common/config/i386/i386-common.c
16047         (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
16048         OPTION_MASK_ISA_MOVDIRI_UNSET,
16049         OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
16050         (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
16051         * config.gcc (movdirintrin.h): New header.
16052         * config/i386/cpuid.h (bit_MOVDIRI,
16053         bit_MOVDIR64B): New bits.
16054         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
16055         and -mmvodir64b.
16056         * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
16057         (VOID, PVOID, PCVOID)): New function types.
16058         * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
16059         __builtin_ia32_directstoreu_u64,
16060         __builtin_ia32_movdir64b): New builtins.
16061         * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
16062         * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
16063         and -mmovdiri.
16064         (ix86_valid_target_attribute_inner_p): Ditto.
16065         (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
16066         and VOID_FTYPE_PUNSIGNED_UNSIGNED.
16067         (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
16068         * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
16069         TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
16070         * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
16071         (movdiri<mode>, movdir64b_<mode>): New.
16072         * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
16073         * config/i386/immintrin.h: Include movdirintrin.h.
16074         * config/i386/movdirintrin.h: New file.
16075         * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
16077 2018-04-19  Richard Biener  <rguenther@suse.de>
16079         PR middle-end/85455
16080         * cfg.c (clear_bb_flags): When loop state says we have
16081         marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
16083 2018-04-19  Richard Biener  <rguenther@suse.de>
16085         PR tree-optimization/84737
16086         * tree-vect-data-refs.c (vect_copy_ref_info): New function
16087         copying restrict info.
16088         (vect_setup_realignment): Use it.
16089         * tree-vectorizer.h (vect_copy_ref_info): Declare.
16090         * tree-vect-stmts.c (vectorizable_store): Copy ref info from
16091         the first DR to all generated stores.
16092         (vectorizable_load): Likewise for loads.
16094 2018-04-19  Jakub Jelinek  <jakub@redhat.com>
16096         PR tree-optimization/85446
16097         * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
16098         the integral and pointer types to have the same precision.
16100         * doc/install.texi: Document --disable-cet being the default and
16101         --enable-cet=auto.
16103 2018-04-18  Martin Liska  <mliska@suse.cz>
16105         * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
16106         style.
16108 2018-04-18  Martin Liska  <mliska@suse.cz>
16110         Revert
16111         2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
16113         PR ipa/83983
16114         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
16115         arguments if they are comparable.
16117 2018-04-18  Martin Liska  <mliska@suse.cz>
16119         Revert
16120         2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
16122         PR lto/84805
16123         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
16124         incomplete types.
16126 2018-04-18  H.J. Lu  <hongjiu.lu@intel.com>
16128         PR target/85388
16129         * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
16130         ENDBR after calling __morestack.
16132 2018-04-18  David Malcolm  <dmalcolm@redhat.com>
16134         PR jit/85384
16135         * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
16136         by using gcc_base_ver to generate a gcc_driver_version, and use
16137         it when generating GCC_DRIVER_NAME.
16138         * configure: Regenerate.
16140 2018-04-18  Jakub Jelinek  <jakub@redhat.com>
16142         PR target/81084
16143         * config.gcc: Obsolete powerpc*-*-*spe*.
16145 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
16147         PR debug/84637
16148         * dbxout.c (dbxout_int): Perform negation in unsigned int type.
16149         (stabstr_D): Change type of unum from unsigned int to
16150         unsigned HOST_WIDE_INT.  Perform negation in unsigned HOST_WIDE_INT
16151         type.
16153 2018-04-17  Jim Wilson  <jimw@sifive.com>
16155         PR 84856
16156         * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
16157         RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
16158         Set arg_pointer_offset after using pretend_args_size.
16160 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
16162         PR rtl-optimization/85431
16163         * dse.c (record_store): Ignore zero width stores.
16165         PR sanitizer/85230
16166         * asan.c (handle_builtin_stack_restore): Adjust comment.  Emit
16167         __asan_allocas_unpoison call and last_alloca_addr = new_sp before
16168         __builtin_stack_restore rather than after it.
16169         * builtins.c (expand_asan_emit_allocas_unpoison): Pass
16170         arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
16171         argument instead of virtual_dynamic_stack_rtx.
16173 2018-04-17  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16175         * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
16176         New prototype.
16177         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
16178         Add note to error message to explain internal mapping of overloaded
16179         built-in function name to non-overloaded built-in function name.
16180         * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
16181         function.
16183 2018-04-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
16185         PR target/85424
16186         * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
16187         where the inputs overlap with the output.
16189 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
16191         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
16192         (=v, v) alternative and explicit "memory" attribute.
16193         (vec_extract_lo_<mode><mask_name>): Likewise.  Also add
16194         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
16195         attributes.
16196         (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
16197         "sselog1" type instead of "sselog".
16198         (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
16199         "sselog".  Remove explicit "memory" attribute.
16200         (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
16201         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
16202         attributes.
16203         (vec_extract_hi_v32hi): Merge all alternatives into one, use
16204         "sselog1" type instead of "sselog".  Remove explicit "memory"
16205         attribute.
16206         (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
16207         use "sselog1" type instead of "sselog".  Remove explicit "memory"
16208         attribute.
16209         (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
16210         "type", "prefix", "prefix_extra", "length_immediate" and "mode"
16211         attributes.
16212         (vec_extract_hi_v64qi): Merge all alternatives into one, use
16213         "sselog1" type instead of "sselog".  Remove explicit "memory"
16214         attribute.
16215         (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
16216         use "sselog1" type instead of "sselog".  Remove explicit "memory"
16217         attribute.
16219         PR target/85430
16220         * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
16222         PR middle-end/85414
16223         * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
16224         on a SUBREG.
16226 2018-04-17  Martin Jambor  <mjambor@suse.cz>
16228         PR ipa/85421
16229         * ipa-cp.c (create_specialized_node): Call
16230         expand_all_artificial_thunks if necessary.
16232 2018-04-17  Martin Liska  <mliska@suse.cz>
16234         PR lto/85405
16235         * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
16236         in message, remote space in between '_G' and '('.
16238 2018-04-17  Jakub Jelinek  <jakub@redhat.com>
16240         PR target/85281
16241         * config/i386/sse.md (reduces<mode><mask_scalar_name>,
16242         avx512f_vmcmp<mode>3<round_saeonly_name>,
16243         avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
16244         avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
16245         avx512f_rndscale<mode><round_saeonly_name>,
16246         avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
16247         avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
16248         Use %<iptr>2 instead of %2 for -masm=intel.
16249         (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
16250         avx512f_vcvttss2usi<round_saeonly_name>,
16251         avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
16252         -masm=intel.
16253         (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
16254         avx512f_vcvttsd2usi<round_saeonly_name>,
16255         avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
16256         Use %q1 instead of %1 for -masm=intel.
16257         (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
16258         avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
16259         of %3 for -masm=intel.
16260         (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
16261         -masm=intel.
16262         (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
16263         -masm=intel.
16264         (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
16265         -masm=intel.
16266         (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
16267         %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
16268         %g1.
16269         (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
16270         -masm=intel.
16271         (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
16272         %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
16273         %g1 and one with %0 and %1.
16274         (avx512er_vmrcp28<mode><round_saeonly_name>,
16275         avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
16276         %1 for -masm=intel.
16277         (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
16278         avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
16279         avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
16280         of %0 and %{%4%} for -masm=intel.
16281         (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
16282         avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
16283         avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
16284         order of %0 and %{%5%}%{z%} for -masm=intel.
16286 2018-04-17  Jan Hubicka  <jh@suse.cz>
16288         PR lto/85405
16289         * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
16291 2018-04-17  Martin Liska  <mliska@suse.cz>
16293         PR ipa/85329
16294         * multiple_target.c (create_dispatcher_calls): Set apostrophes
16295         for target_clone error message.  Make default implementation
16296         clone to be a local declaration.
16297         (separate_attrs): Add new argument and check for an empty
16298         string.
16299         (expand_target_clones): Handle it.
16300         (ipa_target_clone): Make redirection just for target_clones
16301         functions.
16303 2018-04-16  Cesar Philippidis  <cesar@codesourcery.com>
16304             Tom de Vries  <tom@codesourcery.com>
16306         PR middle-end/84955
16307         * omp-expand.c (expand_oacc_for): Add dummy false branch for
16308         tiled basic blocks without omp continue statements.
16310 2018-04-16  Aaron Sawdey  <acsawdey@linux.ibm.com>
16312         PR target/83660
16313         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
16314         vec_extract expression as having side effects to make sure it gets
16315         a cleanup point.
16317 2018-04-16  H.J. Lu  <hongjiu.lu@intel.com>
16319         PR target/85403
16320         * config/i386/i386.c (get_builtin_code_for_version): Check
16321         error_mark_node.
16323 2018-04-16  Olga Makhotina  <olga.makhotina@intel.com>
16325         PR target/84331
16326         * config.gcc: Support "skylake".
16327         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
16328         PROCESSOR_SKYLAKE.
16329         * config/i386/i386.c (m_SKYLAKE): Define.
16330         (processor_target_table): Add "skylake".
16331         (ix86_option_override_internal): Add "skylake".
16332         (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
16333         PROCESSOR_CANNONLAKE.
16334         (get_builtin_code_for_version): Fix priority for
16335         PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
16336         PROCESSOR_SKYLAKE-AVX512.
16337         * config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
16338         (processor_type): Add PROCESSOR_SKYLAKE.
16340 2018-04-16  Paolo Carlini  <paolo.carlini@oracle.com>
16341             Jason Merrill  <jason@redhat.com>
16343         PR c++/85112
16344         * convert.c (convert_to_integer_1): Use direct recursion for
16345         enumeral types and types with a precision less than the number
16346         of bits in their mode.
16348 2018-04-16  Julia Koval  <julia.koval@intel.com>
16350         PR target/84413
16351         * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
16352         X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
16354 2018-04-14  Segher Boessenkool  <segher@kernel.crashing.org>
16356         PR target/85293
16357         * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
16358         * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
16359         and -mno-direct-move.
16361 2018-04-13  Paul A. Clarke  <pc@us.ibm.com>
16363         PR target/83402
16364         * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
16365         Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
16366         Ensure negative shifts result in {0}.
16368 2018-04-13  Vladimir Makarov  <vmakarov@redhat.com>
16370         PR rtl-optimization/79916
16371         * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
16372         regs (if any) to define how to gnerate SD moves when LRA is in
16373         progress.
16375 2018-04-13  Jakub Jelinek  <jakub@redhat.com>
16377         PR rtl-optimization/85393
16378         * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
16379         * except.c (expand_dw2_landing_pad_for_region): Make static.
16380         * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
16381         a label and unconditional jump to old_bb, rather than
16382         expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
16383         basic block.
16385         PR rtl-optimization/85376
16386         * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
16387         zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
16388         instead of a specific value.
16390 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
16391             Bin Cheng  <bin.cheng@arm.com>
16393         PR tree-optimization/82965
16394         PR tree-optimization/83991
16395         * cfgloopanal.c (expected_loop_iterations_unbounded): Add
16396         by_profile_only parameter.
16397         * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
16398         information if the loop was predicted to iterate too many times.
16399         * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
16401 2018-04-13  Jan Hubicka  <hubicka@ucw.cz>
16403         PR lto/71991
16404         * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
16405         always inline.
16407 2018-04-13  Martin Liska  <mliska@suse.cz>
16408             Jakub Jelinek  <jakub@redhat.com>
16410         PR middle-end/81657
16411         * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
16412         * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
16413         * builtins.c (expand_builtin_memory_copy_args): Use
16414         BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
16415         handle dest_addr == pc_rtx.
16417 2018-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
16419         PR target/85291
16420         * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
16421         asked to not generate direct moves.
16422         (fix_trunc<mode>si2_stfiwx): Similar.
16423         (fix_trunc<mode>si2_internal): Similar.
16425 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
16427         PR debug/83157
16428         * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
16429         * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
16430         lookup if dest in some wider mode is known to be const0_rtx and
16431         if so, record permanent equivalence for it to be ZERO_EXTEND of
16432         the narrower mode destination.
16434 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
16436         * lto-streamer-out.c (output_function): Revert 259346.
16437         * omp-expand.c (expand_oacc_for): Likewise.
16439 2018-04-12  Alexander Monakov  <amonakov@ispras.ru>
16441         PR rtl-optimization/85354
16442         * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
16443         * sel-sched.c (sel_global_init): ... here.
16445 2018-04-12  Eric Botcazou  <ebotcazou@adacore.com>
16447         PR target/85238
16448         * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
16449         * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
16450         mode for PE-COFF targets.
16451         * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
16452         (i386_pe_asm_lto_end): Likewise.
16453         * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
16454         (TARGET_ASM_LTO_END): Likewise.
16455         * config/i386/winnt.c (saved_debug_info_level): New static variable.
16456         (i386_pe_asm_lto_start): New function.
16457         (i386_pe_asm_lto_end): Likewise.
16459 2018-04-12  Cesar Philippidis  <cesar@codesourcery.com>
16460             Richard Biener  <rguenther@suse.de>
16462         PR middle-end/84955
16463         * lto-streamer-out.c (output_function): Fix CFG loop state before
16464         streaming out.
16465         * omp-expand.c (expand_oacc_for): Handle calls to internal
16466         functions like regular functions.
16468 2018-04-12  Richard Biener  <rguenther@suse.de>
16470         PR lto/85371
16471         * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
16472         for the early LTO debug to properly generate references to it
16473         during DIE emission.  Do not re-use that for the skeleton for
16474         split-dwarf.
16475         (dwarf2out_early_finish): Likewise.
16477 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
16479         PR target/85328
16480         * config/i386/sse.md
16481         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
16482         <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
16483         vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
16484         vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
16485         and output is a reg, avoid creating invalid lowpart subreg, but
16486         instead split into a 512-bit move.  Don't split if not AVX512VL,
16487         input is xmm16+ reg and output is a mem.
16488         (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
16489         vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
16490         xmm16+ reg and output is a mem.
16492 2018-04-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
16494         * config/s390/s390.c (s390_output_indirect_thunk_function): Check
16495         also for flag_dwarf2_cfi_asm.
16497 2018-04-12  Jakub Jelinek  <jakub@redhat.com>
16499         PR rtl-optimization/85342
16500         * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
16501         a bool scalar var inside of the loop instead.  Don't try to update
16502         recog_data.operand after failed apply_change_group.
16504 2018-04-12  Tom de Vries  <tom@codesourcery.com>
16506         PR target/85296
16507         * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
16508         (nvptx_assemble_decl_begin): Add undefined param.  Declare undefined
16509         array with flexible array member as array without given dimension.
16510         (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
16511         argument for undefined param to true.
16513 2018-04-11  Aaron Sawdey  <acsawdey@linux.ibm.com>
16515         PR target/85321
16516         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
16517         -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
16518         from PowerPC section.
16519         * config/rs6000/sysv4.opt (mcall-): Improve help text.
16520         * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
16521         help text that is too long.
16522         * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
16523         help text that is too long.
16524         * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
16525         help text that is too long.
16527 2018-04-11  Uros Bizjak  <ubizjak@gmail.com>
16529         * config/alpha/alpha.md (stack_probe_internal): Rename
16530         from "probe_stack".  Update all callers.
16532 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
16534         PR rtl-optimization/84566
16535         * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
16536         sched_macro_fuse_insns.
16538 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
16540         PR target/84301
16541         * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
16542         (compute_block_dependences): ... from here.
16544 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
16546         PR tree-optimization/85331
16547         * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
16548         from int to HOST_WIDE_INT.
16550 2018-04-11  Martin Jambor  <mjambor@suse.cz>
16552         PR ipa/84149
16553         * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
16554         (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
16555         not the same as the source val.
16556         (cgraph_edge_brings_value_p): New parameter.
16557         (gather_edges_for_value): Pass destination value to
16558         cgraph_edge_brings_value_p.
16559         (perhaps_add_new_callers): Likewise.
16560         (get_info_about_necessary_edges): Likewise and exclude values brought
16561         only by self-recursive edges.
16562         (create_specialized_node): Redirect only clones of self-calling edges.
16563         (+self_recursive_pass_through_p): New function.
16564         (find_more_scalar_values_for_callers_subset): Use it.
16565         (find_aggregate_values_for_callers_subset): Likewise.
16566         (known_aggs_to_agg_replacement_list): Removed.
16567         (decide_whether_version_node): Re-calculate known constants for all
16568         remaining context clones.
16570 2018-04-11  Richard Biener  <rguenther@suse.de>
16572         PR lto/85339
16573         * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
16574         from early DWARF output.
16575         (dwarf2out_early_finish): Output line info unconditionally into
16576         early DWARF and add reference to it.
16578 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
16580         PR target/85281
16581         * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
16582         (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
16583         other than V2DFmode using iptr mode attribute.
16584         (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
16586 2018-04-11  Alexander Monakov  <amonakov@ispras.ru>
16588         PR rtl-optimization/84659
16589         * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
16591 2018-04-11  Jakub Jelinek  <jakub@redhat.com>
16593         PR debug/85302
16594         * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
16595         SIZEP is NULL.
16596         (output_loc_list): Pass address of a dummy size variable even in the
16597         locview handling loop.
16598         (index_location_lists): Add comment on why skip_loc_list_entry can't
16599         call size_of_locs.
16601 2018-04-11  Thomas Preud'homme  <thomas.preudhomme@arm.com>
16603         PR target/85261
16604         * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
16605         into register.
16607 2018-04-10  Aaron Sawdey  <acsawdey@linux.ibm.com>
16609         PR target/85321
16610         * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
16611         -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
16612         and -mstring-compare-inline-limit.
16614 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
16616         PR target/85287
16617         * config/rs6000/rs6000.md (allocate_stack): Put the residual size
16618         for stack clash protection in a register whenever we need it to be in
16619         a register.
16621 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
16623         * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
16624         Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
16626 2018-04-10  Segher Boessenkool  <segher@kernel.crashing.org>
16628         PR target/85321
16629         * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
16630         the help text.
16631         (mlong-double-): Ditto.
16632         * config/rs6000/sysv4.opt (msdata=): Ditto.
16633         (mtls-size=): Ditto.
16635 2018-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
16637         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
16638         erroneous entries for
16639         "vector int vec_ldl (int, long int *)", and
16640         "vector unsigned int vec_ldl (int, unsigned long int *)".
16641         Add comments and entries for
16642         "vector bool char vec_ldl (int, bool char *)",
16643         "vector bool short vec_ldl (int, bool short *)",
16644         "vector bool int vec_ldl (int, bool int *)",
16645         "vector bool long long vec_ldl (int, bool long long *)",
16646         "vector pixel vec_ldl (int, pixel *)",
16647         "vector long long vec_ldl (int, long long *)",
16648         "vector unsigned long long vec_ldl (int, unsigned long long *)".
16649         * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
16650         type tree bool_long_long_type_node and correct definition of
16651         bool_V2DI_type_node to make reference to this new type tree.
16652         (rs6000_mangle_type): Replace erroneous reference to
16653         bool_long_type_node with bool_long_long_type_node.
16654         * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
16655         comments to emphasize sign distinctions for char and int types and
16656         replace RS6000_BTI_bool_long constant with
16657         RS6000_BTI_bool_long_long constant.  Also add comment to restrict
16658         use of RS6000_BTI_pixel.
16659         (bool_long_type_node): Remove this macro definition.
16660         (bool_long_long_type_node): New macro definition
16662 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
16664         PR rtl-optimization/85300
16665         * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
16666         into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
16667         simplify_unary_operation fails.
16669 2018-04-10  Martin Liska  <mliska@suse.cz>
16671         * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
16672         cgraph_edge and ipa_ref.
16674 2018-04-10  Jakub Jelinek  <jakub@redhat.com>
16676         PR target/85177
16677         PR target/85255
16678         * config/i386/sse.md
16679         (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
16680         computation of the VEC_MERGE selector from mask.
16681         (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
16682         Fix decoding of the VEC_MERGE selector into mask.
16684 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
16686         PR tree-optimization/85286
16687         * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
16689 2018-04-10  Richard Sandiford  <richard.sandiford@linaro.org>
16691         * final.c (final_1): Set insn_last_address as well as
16692         insn_current_address.
16694 2018-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16696         PR target/85173
16697         * explow.c (emit_stack_probe): Call validize_mem on memory location
16698         before passing it to gen_probe_stack.  Create address operand and
16699         legitimize it for the probe_stack_address case.
16701 2018-04-09  Jan Hubicka  <jh@suse.cz>
16703         PR lto/85078
16704         * ipa-devirt.c (rebuild_type_inheritance-hash): New.
16705         * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
16706         * tree.c (free_lang_data_in_type): Fix handling of binfos;
16707         walk basetypes.
16708         (free_lang_data): Rebuild type inheritance graph.
16710 2018-04-09  Martin Sebor  <msebor@redhat.com>
16712         * invoke.texi (-finline-small-functions): Mention other optimization
16713         options.
16714         (-findirect-inlining, -fpartial-inlining): Same.
16715         (-finline-functions-called-once): Same.
16716         (-freorder-blocks-and-partition): Same.
16718 2018-04-09  Jan Hubicka  <jh@suse.cz>
16720         PR rtl/84058
16721         * cfgcleanup.c (try_forward_edges): Do not give up on crossing
16722         jumps; choose last target that matches the criteria (i.e.
16723         no partition changes for non-crossing jumps).
16724         * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
16725         support for redirecting crossing jumps to non-crossing.
16727 2018-04-09  Alexey Brodkin  <abrodkin@synopsys.com>
16729         * config/arc/arc.c (arc_expand_prologue): Set stack usage info
16730         also for naked functions.
16732 2018-04-09  Claudiu Zissulescu  <claziss@synopsys.com>
16734         * config/arc/arc.md (add_shift): New pattern.
16735         (add_shift2): Likewise.
16736         (sub_shift): Likewise.
16737         (sub_shift_cmp0_noout): Likewise.
16738         (compare_si_ashiftsi): Likewise.
16739         (xbfu_cmp0_noout): New combine pattern.
16740         (xbfu_cmp0"): Likewise.
16741         (movsi_set_cc_insn): Place the predicable variant first.
16742         (commutative_binary_cmp0_noout): Remove clobber.
16743         (commutative_binary_cmp0): New pattern.
16744         (noncommutative_binary_cmp0): Likewise.
16745         (noncommutative_binary_cmp0_noout): Likewise.
16746         (noncommutative_binary_comparison_result_used): Removed.
16747         (rsub_cmp0): New pattern.
16748         (rsub_cmp0_noout): Likewise.
16749         (extzvsi): Changed, keep only meaningful variants.
16750         (SQH, SEZ): New iterators.
16751         (SQH_postfix): New mode attribute.
16752         (SEZ_prefix): New code attribute.
16753         (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
16754         (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
16755         * config/arc/predicates.md (cc_set_register): Use CC_REG instead
16756         of numerical value.
16757         (noncommutative_operator): Check the availability of barrel
16758         shifter option.
16760 2018-04-09  Richard Biener  <rguenther@suse.de>
16762         PR tree-optimization/85284
16763         * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
16764         Only use the niter constraining form of simple_iv when the exit
16765         is always executed.
16767 2018-04-09  Tom de Vries  <tom@codesourcery.com>
16769         PR target/84041
16770         * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
16771         (define_expand "*memory_barrier"): New define_expand.
16772         (define_insn "memory_barrier"): New insn.
16774 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
16776         PR rtl-optimization/80463
16777         PR rtl-optimization/83972
16778         PR rtl-optimization/83480
16780         * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
16781         correct producer for the insn.
16782         (tidy_control_flow): Fixup seqnos in case of debug insns.
16784 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
16786         PR rtl-optimization/83913
16788         * sel-sched-ir.c (merge_expr_data): Choose the middle between two
16789         different sched-times when merging exprs.
16791 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
16793         PR rtl-optimization/83962
16795         * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
16796         tidy_fallthru_edge and tidy_control_flow.
16798 2018-04-09  Andrey Belevantsev  <abel@ispras.ru>
16800         PR rtl-optimization/83530
16802         * sel-sched.c (force_next_insn): New global variable.
16803         (remove_insn_for_debug): When force_next_insn is true, also leave only
16804         next insn in the ready list.
16805         (sel_sched_region): When the region wasn't scheduled, make another pass
16806         over it with force_next_insn set to 1.
16808 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
16810         * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
16811         into tm_file.
16812         * config/nds32/constants.md (unspec_volatile_element): Add enum values
16813         for interrupt control.
16814         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
16815         functions for interrupt control.
16816         * config/nds32/nds32-intrinsic.md: Likewise.
16817         * config/nds32/nds32_intrinsic.h: Likewise.
16818         * config/nds32/nds32.h (nds32_builtins): Likewise.
16820 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
16822         * config/nds32/nds32.c (nds32_init_machine_status,
16823         nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
16824         strict_aligned_p field.
16825         (nds32_expand_to_rtl_hook): New function.
16826         (TARGET_EXPAND_TO_RTL_HOOK): Define.
16827         * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
16829 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
16830             Chung-Ju Wu  <jasonwucj@gmail.com>
16832         * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
16833         * config/nds32/nds32-n7.md: New file.
16834         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
16835         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
16836         pipeline.
16837         * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
16838         * config/nds32/nds32.md (pipeline_model): Add n7.
16839         * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
16840         * config/nds32/pipelines.md: Include n7 settings.
16842 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
16843             Chung-Ju Wu  <jasonwucj@gmail.com>
16845         * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
16846         * config/nds32/nds32-e8.md: New file.
16847         * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
16848         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
16849         pipeline.
16850         * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
16851         * config/nds32/nds32.md (pipeline_model): Add e8.
16852         * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
16853         * config/nds32/pipelines.md: Include e8 settings.
16855 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
16856             Chung-Ju Wu  <jasonwucj@gmail.com>
16858         * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
16859         * config/nds32/nds32-n8.md: New file.
16860         * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
16861         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
16862         pipeline.
16863         * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
16864         * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
16865         * config/nds32/nds32.md (pipeline_model): Add n8.
16866         * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
16867         * config/nds32/pipelines.md: Include n8 settings.
16869 2018-04-08  Kito Cheng  <kito.cheng@gmail.com>
16870             Chung-Ju Wu  <jasonwucj@gmail.com>
16872         * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
16873         * config/nds32/nds32-n9-2r1w.md: New file.
16874         * config/nds32/nds32-n9-3r2w.md: New file.
16875         * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
16876         nds32_register_ports): New or modify for cpu n9.
16877         * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
16878         pipeline.
16879         * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
16880         * config/nds32/nds32-utils.c: New file.
16881         * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
16882         TARGET_MUL_SLOW): Define.
16883         * config/nds32/nds32.md (pipeline_model): New attribute.
16884         * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
16885         New options that support cpu n9.
16886         * config/nds32/pipelines.md: Include n9 settings.
16887         * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
16889 2018-04-08  Chung-Ju Wu  <jasonwucj@gmail.com>
16891         * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
16892         information if necessary.
16893         (output_cond_branch_compare_zero): Likewise.
16894         * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
16895         (nds32_target_alignment): Refine for alignment.
16896         * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
16897         (FUNCTION_BOUNDARY): Modify.
16898         * config/nds32/nds32.md (call_internal, call_value_internal): Consider
16899         align case.
16900         * config/nds32/nds32.opt (malways-align, malign-functions): New.
16902 2018-04-08  Monk Chiang  <sh.chiang04@gmail.com>
16904         * config/nds32/constants.md (unspec_volatile_element): Add values for
16905         TLB operation and data prefetch.
16906         * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
16907         functions for TLB operation and data prefetch.
16908         * config/nds32/nds32-intrinsic.md: Likewise.
16909         * config/nds32/nds32_intrinsic.h: Likewise.
16910         * config/nds32/nds32.c (nds32_dpref_names): Likewise.
16911         (nds32_print_operand): Likewise.
16912         * config/nds32/nds32.h (nds32_builtins): Likewise.
16914 2018-04-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
16915         Andrew Pinski <pinsika@gcc.gnu.org>
16917         PR middle-end/82976
16918         * match.pd: Use constant_boolean_node of correct type instead of
16919         boolean_true_node or boolean_false_node for simplifying
16920         pointer comparisons to zero.
16922 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
16924         PR tree-optimization/80021
16925         * tree.c (verify_type_variant): Make error call in verify_variant_match
16926         translatable and remove final full stop.
16928 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
16930         * config/nds32/constants.md (unspec_volatile_element): Add
16931         UNSPEC_VOLATILE_EH_RETURN.
16932         * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
16933         nds32_output_stack_pop): Support dwarf exception handling process.
16934         * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
16935         * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
16936         exception handling process.
16937         (nds32_compute_stack_frame): Likewise.
16938         (nds32_return_addr_rtx): Likewise.
16939         (nds32_initial_elimination_offset): Likewise.
16940         (nds32_expand_prologue): Likewise.
16941         (nds32_expand_epilogue): Likewise.
16942         (nds32_dynamic_chain_address): New function.
16943         * config/nds32/nds32.h (machine_function): Add fields for dwarf
16944         exception handling.
16945         (DYNAMIC_CHAIN_ADDRESS): Define.
16946         (EH_RETURN_DATA_REGNO): Define.
16947         (EH_RETURN_STACKADJ_RTX): Define.
16948         * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
16949         patterns for dwarf exception handling.
16951 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
16953         * config/nds32/nds32.h: Clean up obsolete macros.
16955 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
16957         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16958         Add enum values for particular instructions.
16959         * config/nds32/nds32-intrinsic.c: Implementation of expanding
16960         particular intrinsic functions.
16961         * config/nds32/nds32-intrinsic.md: Likewise.
16962         * config/nds32/nds32_intrinsic.h: Likewise.
16963         * config/nds32/nds32.h (nds32_builtins): Likewise.
16964         * config/nds32/nds32.md (type): Add pbsad and pbsada.
16965         (btst, ave): New patterns for particular instructions.
16967 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
16969         * config/nds32/constants.md (unspec_element, unspec_volatile_element):
16970         Add enum values for atomic load/store and memory sync.
16971         * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
16972         and memory sync.
16973         * config/nds32/nds32-intrinsic.md: Likewise.
16974         * config/nds32/nds32_intrinsic.h: Likewise.
16975         * config/nds32/nds32.h (nds32_builtins): Likewise.
16977 2018-04-07  Jakub Jelinek  <jakub@redhat.com>
16979         PR tree-optimization/85257
16980         * fold-const.c (native_encode_vector): If not all elts could fit
16981         and off is -1, return 0 rather than offset.
16982         * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
16983         (offseti - offset2) / BITS_PER_UNIT as 4th argument to
16984         native_encode_expr.  Verify len * BITS_PER_UNIT >= maxsizei.  Don't
16985         adjust buffer in native_interpret_expr call.
16987 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
16989         * config/nds32/constants.md (unspec_volatile_element): Add cache
16990         control enum values.
16991         * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
16992         * config/nds32/nds32-intrinsic.md: Add cache control patterns.
16993         * config/nds32/nds32.c (nds32_cctl_names): New.
16994         (nds32_print_operand): Handle cache control register names.
16995         * config/nds32/nds32.h (nds32_builtins): New enum values.
16996         * config/nds32/nds32_intrinsic.h: Add cache control enum types and
16997         macros.
16998         * config/nds32/nds32.md (type): Add mmu.
16999         * config/nds32/pipelines.md (simple_insn): Add mmu.
17001 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17003         * config/nds32/nds32.md (type): Remove call.
17004         * config/nds32/pipelines.md (simple_insn): Likewise.
17006 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17008         * config/nds32/constants.md (unspec_volatile_element): Add
17009         UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
17010         UNSPEC_VOLATILE_FMFCFG.
17011         * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
17012         description for fmfcfg and fmfcsr.
17013         (bdesc_1arg): Add fmtcsr.
17014         (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
17015         (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
17016         * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
17017         unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
17018         unspec_fmfcfg): New patterns.
17019         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
17020         NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
17021         NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
17022         * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
17023         __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
17024         __nds32__fmfcfg): Define.
17026 2018-04-07  Monk Chiang  <sh.chiang04@gmail.com>
17028         * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
17029         intrinsic register names.
17030         * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
17031         intrinsic register enum values and macros.
17033 2018-04-07  Chung-Ju Wu  <jasonwucj@gmail.com>
17035         * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
17036         for load/store addressing form.
17037         (nds32_print_operand_address): Likewise.
17039 2018-04-06  Eric Botcazou  <ebotcazou@adacore.com>
17041         PR target/85196
17042         * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
17043         based on LABEL_REF.  Remove useless assertion.
17044         (pic_address_needs_scratch): Fix formatting.
17045         (sparc_legitimize_pic_address): Minor tweaks.
17046         (sparc_delegitimize_address): Adjust assertion accordingly.
17047         * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
17048         into symbolic_operand.
17049         (movsi_high_pic_label_ref): Likewise.
17050         (movsi_lo_sum_pic_label_ref): Likewise.
17051         (movdi_pic_label_ref): Likewise.
17052         (movdi_high_pic_label_ref): Likewise.
17053         (movdi_lo_sum_pic_label_ref): Likewise.
17055 2018-04-06  Amaan Cheval  <amaan.cheval@gmail.com>
17057         * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
17058         custom LIB_SPEC setup.
17060 2018-04-06  Ruslan Bukin  <br@bsdpad.com>
17061             Kito Cheng  <kito.cheng@gmail.com>
17063         * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
17064         * config/riscv/freebsd.h: New.
17066 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
17068         * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
17069         * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
17070         file.
17072 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
17073             Kito Cheng  <kito.cheng@gmail.com>
17075         * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
17076         nds32_output_call, nds32_symbol_binds_local_p): New functions.
17077         * config/nds32/nds32-protos.h (nds32_output_call,
17078         nds32_output_return): Declare.
17079         * config/nds32/nds32.md: Refine all the call and return patterns.
17081 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
17083         PR debug/85252
17084         * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
17085         build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
17087         PR rtl-optimization/84872
17088         * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
17089         nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
17090         EDGE_CROSSING edge.
17092 2018-04-06  Tamar Christina  <tamar.christina@arm.com>
17094         * expr.c (copy_blkmode_to_reg): Revert 254862.
17095         * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
17097 2018-04-06  Richard Biener  <rguenther@suse.de>
17099         PR middle-end/85244
17100         * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
17101         after seeing a component reference with an adjacent field.  Treat
17102         refs to arrays at struct end of external decls similar to
17103         refs to unconstrained commons.
17105 2018-04-06  Jakub Jelinek  <jakub@redhat.com>
17107         PR sanitizer/85213
17108         * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
17109         look through SAVE_EXPRs with non-side-effects argument.  Adjust
17110         recursive calls.
17111         (fold_comparison): Adjust twoval_comparison_p caller, don't handle
17112         save_p here.
17114 2018-04-06  Richard Biener  <rguenther@suse.de>
17116         PR middle-end/85180
17117         * alias.c (find_base_term): New wrapper around find_base_term
17118         unwinding CSELIB_VAL_PTR changes.
17119         (find_base_term): Do not restore CSELIB_VAL_PTR during the
17120         recursion.
17122 2018-04-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
17124         * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
17125         instructions.
17126         * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
17127         constant definitions.
17128         ("nop"): lr 0,0 -> nopr r0
17129         ("nop_lr0", "nop_lr1"): New insn definitions.
17131 2018-04-06  Chung-Ju Wu  <jasonwucj@gmail.com>
17133         * config/nds32/nds32.md (*stack_push, *stack_pop): Use
17134         NDS32_V3PUSH_AVAILABLE_P macro.
17136 2018-04-06  Monk Chiang  <sh.chiang04@gmail.com>
17137             Chung-Ju Wu  <jasonwucj@gmail.com>
17139         * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
17140         (nds32*-*-*): Add float and fpu_config into supported_defaults.
17141         * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
17142         Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
17143         * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
17144         UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
17145         * config/nds32/constraints.md: New constraints and checking for hard
17146         float configuration.
17147         * config/nds32/iterators.md: New mode iterator and attribute for hard
17148         float configuration.
17149         * config/nds32/nds32-doubleword.md: Use hard float alternatives and
17150         patterns.
17151         * config/nds32/nds32-fpu.md: New file.
17152         * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
17153         deal with hard float code generation.
17154         * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
17155         ARCH_V3S.
17156         (abi_type, float_reg_number): New enum type.
17157         * config/nds32/nds32-predicates.c: New predicates for hard float.
17158         * config/nds32/nds32-protos.h: Declare functions for hard float.
17159         * config/nds32/nds32.c: Implementation for hard float configuration.
17160         * config/nds32/nds32.h: Definitions for hard float configuration.
17161         * config/nds32/nds32.md: Include hard float machine description and
17162         modify patterns for hard float configuration.
17163         * config/nds32/nds32.opt: New options for hard float configuration.
17164         * config/nds32/predicates.md: New predicates for hard float
17165         configuration.
17167 2018-04-06  Kuan-Lin Chen  <kuanlinchentw@gmail.com>
17169         * common/config/nds32/nds32-common.c
17170         (nds32_option_optimization_table): Enable -mreleax-hint by default.
17172 2018-04-05  Jakub Jelinek  <jakub@redhat.com>
17174         PR middle-end/85195
17175         * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
17176         CONSTRUCTOR_ELT (ctor, ...)->value.
17178 2018-04-05  Uros Bizjak  <ubizjak@gmail.com>
17180         PR target/85193
17181         * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
17183 2018-04-05  Tom de Vries  <tom@codesourcery.com>
17185         PR target/85204
17186         * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
17187         cond jump.
17189 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
17190             Kito Cheng  <kito.cheng@gmail.com>
17192         * config/nds32/constraints.md (U33): Fine-tune checking condition.
17193         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
17194         * config/nds32/nds32.h (nds32_16bit_address_type): Add
17195         ADDRESS_POST_MODIFY_LO_REG_IMM3U.
17197 2018-04-05  Shiva Chen  <shiva0217@gmail.com>
17198             Kito Cheng  <kito.cheng@gmail.com>
17200         * config/nds32/constraints.md (Ufe): New memory constraint.
17201         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
17202         nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
17203         * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
17204         operands.
17205         * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
17206         * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
17208 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17210         * config/nds32/nds32.md: Use optimize_size in the condition for
17211         alu-shift instructions.
17213 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17215         * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
17217 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17219         * config/nds32/nds32.md (negsi2): Refine pattern.
17221 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
17222             Chung-Ju Wu  <jasonwucj@gmail.com>
17224         * config/nds32/iterators.md (shift_rotate): New code iterator.
17225         (shift): New code attribute.
17226         * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
17227         * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
17228         * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
17229         * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
17230         bit-wise operations.
17231         (andsi3, *andsi3): Ditto.
17232         (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
17233         (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
17234         (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
17235         * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
17236         nds32_ior_operand, nds32_xor_operand): New predicates.
17238 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17240         * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
17241         (addsi3, subsi3): ... this.
17243 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17245         * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
17247 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17249         * config/nds32/nds32.md: Adjust indention.
17251 2018-04-05  Kito Cheng  <kito.cheng@gmail.com>
17253         * config/nds32/nds32.md (feature): New attribute.
17255 2018-04-05  Chung-Ju Wu  <jasonwucj@gmail.com>
17257         * config/nds32/nds32.md (subtype): New attribute.
17259 2018-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
17261         PR target/85203
17262         * config/arm/arm-builtins.c (arm_expand_builtin): Change
17263         expansion to perform a bitwise AND of the argument followed by a
17264         boolean negation of the result.
17266 2018-04-04  Peter Bergner  <bergner@vnet.ibm.com>
17268         PR rtl-optimization/84878
17269         * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
17270         the basic block.  Assert the use reference is not artificial and that
17271         it has an associated insn.
17273 2018-04-04  Michael Matz  <matz@suse.de>
17275         * builtins.c (compute_objsize): Pass correct operand
17276         to array_at_struct_end_p.
17278 2018-04-04  Richard Biener  <rguenther@suse.de>
17280         PR lto/85176
17281         * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
17282         from contexts for DINFO_LEVEL_TERSE and below.
17284 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
17286         * config/nds32/nds32-doubleword.md (move_<mode>): Require
17287         resiter_operand condition.
17288         * config/nds32/nds32.md (*move<mode>): Ditto.
17290 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
17291             Monk Chiang  <sh.chiang04@gmail.com>
17293         * config/nds32/nds32.md (movmisalign<mode>): New pattern.
17295 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
17297         * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
17299 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
17300             Kito Cheng  <kito.cheng@gmail.com>
17302         * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
17303         nds32_cond_code_str, output_cond_branch,
17304         output_cond_branch_compare_zero, nds32_expand_cbranch,
17305         nds32_expand_cstore, nds32_expand_movcc,
17306         nds32_output_cbranchsi4_equality_zero,
17307         nds32_output_cbranchsi4_equality_reg,
17308         nds32_output_cbranchsi4_equality_reg_or_const_int,
17309         nds32_output_cbranchsi4_greater_less_zero: New functions.
17310         * config/nds32/nds32-protos.h (nds32_expand_cbranch,
17311         nds32_expand_cstore, nds32_expand_movcc,
17312         nds32_output_cbranchsi4_equality_zero,
17313         nds32_output_cbranchsi4_equality_reg,
17314         nds32_output_cbranchsi4_equality_reg_or_const_int,
17315         nds32_output_cbranchsi4_greater_less_zero): Declare.
17316         * config/nds32/predicates.md (nds32_movecc_comparison_operator,
17317         nds32_rimm11s_operand): New predicates.
17318         * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
17319         * config/nds32/nds32.md: Rewrite all the branch and conditional move
17320         patterns.
17322 2018-04-04  Kito Cheng  <kito.cheng@gmail.com>
17324         * config/nds32/nds32-doubleword.md: Refine all the instruction type.
17325         * config/nds32/nds32.md: Ditto.
17326         * config/nds32/pipelines.md: Ditto.
17328 2018-04-04  Richard Biener  <rguenther@suse.de>
17330         PR tree-optimization/85168
17331         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
17332         propagating abnormals.
17334 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
17336         * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
17338 2018-04-04  Chung-Ju Wu  <jasonwucj@gmail.com>
17339             Kito Cheng  <kito.cheng@gmail.com>
17341         * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
17342         * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
17343         * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
17344         (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
17345         * config/nds32/nds32.md (sibcall_internal): New.
17346         (sibcall_register): Remove.
17347         (sibcall_immediate): Remove.
17348         (sibcall_value_internal): New.
17349         (sibcall_value_register): Remove.
17350         (sibcall_value_immediate): Remove.
17351         * config/nds32/predicates.md (nds32_general_register_operand): New.
17352         (nds32_call_address_operand): New.
17354 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
17356         PR rtl-optimization/85167
17357         * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
17358         bb_defs if *split_p, instead preinitialize it to NULL.
17360         PR tree-optimization/85156
17361         * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
17362         evaluating the argument multiple times.
17364 2018-04-03  Bill Schmidt  <wschmidt@linux.ibm.com>
17366         * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
17367         than vector.
17368         (_mm_cvtpd_ps): Likewise.
17369         (_mm_cvttpd_epi32): Likewise.
17370         * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
17371         * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
17372         vector, pixel, and bool following altivec.h include.
17374 2018-04-03  Martin Sebor  <msebor@redhat.com>
17376         * doc/extend.texi (Common Function Attributes): Clarify.
17377         (const attribute): Likewise.
17378         (pure attribute): Likewise.
17380 2018-04-03  Jakub Jelinek  <jakub@redhat.com>
17382         PR target/85169
17383         * config/i386/i386.c (ix86_expand_vector_set): Use
17384         HOST_WIDE_INT_1U << elt instead of 1 << elt.  Formatting fix.
17386 2018-04-03  Uros Bizjak  <ubizjak@gmail.com>
17388         * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
17389         instructions when changing rounding bits to preserve precision bits
17390         in the x87 control word.
17392 2018-04-03  Martin Liska  <mliska@suse.cz>
17394         PR tree-optimization/82491
17395         * rtl.h (strip_offset_and_add): Replace += suboffset with
17396         poly_uint64 () + suboffset.
17398 2018-03-29  Martin Liska  <mliska@suse.cz>
17399             Martin Jambor  <mjambor@suse.cz>
17401         PR ipa/84947
17402         * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
17403         param_type is not an integral or pointer type.
17405 2018-04-03  Richard Biener  <rguenther@suse.de>
17407         * sese.h (recompute_all_dominators): Remove.
17409 2018-04-02  Martin Sebor  <msebor@redhat.com>
17411         * doc/invoke.texi (-Wrestrict): Fix typos.
17413 2018-04-02  Jim Wilson  <jimw@sifive.com>
17415         * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
17416         * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
17417         (<optab>di3, <optab>si3_extend): Likewise.
17418         (<optab>si3_mask, <optab>si3_mask_1): New.
17419         (<optab>di3_mask, <optab>di3_mask_1): New.
17420         (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
17421         (lshrsi3_zero_extend_1): Use VOIDmode shift count.
17422         * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
17424 2018-04-02  Gerald Pfeifer  <gerald@pfeifer.com>
17426         * doc/cpp.texi (Variadic Macros): Fix line continuation in an
17427         example.
17429 2018-04-02  Chung-Ju Wu  <jasonwucj@gmail.com>
17431         * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
17432         (nds32_canonicalize_comparison): New function.
17434 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
17435             Kito Cheng  <kito.cheng@gmail.com>
17436             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
17438         * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
17439         * config/nds32/constants.md (unspec_volatile_element): Add
17440         UNSPEC_VOLATILE_RELAX_GROUP.
17441         * config/nds32/nds32-relax-opt.c: New file.
17442         * config/nds32/nds32-predicates.c
17443         (nds32_symbol_load_store_p): New function.
17444         * config/nds32/nds32-protos.h
17445         (nds32_symbol_load_store_p): Declare function.
17446         (make_pass_nds32_relax_opt): Declare new rtl pass function.
17447         * config/nds32/nds32.c
17448         (nds32_register_pass): New function to register pass.
17449         (nds32_register_passes): New function to register passes.
17450         * config/nds32/nds32.md (relax_group): New pattern.
17451         * config/nds32/nds32.opt (mrelax-hint): New option.
17452         * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
17454 2018-04-01  Kito Cheng  <kito.cheng@gmail.com>
17456         * config/nds32/t-nds32: Modify files dependency.
17458 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
17460         * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
17461         (PROFILE_HOOK): Define its implementation.
17463 2018-04-01  Chung-Ju Wu  <jasonwucj@gmail.com>
17465         * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
17466         type and 32-bit size.
17468 2018-04-01  Jakub Jelinek  <jakub@redhat.com>
17470         PR middle-end/85090
17471         * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
17472         (V_128_256): New mode iterator.
17473         (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
17474         (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
17475         (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
17476         of V.
17477         * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
17478         V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
17480 2018-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
17482         PR target/83315
17483         * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
17484         NaN inputs correctly.
17486 2018-03-30  Peter Bergner  <bergner@vnet.ibm.com>
17488         PR target/80546
17489         * config/rs6000/vsx.md (??r): New mode attribute.
17490         (*vsx_mov<mode>_64bit): Use it.
17491         (*vsx_mov<mode>_32bit): Likewise.
17493 2018-03-30  Martin Sebor  <msebor@redhat.com>
17495         PR tree-optimization/84818
17496         * builtins.c (check_access): Use warning_n.
17498 2018-03-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
17500         PR target/83822
17501         * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
17502         condition.
17503         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
17504         condition.
17506 2018-03-30  Julia Koval  <julia.koval@intel.com>
17508         PR target/84413
17509         * x86-tune.def (movx, partial_reg_dependency): Enable for
17510         m_SKYLAKE_AVX512.
17512 2018-03-29  Vladimir Makarov  <vmakarov@redhat.com>
17514         PR inline-asm/84985
17515         * lra-constraints.c (process_alt_operands): Move setting
17516         this_alternative_matches below.
17518 2018-03-29  Martin Liska  <mliska@suse.cz>
17520         PR lto/84995.
17521         * doc/invoke.texi: Document how LTO works with debug info.
17522         Describe auto-load support of binutils.  Mention 'x86-64'
17523         as valid option value of -march option.
17525 2018-03-29  Jakub Jelinek  <jakub@redhat.com>
17527         * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
17529         PR c/85094
17530         * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
17531         For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
17532         OEP_NO_HASH_CHECK for recursive call, to avoid exponential
17533         checking.
17535 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
17537         PR target/84912
17538         * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
17539         (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
17540         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
17541         for RS6000_BTM_POWERPC64.
17542         (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
17543         (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
17544         * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
17545         definition.
17546         (DIVDE): Use it.
17547         (DIVDEU): Likewise.
17549 2018-03-28  Carl Love  <cel@us.ibm.com>
17551         Revert
17552         2017-09-27  Carl Love  <cel@us.ibm.com>
17554         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
17555         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
17556         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
17557         fctiw instruction.
17559 2018-03-28  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17561         * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
17562         instead of __vector bool.
17563         (_mm_max_pu8): Likewise.
17564         (_mm_min_pi16): Likewise.
17566 2018-03-28  Peter Bergner  <bergner@vnet.ibm.com>
17568         PR target/84912
17569         * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
17570         (DIVWEUO): Likewise.
17571         (DIVDEO): Likewise.
17572         (DIVDEUO): Likewise.
17573         * config/rs6000/rs6000.c (builtin_function_type): Remove support for
17574         DIVWEUO and DIVDEUO.
17575         * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
17576         (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
17577         (div_extend): Likewise.
17578         * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
17579         builtin function.
17580         (__builtin_divweuo): Likewise.
17581         (__builtin_divdeo): Likewise.
17582         (__builtin_divdeuo): Likewise.
17584 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
17586         PR target/85095
17587         * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
17588         *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
17590         PR tree-optimization/82004
17591         * gimple-match-head.c (optimize_pow_to_exp): New function.
17592         * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
17593         Don't fold to exp if optimize_pow_to_exp is false.
17595 2018-03-28  Martin Liska  <mliska@suse.cz>
17597         PR other/84819
17598         * calls.c (initialize_argument_information): Fix trailing space.
17599         * common.opt: Fix typo and provide better explanation for
17600         -fsanitize-coverage option.
17601         * config/i386/i386.opt: Fix typo.
17603 2018-03-28  Jakub Jelinek  <jakub@redhat.com>
17604             Martin Liska  <mliska@suse.cz>
17606         PR sanitizer/85081
17607         * gimplify.c (asan_poison_variable): Don't do the check for
17608         gimplify_omp_ctxp here.
17609         (gimplify_decl_expr): Do it here.
17610         (gimplify_target_expr): Likewise.
17612 2018-03-28  Martin Liska  <mliska@suse.cz>
17614         PR target/84988
17615         * config/i386/i386.c (ix86_function_arg_advance): Do not call
17616         chkp_type_bounds_count if MPX is not enabled.
17618 2018-03-27  Chung-Ju Wu  <jasonwucj@gmail.com>
17620         * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
17622 2018-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
17624         PR target/84914
17625         * config/rs6000/rs6000.c (create_complex_muldiv): New helper
17626         function to create the function decl for complex long double
17627         multiply and divide for -mabi=ieeelongdouble.
17628         (init_float128_ieee): Call it.
17630 2018-03-27  H.J. Lu  <hongjiu.lu@intel.com>
17632         PR target/85044
17633         * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
17634         -fcf-protection=branch -mibt.
17635         * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
17637 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
17639         PR target/81863
17640         * config/arm/arm.c (arm_valid_symbolic_address): Handle
17641         arm_word_relocations.
17643 2018-03-27  Cesar Philippidis  <cesar@codesourcery.com>
17645         PR target/85056
17646         * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
17647         extern array declarations.
17649 2018-03-27  Richard Biener  <rguenther@suse.de>
17651         PR middle-end/84067
17652         * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
17653         explicit single_use checks.
17655 2018-03-27  Richard Biener  <rguenther@suse.de>
17657         PR tree-optimization/85082
17658         * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
17659         Valueize the VUSE.
17661 2018-03-27  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
17663         * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
17664         * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
17665         Turn on fasynchronous-unwind-tables and funwind-tables.
17667 2018-03-26  Uros Bizjak  <ubizjak@gmail.com>
17669         PR target/85073
17670         * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
17671         (*bmi_blsr_<mode>_ccz): Ditto.
17673 2018-03-26  Tom de Vries  <tom@codesourcery.com>
17675         PR tree-optimization/85063
17676         * omp-general.c (offloading_function_p): New function.  Factor out
17677         of ...
17678         * omp-offload.c (pass_omp_target_link::gate): ... here.
17679         * omp-general.h (offloading_function_p): Declare.
17680         * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
17681         with attribute omp declare target for offloading functions.
17683 2018-03-24  Richard Sandiford  <richard.sandiford@linaro.org>
17685         PR tree-optimization/84005
17686         * tree-data-ref.h (get_base_for_alignment): Declare.
17687         * tree-data-ref.c (get_base_for_alignment_1): New function.
17688         (get_base_for_alignment): Likewise.
17689         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
17690         get_base_for_alignment to find a suitable base object, instead
17691         of always using drb->base_address.
17693 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
17695         PR inline-asm/85022
17696         * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
17697         known size by default.
17699 2018-03-23  Vladimir Makarov  <vmakarov@redhat.com>
17701         PR inline-asm/85030
17702         * lra-constraints.c (process_alt_operands): Don't match BLKmode
17703         and non BLKmode operands.
17705 2018-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17707         PR target/85026
17708         * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
17709         Clean up attributes.
17711 2018-03-23  Richard Biener  <rguenther@suse.de>
17713         PR debug/85020
17714         * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
17715         we are going to emit early debug for LTO.
17717 2018-03-23  Jakub Jelinek  <jakub@redhat.com>
17719         PR inline-asm/85034
17720         * function.c (match_asm_constraints_1): Don't optimize if input
17721         doesn't satisfy general_operand predicate for output's mode.
17723         PR inline-asm/85022
17724         * alias.c (write_dependence_p): Don't require for x_canonicalized
17725         non-VOIDmode if x has VOIDmode.
17727         PR sanitizer/85029
17728         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
17729         just don't try to optimize it rather than assert it never happens.
17731 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17733         * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
17734         macro expansions for definition of ST_INTERNAL_<mode> and
17735         LD_INTERNAL_<mode> builtins.
17736         * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
17737         Remove prototype.
17738         * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
17739         function.
17740         (altivec_expand_st_builtin): Likewise.
17741         (altivec_expand_builtin): Remove calls to deleted functions.
17742         (rs6000_address_for_altivec): Delete this function.
17743         * config/rs6000/vector.md: Remove expands for
17744         vector_altivec_load_<mode> and vector_altivec_store_<mode>.
17746 2018-03-22  Sudakshina Das  <sudi.das@arm.com>
17748         PR target/84826
17749         * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
17750         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
17751         re-computing once computed.
17752         (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
17753         (arm_init_machine_status): Initialize
17754         machine->static_chain_stack_bytes.
17756 2018-03-22  Kelvin Nilsen  <kelvin@gcc.gnu.org>
17758         PR target/84760
17759         * doc/extend.texi: Add four new prototypes for vec_ld.
17760         * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
17761         definitions for more logical presentation.
17762         * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
17763         entries for V1TI variants of __builtin_altivec_ld builtin.
17764         * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
17765         handling of V1TI variant of LVX icode pattern.
17766         (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
17767         (rs6000_gimple_fold_builtin): Likewise.
17768         (altivec_init_builtins): Add code to define
17769         __builtin_altivec_lvx_v1ti function.
17771 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
17773         PR inline-asm/84941
17774         * function.c (match_asm_constraints_1): Don't do the optimization
17775         if input isn't a REG, SUBREG, MEM or constant.
17777 2018-03-22  Tom de Vries  <tom@codesourcery.com>
17779         PR tree-optimization/84956
17780         * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
17781         bb_has_abnormal_pred.
17783 2018-03-22  Jakub Jelinek  <jakub@redhat.com>
17785         PR sanitizer/85018
17786         * dwarf2asm.c (dw2_output_indirect_constant_1): Set
17787         DECL_INITIAL (decl) to decl at the end.
17788         * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
17789         adjust the comment.
17791 2018-03-21  Joseph Myers  <joseph@codesourcery.com>
17793         * doc/extend.texi (__builtin_tgmath): Document when complex
17794         integer types are treated as _Complex _Float64.
17796 2018-03-21  Tom de Vries  <tom@codesourcery.com>
17798         * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
17800 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
17802         PR tree-optimization/84960
17803         * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
17804         if it is ENTRY block, move them into single succ of ENTRY in that case.
17806 2018-03-21  Richard Sandiford  <richard.sandiford@linaro.org>
17808         PR tree-optimization/84811
17809         * poly-int.h (poly_span_traits): Remove the T3 parameter and
17810         promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
17811         (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
17812         (known_subrange_p): Update accordingly.  Cast each value involved
17813         in the size comparison, rather than casting the result of the
17814         subtraction.
17816 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
17818         PR tree-optimization/84982
17819         * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
17820         by flipping the least significant bit rather than all bits from
17821         bitpos to bitpos + bitsize - 1.
17823 2018-03-21  Nathan Sidwell  <nathan@acm.org>
17825         * doc/extend.texi (Deprecated Features): Remove mention of
17826         long-deleted deprecations.
17828 2018-03-21  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
17830         PR jit/84288
17831         * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
17832         * configure: Regenerate.
17834 2018-03-21  Tom de Vries  <tom@codesourcery.com>
17836         PR tree-optimization/83126
17837         * tree-parloops.c (num_phis): New function.
17838         (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
17840 2018-03-21  Nathan Sidwell  <nathan@acm.org>
17842         * doc/extend.texi (Deprecated Features): Update deprecated flags,
17843         mention anon-struct/union members and trailing attributes.
17845 2018-03-21  Bin Cheng  <bin.cheng@arm.com>
17847         PR tree-optimization/84969
17848         * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
17849         builtin memset partitions if they set different rhs values.
17851 2018-03-21  Jakub Jelinek  <jakub@redhat.com>
17853         PR rtl-optimization/84989
17854         * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
17855         VEC_DUPLICATE with scalar result mode.
17857 2018-03-21  Martin Liska  <mliska@suse.cz>
17859         PR ipa/84963
17860         * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
17861         not intended return statement.
17863 2018-03-21  Martin Liska  <mliska@suse.cz>
17865         PR target/84988
17866         * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
17867         (chkp_find_bound_slots_1): Limit number of iterations.
17869 2018-03-20  David H. Gutteridge  <dhgutteridge@sympatico.ca>
17871         PR target/84838
17872         * Minor grammar fixes for x86 options.
17874 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
17876         PR debug/84875
17877         * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
17878         holding REG_CFA_RESTORE notes, instead turn them into a USE.
17880 2018-03-20  Peter Bergner  <bergner@vnet.ibm.com>
17882         PR target/83789
17883         * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
17884         (altivec_lvx_<mode>_1op): Likewise.
17885         (altivec_stvx_<mode>_2op): Likewise.
17886         (altivec_stvx_<mode>_1op): Likewise.
17887         (altivec_lvx_<VM2:mode>): New define_expand.
17888         (altivec_stvx_<VM2:mode>): Likewise.
17889         (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
17890         (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
17891         (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
17892         (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
17893         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
17894         (rs6000_gen_lvx): Likewise.
17895         * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
17896         (altivec_expand_stv_builtin): Likewise.
17897         (altivec_expand_builtin): Likewise.
17898         * config/rs6000/vector.md: Likewise.
17900 2018-03-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
17902         PR target/82518
17903         * config/arm/arm.c (arm_array_mode_supported_p): Return false for
17904         BYTES_BIG_ENDIAN.
17906 2018-03-20  Richard Biener  <rguenther@suse.de>
17908         PR target/84986
17909         * config/i386/i386.c (ix86_add_stmt_cost): Only cost
17910         sign-conversions as zero, fall back to standard scalar_stmt
17911         cost for the rest.
17913 2018-03-20  Martin Liska  <mliska@suse.cz>
17915         PR ipa/84825
17916         * predict.c (rebuild_frequencies): Handle case when we have
17917         PROFILE_ABSENT, but flag_guess_branch_prob is false.
17919 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
17921         PR target/84990
17922         * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
17923         flag_section_anchors.
17924         * varasm.c (use_blocks_for_decl_p): Remove hack for
17925         dw2_force_const_mem.
17927         PR target/84845
17928         * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
17929         to ...
17930         (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this.  If pseudos can't
17931         be created, use lowpart_subreg of operands[0] rather than operands[0]
17932         itself.
17933         (*aarch64_reg_<mode>3_minus_mask): Rename to ...
17934         (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
17935         (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
17936         and n constraint instead of aarch64_shift_imm_di and Usd.
17937         (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
17938         (*aarch64_<optab>_reg_minus<mode>3): ... this.
17940 2018-03-20  Sudakshina Das  <sudi.das@arm.com>
17942         PR target/82989
17943         * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
17944         to favor GPR over NEON registers.
17945         (<shift>di3_neon): Likewise.
17947 2018-03-20  Tom de Vries  <tom@codesourcery.com>
17949         PR target/84952
17950         * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
17951         (nvptx_process_pars): Emit bar.sync asap and alap.
17953 2018-03-20  Tom de Vries  <tom@codesourcery.com>
17955         PR target/84954
17956         * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
17957         seen_label if seen_label is already set.
17959 2018-03-20  Jakub Jelinek  <jakub@redhat.com>
17961         PR target/84945
17962         * config/i386/i386.c (fold_builtin_cpu): For features above 31
17963         use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
17964         Use 1U instead of 1.  Formatting fixes.
17966         PR c/84953
17967         * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
17968         instead of TREE_TYPE (s1) for the return value.
17970 2018-03-19  Jakub Jelinek  <jakub@redhat.com>
17972         PR tree-optimization/84946
17973         * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
17974         bitsize + bitsize in poly_uint64 rather than poly_int64.
17976         PR sanitizer/78651
17977         * dwarf2asm.c: Include fold-const.c.
17978         (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
17979         of decl rather than decl itself.
17981         PR rtl-optimization/84643
17982         * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
17984 2018-03-19  Maxim Ostapenko  <m.ostapenko@samsung.com>
17986         PR sanitizer/78651
17987         * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
17988         calling assemble_variable.
17990 2018-03-19  Sudakshina Das  <sudi.das@arm.com>
17992         PR target/81647
17993         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
17994         instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
17996 2018-03-19  Jim Wilson  <jimw@sifive.com>
17998         PR bootstrap/84856
17999         * config/riscv/riscv.c (riscv_function_arg_boundary): Use
18000         PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
18001         (riscv_first_stack_step): Likewise.
18002         (riscv_option_override): Use STACK_BOUNDARY instead of
18003         MIN_STACK_BOUNDARY.
18004         * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
18005         MIN_STACK_BOUNDARY.
18006         (BIGGEST_ALIGNMENT): Set to 128.
18007         (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
18008         (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
18009         STACK_BOUNDARY.
18011 2018-03-19  Richard Biener  <rguenther@suse.de>
18013         PR tree-optimization/84933
18014         * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
18015         values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
18017 2018-03-19  Richard Biener  <rguenther@suse.de>
18019         PR tree-optimization/84859
18020         * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
18021         (cond_if_else_store_replacement): Perform sinking operation on
18022         single-store BBs regardless of MAX_STORES_TO_SINK setting.
18023         Generalize what a BB with a single eligible store is.
18025 2018-03-19  Richard Biener  <rguenther@suse.de>
18027         PR tree-optimization/84929
18028         * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
18029         chrec_is_positive against non-chrec arg.
18031 2018-03-19  Tamar Christina  <tamar.christina@arm.com>
18033         PR target/84711
18034         * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
18036 2018-03-18  Martin Liska  <mliska@suse.cz>
18038         PR rtl-optimization/84635
18039         * regrename.c (build_def_use): Use matches_mode only when
18040         matches >= 0.
18042 2018-03-18  Richard Sandiford  <richard.sandiford@linaro.org>
18044         PR tree-optimization/84913
18045         * tree-vect-loop.c (vectorizable_reduction): Don't try to
18046         vectorize chains of COND_EXPRs.
18048 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
18050         * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
18052 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
18054         * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
18056 2018-03-18  Chung-Ju Wu  <jasonwucj@gmail.com>
18058         * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
18060 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
18061             Kito Cheng  <kito.cheng@gmail.com>
18063         * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
18064         * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
18065         (nds32_adjust_reg_alloc_order): New function.
18066         * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
18068 2018-03-17  Kito Cheng  <kito.cheng@gmail.com>
18070         * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
18071         nds32_print_operand, nds32_print_operand_address): Use
18072         HOST_WIDE_INT_PRINT_DEC instead.
18074 2018-03-17  Chung-Ju Wu  <jasonwucj@gmail.com>
18076         * config/nds32/nds32.c (nds32_register_priority): Modify cost.
18078 2018-03-17  Jakub Jelinek  <jakub@redhat.com>
18080         PR target/84902
18081         * config/i386/i386.c (initial_ix86_tune_features,
18082         initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
18083         unsigned long long.
18084         (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
18085         to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
18086         rather than 1u << ix86_tune.  Formatting fix.
18087         (ix86_option_override_internal): Change ix86_arch_mask from
18088         unsigned int to unsigned HOST_WIDE_INT, initialize to
18089         HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
18090         (ix86_function_specific_restore): Likewise.
18092 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
18094         PR target/84899
18095         * postreload.c (reload_combine_recognize_pattern): Perform
18096         INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
18097         truncate_int_for_mode the result for the destination's mode.
18099         PR c/84909
18100         * hsa-gen.c (mem_type_for_type): Fix comment typo.
18101         * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
18102         Likewise.
18103         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
18104         Likewise.
18106 2018-03-16  Vladimir Makarov  <vmakarov@redhat.com>
18108         PR target/84876
18109         * lra-assigns.c (lra_split_hard_reg_for): Don't use
18110         regno_allocno_class_array and sorted_pseudos.
18111         * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
18112         insns where regno is used.
18114 2018-03-16  Martin Liska  <mliska@suse.cz>
18116         PR ipa/84833
18117         * multiple_target.c (create_dispatcher_calls): Redirect
18118         reference in the symbol table.
18120 2018-03-16  Martin Liska  <mliska@suse.cz>
18122         PR ipa/84722
18123         * multiple_target.c (create_dispatcher_calls): Redirect also
18124         an alias.
18126 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
18128         PR c++/79937
18129         PR c++/82410
18130         * tree.h (TARGET_EXPR_NO_ELIDE): Define.
18131         * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
18132         TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
18134 2018-03-16  Julia Koval  <julia.koval@intel.com>
18136         * doc/invoke.texi (Skylake Server): Add CLWB.
18137         Cannonlake): Remove CLWB.
18139 2018-03-16  Jakub Jelinek  <jakub@redhat.com>
18141         PR tree-optimization/84841
18142         * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
18143         1 << 3.
18144         (FLOAT_ONE_CONST_TYPE): Define.
18145         (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
18146         (sort_by_operand_rank): Put entries with higher constant_type last
18147         rather than first to match comments.
18149 2018-03-15  Sandra Loosemore  <sandra@codesourcery.com>
18151         * config/nios2/nios2.md (movsi_internal): Fix thinko in
18152         split predicate.
18154 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
18156         PR c++/79085
18157         * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
18158         check and use address of target always.
18160 2018-03-15  H.J. Lu  <hongjiu.lu@intel.com>
18162         PR target/84574
18163         * config/i386/i386.c (indirect_thunk_needed): Update comments.
18164         (indirect_thunk_bnd_needed): Likewise.
18165         (indirect_thunks_used): Likewise.
18166         (indirect_thunks_bnd_used): Likewise.
18167         (indirect_return_needed): New.
18168         (indirect_return_bnd_needed): Likewise.
18169         (output_indirect_thunk_function): Add a bool argument for
18170         function return.
18171         (output_indirect_thunk_function): Don't generate alias for
18172         function return thunk.
18173         (ix86_code_end): Call output_indirect_thunk_function to generate
18174         function return thunks.
18175         (ix86_output_function_return): Set indirect_return_bnd_needed
18176         and indirect_return_needed instead of indirect_thunk_bnd_needed
18177         and indirect_thunk_needed.
18179 2018-03-15  Olga Makhotina  <olga.makhotina@intel.com>
18181         * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
18182         (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
18183         (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
18185 2018-03-15  David Malcolm  <dmalcolm@redhat.com>
18186             Paul Hua <paul.hua.gm@gmail.com>
18188         PR c/84852
18189         * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
18191 2018-03-15  Segher Boessenkool  <segher@kernel.crashing.org>
18193         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
18194         TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
18195         resp. SFmode cases.
18197 2018-03-15  Tamar Christina  <tamar.christina@arm.com>
18199         PR target/84711
18200         * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
18201         instead of GET_MODE_SIZE when comparing Units.
18203 2018-03-15  Vladimir Mezentsev  <vladimir.mezentsev@oracle.com>
18205         PR target/68256
18206         * varasm.c (hash_section): Return an unchangeble hash value
18207         * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
18208         Return !aarch64_can_use_per_function_literal_pools_p ().
18210 2018-03-15  Jakub Jelinek  <jakub@redhat.com>
18212         PR target/84860
18213         * optabs.c (emit_conditional_move): Pass address of cmode's copy
18214         rather than address of cmode as last argument to prepare_cmp_insn.
18216 2018-03-15  Julia Koval  <julia.koval@intel.com>
18218         * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
18219         F_AVX512VNNI, F_AVX512BITALG): New.
18221 2018-03-14  John David Anglin  <danglin@gcc.gnu.org>
18223         PR target/83451
18224         * config/pa/pa.c (pa_emit_move_sequence):  Always emit secondary reload
18225         insn for floating-point loads and stores.
18227 2018-03-14  Carl Love  <cel@us.ibm.com>
18229         * config/rs6000/rs6000-c.c: Add macro definitions for
18230         ALTIVEC_BUILTIN_VEC_PERMXOR.
18231         * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
18232         * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
18233         * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
18234         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
18235         UNSPEC_VPERMXOR.
18236         * config/doc/extend.texi: Add prototypes for vec_permxor.
18238 2018-03-14  David Malcolm  <dmalcolm@redhat.com>
18240         PR c/84852
18241         * diagnostic-show-locus.c (class layout_point): Convert m_line
18242         from int to linenum_type.
18243         (line_span::comparator): Use linenum "compare" function when
18244         comparing line numbers.
18245         (test_line_span): New function.
18246         (layout_range::contains_point): Convert param "row" from int to
18247         linenum_type.
18248         (layout_range::intersects_line_p): Likewise.
18249         (layout::will_show_line_p): Likewise.
18250         (layout::print_source_line): Likewise.
18251         (layout::should_print_annotation_line_p): Likewise.
18252         (layout::print_annotation_line): Likewise.
18253         (layout::print_leading_fixits): Likewise.
18254         (layout::annotation_line_showed_range_p): Likewise.
18255         (struct line_corrections): Likewise for field m_row.
18256         (line_corrections::line_corrections): Likewise for param "row".
18257         (layout::print_trailing_fixits): Likewise.
18258         (layout::get_state_at_point): Likewise.
18259         (layout::get_x_bound_for_row): Likewise.
18260         (layout::print_line): Likewise.
18261         (diagnostic_show_locus): Likewise for locals "last_line" and "row".
18262         (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
18263         * input.c (selftest::test_linenum_comparisons): New function.
18264         (selftest::input_c_tests): Call it.
18265         * selftest.c (selftest::test_assertions): Test ASSERT_GT,
18266         ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
18267         * selftest.h (ASSERT_GT): New macro.
18268         (ASSERT_GT_AT): New macro.
18269         (ASSERT_LT): New macro.
18270         (ASSERT_LT_AT): New macro.
18272 2018-03-14  Segher Boessenkool  <segher@kernel.crashing.org>
18274         PR rtl-optimization/84780
18275         * combine.c (distribute_links): Don't make a link based on pc_rtx.
18277 2018-03-14  Martin Liska  <mliska@suse.cz>
18279         * tree.c (record_node_allocation_statistics): Use
18280         get_stats_node_kind.
18281         (get_stats_node_kind): New function extracted from
18282         record_node_allocation_statistics.
18283         (free_node): Use get_stats_node_kind.
18285 2018-03-14  Richard Biener  <rguenther@suse.de>
18287         * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
18288         that the value-set of ANTIC_IN doesn't grow.
18290         Revert
18291         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
18292         member.
18293         (BB_VISITED_WITH_VISITED_SUCCS): New define.
18294         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
18296 2018-03-14  Julia Koval  <julia.koval@intel.com>
18298         * config.gcc (icelake-client, icelake-server): New.
18299         (icelake): Remove.
18300         * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
18301         (initial_ix86_arch_features): Ditto.
18302         (PTA_SKYLAKE): Add SGX.
18303         (PTA_ICELAKE): Remove.
18304         (PTA_ICELAKE_CLIENT): New.
18305         (PTA_ICELAKE_SERVER): New.
18306         (ix86_option_override_internal): Split up icelake on icelake client and
18307         icelake server.
18308         (get_builtin_code_for_version): Ditto.
18309         (fold_builtin_cpu): Ditto.
18310         * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
18311         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
18312         * config/i386/i386.h (processor_type): Ditto.
18313         * doc/invoke.texi: Ditto.
18315 2018-03-14  Jakub Jelinek  <jakub@redhat.com>
18317         PR sanitizer/83392
18318         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
18319         INTEGER_CST offset, add it together with bitpos / 8 and
18320         sign extend based on POINTER_SIZE.
18322         PR target/84844
18323         Revert
18324         2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
18326         PR target/78090
18327         * config/i386/constraints.md (Yc): New register constraint.
18328         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
18329         Use Yc constraint for alternative 2 of operand 0.  Remove
18330         preferred_for_speed attribute.
18332 2018-03-14  Richard Biener  <rguenther@suse.de>
18334         PR tree-optimization/84830
18335         * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
18336         with the old one to avoid oscillations.
18338 2018-03-13  Vladimir Makarov  <vmakarov@redhat.com>
18340         PR target/83712
18341         * lra-assigns.c (find_all_spills_for): Ignore uninteresting
18342         pseudos.
18343         (assign_by_spills): Return a flag of reload assignment failure.
18344         Do not process the reload assignment failures.  Do not spill other
18345         reload pseudos if they has the same reg class.  Update n if
18346         necessary.
18347         (lra_assign): Add a return arg.  Set up from the result of
18348         assign_by_spills call.
18349         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18350         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
18351         usage_insns if it is not NULL.
18352         (spill_hard_reg_in_range): New function.
18353         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18354         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18355         function prototypes.
18356         (lra_assign): Change prototype.
18357         * lra.c (lra): Add code to deal with fails by splitting hard reg
18358         live ranges.
18360 2018-03-01  Palmer Dabbelt  <palmer@sifive.com>
18362         * config/riscv/riscv.opt (mrelax): New option.
18363         * config/riscv/riscv.c (riscv_file_start): Emit ".option
18364         "norelax" when riscv_mrelax is disabled.
18365         * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
18367 2018-03-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
18369         PR target/84743
18370         * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
18371         reassociation for int modes.
18373 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
18375         * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
18376         Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
18377         for big-endian.
18378         * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
18379         * config/aarch64/aarch64-sve.md
18380         (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
18381         (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
18382         (*extend<mode><Vwide>2): Rename to...
18383         (aarch64_sve_extend<mode><Vwide>2): ...this.
18384         (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
18385         renaming the old pattern to...
18386         (aarch64_sve_punpk<perm_hilo>_<mode>): ...this.  Only define
18387         unsigned packs.
18388         (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
18389         define_expand, renaming the old pattern to...
18390         (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
18391         (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
18392         (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
18393         account when deciding which SVE instruction the optab should use.
18394         (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
18396 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
18398         * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
18399         (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
18400         (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
18401         (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
18402         (tlsdesc_small_<mode>): Turn a define_expand and use
18403         tlsdesc_small_sve_<mode> for SVE.  Rename original define_insn to...
18404         (tlsdesc_small_advsimd_<mode>): ...this.
18405         (tlsdesc_small_sve_<mode>): New pattern.
18407 2018-03-13  Richard Sandiford  <richard.sandiford@linaro.org>
18409         * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
18410         (UNSPEC_UMUL_HIGHPART): New constants.
18411         (MUL_HIGHPART): New int iteraor.
18412         (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
18413         * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
18414         define_expand.
18415         (*<su>mul<mode>3_highpart): New define_insn.
18417 2018-03-13  Eric Botcazou  <ebotcazou@adacore.com>
18419         PR lto/84805
18420         * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
18421         incomplete types.
18423 2018-03-13  Martin Liska  <mliska@suse.cz>
18425         PR ipa/84658.
18426         * (sem_item_optimizer::sem_item_optimizer): Initialize new
18427         vector.
18428         (sem_item_optimizer::~sem_item_optimizer): Release it.
18429         (sem_item_optimizer::merge_classes): Register variable aliases.
18430         (sem_item_optimizer::fixup_pt_set): New function.
18431         (sem_item_optimizer::fixup_points_to_sets): Likewise.
18432         * ipa-icf.h: Declare new variables and functions.
18434 2018-03-13  Jakub Jelinek  <jakub@redhat.com>
18436         PR middle-end/84834
18437         * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
18438         integer_pow2p@2 and test integer_pow2p in condition.
18439         (A < 0 ? C : 0): Similarly for @1.
18441         PR middle-end/84831
18442         * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
18443         characters starting at p contain '\0' character, don't look beyond
18444         that.
18446         PR target/84827
18447         * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
18448         pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
18450         PR target/84828
18451         * reg-stack.c (change_stack): Change update_end var from int to
18452         rtx_insn *, if non-NULL don't update just BB_END (current_block), but
18453         also call set_block_for_insn on the newly added insns and rescan.
18455         PR target/84786
18456         * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
18457         on the last operand.
18459         PR c++/84704
18460         * tree.c (stabilize_reference_1): Return save_expr (e) for
18461         STATEMENT_LIST even if it doesn't have side-effects.
18463 2018-03-12  Jonathan Wakely  <jwakely@redhat.com>
18465         * doc/invoke.texi (-mclflushopt): Fix spelling of option.
18467 2018-03-12  Renlin Li  <renlin.li@arm.com>
18469         * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
18470         aarch64_output_scalar_simd_mov_immediate.
18472 2018-03-12  Martin Sebor  <msebor@redhat.com>
18474         PR tree-optimization/83456
18475         * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
18476         for perfectly overlapping calls to memcpy.
18477         (gimple_fold_builtin_memory_chk): Same.
18478         (gimple_fold_builtin_strcpy): Handle no-warning.
18479         (gimple_fold_builtin_stxcpy_chk): Same.
18480         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
18482 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
18484         * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
18485         parameter.  Use it for SFmode.
18486         (rs6000_function_arg_advance_1): Adjust.
18487         (rs6000_function_arg): Adjust.
18488         (rs6000_gimplify_va_arg): Pass false for that new parameter.
18490 2018-03-12  Segher Boessenkool  <segher@kernel.crashing.org>
18492         PR rtl-optimization/84169
18493         PR rtl-optimization/84780
18494         * combine.c (can_combine_p): Check for a 2-insn combination whether
18495         the destination register is used between the two insns, too.
18497 2018-03-12  Richard Biener  <rguenther@suse.de>
18499         PR tree-optimization/84803
18500         * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
18501         for refs DR analysis didn't process.
18503 2018-03-12  Richard Biener  <rguenther@suse.de>
18505         PR tree-optimization/84777
18506         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
18507         force-vectorize loops ignore whether we are optimizing for size.
18509 2018-03-12  Chung-Ju Wu  <jasonwucj@gmail.com>
18511         * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
18512         (TARGET_MD_ASM_ADJUST): Define.
18514 2018-03-12  Monk Chiang  <sh.chiang04@gmail.com>
18515             Kito Cheng  <kito.cheng@gmail.com>
18516             Chung-Ju Wu  <jasonwucj@gmail.com>
18518         * config/nds32/nds32.c (nds32_compute_stack_frame,
18519         nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
18520         nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
18521         nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
18522         nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
18523         * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
18524         NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
18525         * config/nds32/nds32.md (prologue, epilogue): Use macro
18526         NDS32_V3PUSH_AVAILABLE_P to do checking.
18528 2018-03-11  Jakub Jelinek  <jakub@redhat.com>
18530         PR debug/58150
18531         * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
18532         DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
18533         but on TYPE_SIZE.  Don't do anything for ENUM_IS_OPAQUE if not creating
18534         a new die.  Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE.  Guard
18535         addition of most attributes on !orig_type_die or the attribute not
18536         being present already.  Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
18538 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
18539             Chung-Ju Wu  <jasonwucj@gmail.com>
18541         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
18542         __NDS32_VH__ macro.
18543         * config/nds32/nds32.opt (mvh): New option.
18545 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
18546             Chung-Ju Wu  <jasonwucj@gmail.com>
18548         * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
18549         function.
18550         * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
18551         * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
18552         definition.
18554 2018-03-11  Kito Cheng  <kito.cheng@gmail.com>
18555             Chung-Ju Wu  <jasonwucj@gmail.com>
18557         * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
18558         function.
18559         * config/nds32/nds32-multiple.md (strlensi): New pattern.
18560         * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
18562 2018-03-11  Monk Chiang  <sh.chiang04@gmail.com>
18563             Kito Cheng  <kito.cheng@gmail.com>
18564             Chung-Ju Wu  <jasonwucj@gmail.com>
18566         * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
18567         UNSPEC_FFMISM and UNSPEC_FLMISM.
18568         * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
18569         for ffb, ffmism and flmism.
18570         * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
18571         (unspec_ffmism): Ditto.
18572         (unspec_flmism): Ditto.
18573         (nds32_expand_builtin_impl): Check if string extension is available.
18574         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
18575         NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
18577 2018-03-10  Vladimir Makarov  <vmakarov@redhat.com>
18579         Reverting patch:
18580         2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
18582         PR target/83712
18583         * lra-assigns.c (assign_by_spills): Return a flag of reload
18584         assignment failure.  Do not process the reload assignment
18585         failures.  Do not spill other reload pseudos if they has the same
18586         reg class.
18587         (lra_assign): Add a return arg.  Set up from the result of
18588         assign_by_spills call.
18589         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18590         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
18591         usage_insns if it is not NULL.
18592         (spill_hard_reg_in_range): New function.
18593         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18594         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18595         function prototypes.
18596         (lra_assign): Change prototype.
18597         * lra.c (lra): Add code to deal with fails by splitting hard reg
18598         live ranges.
18600 2018-03-10  H.J. Lu  <hongjiu.lu@intel.com>
18602         PR target/84807
18603         * config/i386/i386.opt: Replace Enforcment with Enforcement.
18605 2018-03-10  Alexandre Oliva  <aoliva@redhat.com>
18607         PR debug/84620
18608         * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
18609         (dw_val_node): Add val_symbolic_view.
18610         * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
18611         (symview_upper_bound): New.
18612         (new_line_info_table): Initialize symviews_since_reset.
18613         (dwarf2out_source_line): Count symviews_since_reset and set
18614         symview_upper_bound.
18615         (dw_val_equal_p): Handle symview.
18616         (add_AT_symview): New.
18617         (print_dw_val): Handle symview.
18618         (attr_checksum, attr_checksum_ordered): Likewise.
18619         (same_dw_val_p, size_of_die): Likewise.
18620         (value_format, output_die): Likewise.
18621         (add_high_low_attributes): Use add_AT_symview for entry_view.
18622         (dwarf2out_finish): Reset symview_upper_bound, clear
18623         zero_view_p.
18625 2018-03-09  Peter Bergner  <bergner@vnet.ibm.com>
18627         PR target/83969
18628         * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
18629         Add strict argument and use it.
18630         (rs6000_split_multireg_move): Update for new strict argument.
18631         (mem_operand_gpr): Disallow all non-offsettable addresses.
18632         * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
18634 2018-03-09  Jakub Jelinek  <jakub@redhat.com>
18636         PR target/84772
18637         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
18638         temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
18639         * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
18641         PR c++/84767
18642         * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
18643         decl, use remap_type if we want to use the type.
18645 2018-03-09  Martin Sebor  <msebor@redhat.com>
18647         PR tree-optimization/84526
18648         * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
18649         Remove dead code.
18650         (builtin_access::generic_overlap): Be prepared to handle non-array
18651         base objects.
18653 2018-03-09  Alexandre Oliva  <aoliva@redhat.com>
18655         PR rtl-optimization/84682
18656         * lra-constraints.c (process_address_1): Check is_address flag
18657         for address constraints.
18658         (process_alt_operands): Likewise.
18659         * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
18660         preprocess_constraints.
18661         * recog.h (preprocess_constraints): Add oploc parameter.
18662         Adjust callers.
18663         * recog.c (preprocess_constraints): Test address_operand for
18664         CT_ADDRESS constraints.
18666 2018-03-09  Vladimir Makarov  <vmakarov@redhat.com>
18668         PR target/83712
18669         * lra-assigns.c (assign_by_spills): Return a flag of reload
18670         assignment failure.  Do not process the reload assignment
18671         failures.  Do not spill other reload pseudos if they has the same
18672         reg class.
18673         (lra_assign): Add a return arg.  Set up from the result of
18674         assign_by_spills call.
18675         (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
18676         * lra-constraints.c (split_reg): Add a new arg.  Use it instead of
18677         usage_insns if it is not NULL.
18678         (spill_hard_reg_in_range): New function.
18679         (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
18680         * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
18681         function prototypes.
18682         (lra_assign): Change prototype.
18683         * lra.c (lra): Add code to deal with fails by splitting hard reg
18684         live ranges.
18686 2018-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18688         PR target/83193
18689         * common/config/arm/arm-common.c (arm_parse_arch_option_name):
18690         Accept complain bool parameter.  Only emit errors if it is true.
18691         (arm_parse_cpu_option_name): Likewise.
18692         (arm_target_thumb_only): Adjust callers of the above.
18693         * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
18694         prototype to take a default true bool parameter.
18695         (arm_parse_arch_option_name): Likewise.
18697 2018-03-09  David Malcolm  <dmalcolm@redhat.com>
18698             Francois-Xavier Coudert  <fxcoudert@gcc.gnu.org>
18700         PR jit/64089
18701         PR jit/84288
18702         * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
18703         * configure: Regenerate.
18704         * configure.ac ("linker --version-script option"): New.
18705         ("linker soname option"): New.
18707 2018-03-09  Richard Biener  <rguenther@suse.de>
18709         PR tree-optimization/84775
18710         * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
18711         immediate uses of predicate stmts and mark them modified.
18713         Revert
18714         PR tree-optimization/84178
18715         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
18716         to caller.
18717         (version_loop_for_if_conversion): Delay update_ssa call.
18718         (tree_if_conversion): Delay update_ssa until after predicate
18719         insertion.
18721 2018-03-09  Eric Botcazou  <ebotcazou@adacore.com>
18723         PR target/84763
18724         * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
18725         when the function accesses prior frames.
18727 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
18729         PR debug/84456
18730         * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
18731         gen_llsym, otherwise call maybe_gen_llsym.
18733         PR inline-asm/84742
18734         * recog.c (asm_operand_ok): Return 0 if multi-character constraint
18735         has ',' character inside of it.
18737 2018-03-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
18739         PR target/84748
18740         * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
18741         as clobbering CC_REGNUM.
18743 2018-03-08  Richard Biener  <rguenther@suse.de>
18745         PR middle-end/84552
18746         * tree-scalar-evolution.c: Include tree-into-ssa.h.
18747         (follow_copies_to_constant): Do not follow SSA names registered
18748         for update.
18750 2018-03-08  Richard Biener  <rguenther@suse.de>
18752         PR tree-optimization/84178
18753         * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
18754         to caller.
18755         (version_loop_for_if_conversion): Delay update_ssa call.
18756         (tree_if_conversion): Delay update_ssa until after predicate
18757         insertion.
18759 2018-03-08  David Malcolm  <dmalcolm@redhat.com>
18761         PR tree-optimization/84178
18762         * tree-if-conv.c (release_bb_predicate): Remove the
18763         the assertion that the stmts have NULL use_ops.
18764         Discard the statements, asserting that they haven't
18765         yet been added to a BB.
18767 2018-03-08  Richard Biener  <rguenther@suse.de>
18769         PR tree-optimization/84746
18770         * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
18771         (phi_translate): Pass in destination ANTIC_OUT set.
18772         (phi_translate_1): Likewise.  For a simplified result lookup
18773         a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
18774         (phi_translate_set): Adjust.
18775         (do_pre_regular_insertion): Likewise.
18776         (do_pre_partial_partial_insertion): Likewise.
18778 2018-03-08  Martin Liska  <mliska@suse.cz>
18780         PR gcov-profile/84735
18781         * doc/gcov.texi: Document usage of profile files.
18782         * gcov-io.h: Document changes in the format.
18784 2018-03-08  Alexandre Oliva  <aoliva@redhat.com>
18786         PR debug/84404
18787         PR debug/84408
18788         * dwarf2out.c (struct dw_line_info_table): Update comments for
18789         view == -1.
18790         (FORCE_RESET_NEXT_VIEW): New.
18791         (FORCE_RESETTING_VIEW_P): New.
18792         (RESETTING_VIEW_P): Check for -1 too.
18793         (ZERO_VIEW_P): Likewise.
18794         (new_line_info_table): Force-reset next view.
18795         (dwarf2out_begin_function): Likewise.
18796         (dwarf2out_source_line): Simplify zero_view_p initialization.
18797         Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
18798         view directly.  Omit view when omitting .loc at line 0.
18800 2018-03-08  Jakub Jelinek  <jakub@redhat.com>
18802         PR tree-optimization/84740
18803         * tree-switch-conversion.c (process_switch): Call build_constructors
18804         only if info.phi_count is non-zero.
18806         PR tree-optimization/84739
18807         * tree-tailcall.c (find_tail_calls): Check call arguments against
18808         DECL_ARGUMENTS (current_function_decl) rather than
18809         DECL_ARGUMENTS (func) when checking for tail recursion.
18811 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
18813         * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
18814         Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
18815         Volker Reichelt's entry and add entries for people that perform
18816         GCC fuzzy testing and report numerous bugs.
18818 2018-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
18820         PR target/82411
18821         * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
18822         readonly data in sdata, if that is disabled.
18823         * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
18824         * doc/invoke.texi (RS/6000 and PowerPC Options): Document
18825         -mreadonly-in-sdata option.
18827 2018-03-07  Martin Sebor  <msebor@redhat.com>
18829         PR tree-optimization/84468
18830         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
18831         basic block when looking for nul assignment.
18833 2018-03-07  Eric Botcazou  <ebotcazou@adacore.com>
18835         PR target/84277
18836         * except.h (output_function_exception_table): Adjust prototype.
18837         * except.c (output_function_exception_table): Remove FNNAME parameter
18838         and add SECTION parameter.  Ouput one part of the table at a time.
18839         * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
18840         the first part of the exception table and emit unwind directives.
18841         * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
18842         (i386_pe_seh_cold_init): Likewise.
18843         * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
18844         (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
18845         * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
18846         (ix86_output_call_insn): Emit a nop in one more case for SEH.
18847         * config/i386/winnt.c: Include except.h.
18848         (struct seh_frame_state): Add reg_offset, after_prologue and
18849         in_cold_section fields.
18850         (i386_pe_seh_end_prologue): Set seh->after_prologue.
18851         (i386_pe_seh_cold_init): New function.
18852         (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
18853         to seh->in_cold_section.
18854         (seh_emit_push): Record the offset of the push.
18855         (seh_emit_save): Record the offet of the save.
18856         (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
18857         Test seh->after_prologue to disregard the epilogue.
18858         (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
18859         (i386_pe_end_cold_function): New function.
18861 2018-03-07  Jakub Jelinek  <jakub@redhat.com>
18863         PR fortran/84565
18864         * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
18865         aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
18867         PR c++/84704
18868         * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
18869         on tmp_var.
18870         * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
18871         don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
18873         PR middle-end/84723
18874         * multiple_target.c: Include tree-inline.h and intl.h.
18875         (expand_target_clones): Diagnose and fail if node->definition and
18876         !tree_versionable_function_p (node->decl).
18878 2018-03-06  John David Anglin  <danglin@gcc.gnu.org>
18880         * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
18881         sprint_ul.
18882         (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
18883         (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
18884         * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
18886 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
18888         PR target/84710
18889         * combine.c (try_combine): Use reg_or_subregno instead of handling
18890         just paradoxical SUBREGs and REGs.
18892 2018-03-06  Claudiu Zissulescu  <claziss@synopsys.com>
18894         * config/arc/arc.c (arc_finalize_pic): Remove function.
18895         (arc_must_save_register): We use single base PIC register, remove
18896         checks to save/restore the PIC register.
18897         (arc_expand_prologue): Likewise.
18898         * config/arc/arc-protos.h (arc_set_default_type_attributes):
18899         Remove.
18900         (arc_verify_short): Likewise.
18901         (arc_attr_type): Likewise.
18902         * config/arc/arc.c (arc_set_default_type_attributes): Remove.
18903         (walk_stores): Likewise.
18904         (arc_address_cost): Make it static.
18905         (arc_verify_short): Likewise.
18906         (branch_dest): Likewise.
18907         (arc_attr_type): Likewise.
18908         * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
18909         (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
18910         (arc_final_prescan_insn): Remove inserting the nops due to
18911         hardware hazards.  It is done in reorg step.
18912         (insn_length_variant_t): Remove.
18913         (insn_length_parameters_t): Likewise.
18914         (arc_insn_length_parameters): Likewise.
18915         (arc_get_insn_variants): Likewise.
18916         * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
18918 2018-03-06  Jakub Jelinek  <jakub@redhat.com>
18920         PR inline-asm/84683
18921         * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
18922         assertion failure.
18924         PR tree-optimization/84687
18925         * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
18926         on new_node->decl.
18927         * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
18929 2018-03-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18931         * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
18932         Rename to ppc_speculation_barrier.
18933         * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
18934         __builtin_ppc_speculation_barrier.
18936 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
18938         PR target/84700
18939         * combine.c (combine_simplify_rtx): Don't try to simplify if
18940         if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
18941         are equal to x.
18943 2018-03-05  Segher Boessenkool  <segher@kernel.crashing.org>
18945         * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
18946         to 32 bytes when compiling for POWER9.
18948 2018-03-05  Jakub Jelinek  <jakub@redhat.com>
18950         PR target/84564
18951         * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
18952         regparm >= 3 with no arg reg available also for calls with
18953         flag_force_indirect_call.  Pass decl to ix86_function_regparm.
18955         PR target/84524
18956         * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
18957         orig,vex.
18958         (*<plusminus_insn><mode>3): Likewise.  Remove <mask_operand3> uses.
18960 2018-03-05  Peter Bergner  <bergner@vnet.ibm.com>
18962         PR target/84264
18963         * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
18965 2018-03-05  Richard Biener  <rguenther@suse.de>
18967         PR tree-optimization/84486
18968         * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
18969         When inserting a __builtin_assume_aligned call set the LHS
18970         SSA name alignment info accordingly.
18972 2018-03-05  Wilco Dijkstra  <wdijkstr@arm.com>
18974         PR tree-optimization/84114
18975         * config/aarch64/aarch64.c (aarch64_reassociation_width)
18976         Avoid reassociation of FLOAT_MODE addition.
18978 2018-03-05  Olga Makhotina  <olga.makhotina@intel.com>
18980         * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
18981         OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
18982         OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
18983         (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
18984         * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
18985         * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
18986         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
18987         and -mwbnoinvd.
18988         * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
18989         __builtin_ia32_wbinvd): New builtins.
18990         (SPECIAL_ARGS2): New.
18991         * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
18992         (SPECIAL_ARGS2): New.
18993         * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
18994         (ix86_valid_target_attribute_inner_p): Ditto.
18995         (ix86_init_mmx_sse_builtins): Add special_args2.
18996         * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
18997         TARGET_WBNOINVD_P): New.
18998         * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
18999         (define_insn "wbinvd", define_insn "wbnoinvd"): New.
19000         * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
19001         * config/i386/immintrin.h (_wbinvd): New intrinsic.
19002         * config/i386/pconfigintrin.h: New file.
19003         * config/i386/wbnoinvdintrin.h: Ditto.
19004         * config/i386/x86intrin.h: Add headers pconfigintrin.h and
19005         wbnoinvdintrin.h.
19006         * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
19008 2018-03-05  Richard Biener  <rguenther@suse.de>
19010         PR tree-optimization/84670
19011         * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
19012         member.
19013         (BB_VISITED_WITH_VISITED_SUCCS): New define.
19014         (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
19015         (compute_antic_aux): Only assert the number of values in ANTIC_IN
19016         doesn't grow if all successors (recursively) were visited at least
19017         once.
19019 2018-03-05  Richard Biener  <rguenther@suse.de>
19021         PR tree-optimization/84650
19022         * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
19023         if executed in the loop pipeline.
19025 2018-03-05  Sandra Loosemore  <sandra@codesourcery.com>
19027         * doc/configfiles.texi (Configuration Files): Move info about
19028         conditionalizing $target-protos.h to...
19029         * doc/sourcebuild.texi (Back End): Here.  Explain how $target.h
19030         differs from $target-protos.h.
19032 2018-03-05  Kito Cheng  <kito.cheng@gmail.com>
19033             Chung-Ju Wu  <jasonwucj@gmail.com>
19035         * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
19036         * config/nds32/nds32-multiple.md (setmemsi): Define.
19037         * config/nds32/nds32-memory-manipulation.c
19038         (nds32_gen_dup_4_byte_to_word_value): New.
19039         (emit_setmem_word_loop): New.
19040         (emit_setmem_byte_loop): New.
19041         (nds32_expand_setmem_loop): New.
19042         (nds32_expand_setmem_loop_v3m): New.
19043         (nds32_expand_setmem_unroll): New.
19044         (nds32_expand_setmem): New.
19046 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
19047             Chung-Ju Wu  <jasonwucj@gmail.com>
19049         * config/nds32/nds32-memory-manipulation.c
19050         (nds32_emit_load_store): New.
19051         (nds32_emit_post_inc_load_store): New.
19052         (nds32_emit_mem_move): New.
19053         (nds32_emit_mem_move_block): New.
19054         (nds32_expand_movmemsi_loop_unknown_size): New.
19055         (nds32_expand_movmemsi_loop_known_size): New.
19056         (nds32_expand_movmemsi_loop): New.
19057         (nds32_expand_movmemsi_unroll): New.
19058         (nds32_expand_movmemqi): Rename ...
19059         (nds32_expand_movmemsi): ... to this.
19060         * config/nds32/nds32-multiple.md (movmemqi): Rename ...
19061         (movmemsi): ... to this.
19062         * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
19063         (nds32_expand_movmemsi): ... to this.
19065 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
19066             Monk Chiang  <sh.chiang04@gmail.com>
19067             Chung-Ju Wu  <jasonwucj@gmail.com>
19069         * config/nds32/nds32-protos.h
19070         (nds32_expand_load_multiple): New arguments.
19071         (nds32_expand_store_multiple): Ditto.
19072         (nds32_valid_multiple_load_store): Rename ...
19073         (nds32_valid_multiple_load_store_p): ... to this.
19074         * config/nds32/nds32-memory-manipulation.c
19075         (nds32_expand_load_multiple): Refine implementation.
19076         (nds32_expand_store_multiple): Ditto.
19077         * config/nds32/nds32-multiple.md
19078         (load_multiple): Update nds32_expand_load_multiple interface.
19079         (store_multiple): Update nds32_expand_store_multiple interface.
19080         * config/nds32/nds32-predicates.c
19081         (nds32_valid_multiple_load_store): Rename ...
19082         (nds32_valid_multiple_load_store_p): ... to this and refine
19083         implementation.
19084         * config/nds32/predicates.md
19085         (nds32_load_multiple_and_update_address_operation): New predicate.
19086         (nds32_store_multiple_and_update_address_operation): New predicate.
19088 2018-03-04  Kito Cheng  <kito.cheng@gmail.com>
19089             Chung-Ju Wu  <jasonwucj@gmail.com>
19091         * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
19092         (combo): New attribute.
19093         * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
19095 2018-03-03  Chung-Ju Wu  <jasonwucj@gmail.com>
19097         * config/nds32/nds32.opt: Change -mcmodel= default value.
19099 2018-03-03  Kito Cheng  <kito.cheng@gmail.com>
19100             Monk Chiang  <sh.chiang04@gmail.com>
19101             Chung-Ju Wu  <jasonwucj@gmail.com>
19103         * config/nds32/constants.md (unspec_element): New enum.
19104         * config/nds32/constraints.md (Umw): New constraint.
19105         * config/nds32/nds32-intrinsic.c: Add more builtin functions.
19106         * config/nds32/nds32-intrinsic.md: Likewise.
19107         * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
19108         (nds32_valid_smw_lwm_base_p): New.
19109         (nds32_output_smw_single_word): New.
19110         (nds32_output_lmw_single_word): New.
19111         (nds32_expand_unaligned_load): New.
19112         (nds32_expand_unaligned_store): New.
19113         * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
19114         (nds32_output_smw_single_word): Declare.
19115         (nds32_output_lmw_single_word): Declare.
19116         (nds32_expand_unaligned_load): Declare.
19117         (nds32_expand_unaligned_store): Declare.
19118         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
19119         NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
19120         NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
19121         NDS32_BUILTIN_UASTORE_DW.
19122         * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
19123         predicate.
19125 2018-03-03  Monk Chiang  <sh.chiang04@gmail.com>
19126             Kito Cheng  <kito.cheng@gmail.com>
19127             Chung-Ju Wu  <jasonwucj@gmail.com>
19129         * config/nds32/nds32-intrinsic.c
19130         (nds32_expand_builtin_null_ftype_reg): Delete.
19131         (nds32_expand_builtin_reg_ftype_imm): Ditto.
19132         (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
19133         (nds32_read_argument): New.
19134         (nds32_legitimize_target): Ditto.
19135         (nds32_legitimize_argument): Ditto.
19136         (nds32_check_constant_argument): Ditto.
19137         (nds32_expand_unop_builtin): Ditto.
19138         (nds32_expand_unopimm_builtin): Ditto.
19139         (nds32_expand_binop_builtin): Ditto.
19140         (nds32_builtin_decl_impl): Ditto.
19141         (builtin_description): Ditto.
19142         (nds32_expand_builtin_impl): Rewrite with new infrastructure.
19143         (nds32_init_builtins_impl): Ditto.
19144         * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
19145         (nds32_builtin_decl): New.
19146         * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
19147         * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
19149 2018-03-02  Jeff Law  <law@redhat.com>
19151         * reorg.c (stop_search_p): Handle DEBUG_INSN.
19152         (redundant_insn, fill_simple_delay_slots): Likewise.
19153         (fill_slots_from_thread): Likewise.
19154         * resource.c (mark_referenced_resources): Likewise.
19155         (mark_set_resources, find_dead_or_set_registers): Likewise.
19157 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
19159         * substring-locations.h (format_warning_va): Formatting fix for
19160         ATTRIBUTE_GCC_DIAG.
19161         (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
19162         argument.
19163         (format_warning_n_va, format_warning_at_substring_n): New prototypes.
19164         * substring-locations.c: Include intl.h.
19165         (format_warning_va): Turned into small wrapper around
19166         format_warning_n_va, renamed to ...
19167         (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
19168         rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
19169         use ngettext.
19170         (format_warning_at_substring_n): New function.
19171         * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
19172         (fmtwarn): Add ATTRIBUTE_GCC_DIAG.  Turn into a copy of
19173         format_warning_at_substring with just a shorter name instead of
19174         const function pointer.
19175         (fmtwarn_n): New function.
19176         (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
19177         appropriate, get rid of all the fmtstr temporaries, move conditionals
19178         with G_() wrapped string literals directly into fmtwarn arguments,
19179         cast dir.len to (int), formatting fixes.
19181 2018-03-02  Thomas Schwinge  <thomas@codesourcery.com>
19183         * doc/invoke.texi: Remove "Cilk Plus" references.
19185 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
19186             Richard Biener  <rguenther@suse.de>
19188         PR ipa/84628
19189         * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
19190         for error or warning attributes if CALL_FROM_THUNK_P is set.
19191         Formatting fixes.
19193 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
19195         PR target/56540
19196         * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
19197         __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
19199         PR target/56540
19200         * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
19201         __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
19203         * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
19204         instead of -1U in last predictors element's probability member.
19206 2018-03-02  Eric Botcazou  <ebotcazou@adacore.com>
19208         PR ipa/83983
19209         * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
19210         arguments if they are comparable.
19212 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
19214         PR tree-optimization/84634
19215         * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
19216         masks and masked_loop_p with a single loop_masks, making sure it's
19217         null for bb vectorization.
19219 2018-03-02  Richard Sandiford  <richard.sandiford@linaro.org>
19221         * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
19222         (vect_analyze_data_ref_access): Use loop->safe_len rather than
19223         loop->force_vectorize to check whether there is no alias.
19225 2018-03-02  Jakub Jelinek  <jakub@redhat.com>
19227         PR target/84614
19228         * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
19229         prototypes.
19230         * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
19231         comments.
19232         (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
19233         * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
19234         instead of a loop around prev_real_insn.
19235         * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
19236         prev_real_insn.
19238         PR inline-asm/84625
19239         * config/i386/i386.c (ix86_print_operand): Use conditional
19240         output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
19241         zero vector.
19243 2018-03-02  Richard Biener  <rguenther@suse.de>
19245         PR tree-optimization/84427
19246         * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
19247         (bitmap_set_subtract_values): Rewrite to handle multiple
19248         exprs per value.
19249         (clean): Likewise.
19250         (prune_clobbered_mems): Likewise.
19251         (phi_translate): Take edge instead of pred/phiblock.
19252         (phi_translate_1): Likewise.
19253         (phi_translate_set): Likewise.  Insert all translated
19254         exprs for a value into the set, keeping possibly multiple
19255         expressions per value.
19256         (compute_antic_aux): Adjust for phi_translate changes.
19257         When intersecting union the expressions and prune those
19258         not in the final value set, keeping possibly multiple
19259         expressions per value.  Do not use value-insertion
19260         for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
19261         all expressions.  Add verification that the value-sets
19262         only shrink during iteration.
19263         (compute_partial_antic_aux): Adjust for the phi_translate changes.
19264         (do_pre_regular_insertion): Likewise.
19265         (do_pre_partial_partial_insertion): Likewise.
19267 2018-03-02  Richard Biener  <rguenther@suse.de>
19269         PR target/82005
19270         * config/darwin.c (saved_debug_info_level): New static global.
19271         (darwin_asm_lto_start): Disable debug info generation for LTO out.
19272         (darwin_asm_lto_end): Restore debug info generation settings.
19274 2018-03-01  Martin Liska  <mliska@suse.cz>
19276         PR sanitizer/82484
19277         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
19278         volatile arguments.
19280 2018-03-01  Richard Biener  <rguenther@suse.de>
19282         PR debug/84645
19283         * dwarf2out.c (gen_variable_die): Properly handle late VLA
19284         type annotation with LTO when debug was disabled at compile-time.
19286 2018-03-01  Matthew Fortune  <mfortune@gmail.com>
19288         * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
19289         XINT with INTVAL.
19290         (mips_final_postscan_insn): Likewise.
19292 2018-03-01  Richard Sandiford  <richard.sandiford@linaro.org>
19294         PR rtl-optimization/84528
19295         * alias.c (init_alias_target): Add commentary.
19296         (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
19297         a unique base value if the frame pointer is not eliminated
19298         to the stack pointer.
19300 2018-03-01  Tom de Vries  <tom@codesourcery.com>
19302         PR rtl-optimization/83327
19303         * lra-int.h (hard_regs_spilled_into): Declare.
19304         * lra.c (hard_regs_spilled_into): Define.
19305         (init_reg_info): Init hard_regs_spilled_into.
19306         * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
19307         * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
19308         (process_bb_lives): Handle hard_regs_spilled_into.
19309         (lra_create_live_ranges_1): Before doing liveness propagation, clear
19310         regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
19312 2018-02-28  David Edelsohn  <dje.gcc@gmail.com>
19314         * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
19315         (powerpc-ibm-aix[789]*): Default to AIX 7.2.
19316         * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
19317         * config/rs6000/aix72.h: New file.
19319 2018-02-28  Jakub Jelinek  <jakub@redhat.com>
19321         * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
19322         instead of warning_at with conditional singular and plural messages
19323         where possible.
19325         PR target/52991
19326         * stor-layout.c (update_alignment_for_field): For
19327         targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
19328         && !DECL_PACKED (field), do the alignment update, just use
19329         only desired_align instead of MAX (type_align, desired_align)
19330         as the alignment.
19331         (place_field): Don't do known_align < desired_align handling
19332         early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
19333         is non-NULL, instead do it after rli->prev_field handling and
19334         only if not within a bitfield word.  For DECL_PACKED (field)
19335         use type_align of BITS_PER_UNIT.
19337 2018-02-28  Eric Botcazou  <ebotcazou@adacore.com>
19339         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
19340         superfluous parentheses and trailing spaces.
19342 2018-02-28  Richard Biener  <rguenther@suse.de>
19344         PR tree-optimization/84584
19345         * graphite-scop-detection.c (scop_detection::add_scop): Discard
19346         SCoPs with fake exit edge.
19348 2018-02-28  Martin Liska  <mliska@suse.cz>
19350         PR testsuite/84597
19351         * timevar.c (timer::print): Fix format to properly print 100%
19352         values.
19354 2018-02-28  Richard Biener  <rguenther@suse.de>
19356         PR middle-end/84607
19357         * genmatch.c (capture_info::walk_match): Do not mark
19358         captured expressions without operands as expr_p given
19359         they act more like predicates and should be subject to
19360         "lost tail" side-effect preserving.
19362 2018-02-28  Alexandre Oliva  <aoliva@redhat.com>
19364         PR rtl-optimization/81611
19365         * auto-inc-dec.c (attempt_change): Move dead note from
19366         mem_insn if it's the next use of regno
19367         (find_address): Take address use of reg holding
19368         non-incremented value.  Add parm to limit search to the named
19369         reg only.
19370         (merge_in_block): Attempt to use a mem insn that is the next
19371         use of the original regno.
19373 2018-02-27  Martin Sebor  <msebor@redhat.com>
19375         PR c++/83871
19376         * doc/invoke.texi (-Wmissing-attributes): New option.
19377         * print-tree.c (print_node): Handle DECL_UNINLINABLE.
19379 2018-02-27  Martin Sebor  <msebor@redhat.com>
19381         PR translation/84207
19382         * diagnostic-core.h (warning_n, error_n, inform_n): Change
19383         n argument to unsigned HOST_WIDE_INT.
19384         * diagnostic.c (warning_n, error_n, inform_n): Ditto.
19385         (diagnostic_n_impl): Ditto.  Handle arguments in excess of LONG_MAX.
19386         * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
19387         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
19389 2018-02-27  Richard Biener  <rguenther@suse.de>
19391         PR tree-optimization/84512
19392         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
19393         Do not use the estimate returned from record_stmt_cost for
19394         the scalar iteration cost but sum properly using add_stmt_cost.
19396 2018-02-27  Richard Biener  <rguenther@suse.de>
19398         PR tree-optimization/84466
19399         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
19400         Adjust last change to less strictly validate use operands.
19402 2018-02-27  Martin Liska  <mliska@suse.cz>
19404         PR gcov-profile/84548
19405         * gcov.c (process_file): Allow partial overlap and consider it
19406         also as group functions.
19407         (output_lines): Properly calculate range of lines for a group.
19409 2018-02-27  Martin Liska  <mliska@suse.cz>
19411         * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
19412         'ggc' suffixes.  Change first column width.
19413         (timer::print): Fix formatting of the column.
19415 2018-02-27  Alexandre Oliva  <aoliva@redhat.com>
19417         * tree-ssa-live.c (remove_unused_scope_block_p): Do not
19418         preserve inline entry blocks for the sake of debug inline
19419         entry point markers alone.
19420         (remove_unused_locals): Suggest in comments a better place to
19421         force the preservation of inline entry blocks that are
19422         otherwise unused, but do not preserve them.
19424 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
19426         * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
19428 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
19430         PR target/84039
19431         * config/i386/constraints.md (Bs): Replace
19432         ix86_indirect_branch_register with
19433         TARGET_INDIRECT_BRANCH_REGISTER.
19434         (Bw): Likewise.
19435         * config/i386/i386.md (indirect_jump): Likewise.
19436         (tablejump): Likewise.
19437         (*sibcall_memory): Likewise.
19438         (*sibcall_value_memory): Likewise.
19439         Peepholes of indirect call and jump via memory: Likewise.
19440         (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
19441         (*sibcall_value_GOT_32): Likewise.
19442         * config/i386/predicates.md (indirect_branch_operand): Likewise.
19443         (GOT_memory_operand): Likewise.
19444         (call_insn_operand): Likewise.
19445         (sibcall_insn_operand): Likewise.
19446         (GOT32_symbol_operand): Likewise.
19447         * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
19449 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
19451         PR rtl-optimization/83496
19452         * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
19453         booleans to RTXes.  Call fix_reg_dead_note on every non-null element.
19454         (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
19455         redundant insn, if any.
19456         (relax_delay_slots): Likewise.
19457         (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
19459 2018-02-26  Richard Sandiford  <richard.sandiford@linaro.org>
19461         PR tree-optimization/83965
19462         * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
19463         that grouped statements are part of a reduction chain.  Return
19464         true if the statement is not marked as a reduction itself but
19465         is part of a group.
19466         (vect_recog_dot_prod_pattern): Don't check whether the statement
19467         is part of a group here.
19468         (vect_recog_sad_pattern): Likewise.
19469         (vect_recog_widen_sum_pattern): Likewise.
19471 2018-02-26  Eric Botcazou  <ebotcazou@adacore.com>
19473         PR debug/84545
19474         * final.c (rest_of_clean_state): Also look for calls inside sequences.
19476 2018-02-26  H.J. Lu  <hongjiu.lu@intel.com>
19478         PR target/84530
19479         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
19480         the bool argument.
19481         (ix86_output_indirect_function_return): New prototype.
19482         (ix86_split_simple_return_pop_internal): Likewise.
19483         * config/i386/i386.c (indirect_return_via_cx): New.
19484         (indirect_return_via_cx_bnd): Likewise.
19485         (indirect_thunk_name): Handle return va CX_REG.
19486         (output_indirect_thunk_function): Create alias for
19487         __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
19488         (ix86_output_indirect_jmp): Remove the bool argument.
19489         (ix86_output_indirect_function_return): New function.
19490         (ix86_split_simple_return_pop_internal): Likewise.
19491         * config/i386/i386.md (*indirect_jump): Don't pass false
19492         to ix86_output_indirect_jmp.
19493         (*tablejump_1): Likewise.
19494         (simple_return_pop_internal): Change it to define_insn_and_split.
19495         Call ix86_split_simple_return_pop_internal to split it for
19496         -mfunction-return=.
19497         (simple_return_indirect_internal): Call
19498         ix86_output_indirect_function_return instead of
19499         ix86_output_indirect_jmp.
19501 2018-02-26  Jakub Jelinek  <jakub@redhat.com>
19503         PR bootstrap/84405
19504         * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
19505         memset and value initialization afterwards.
19507 2018-02-26  Christophe Lyon  <christophe.lyon@linaro.org>
19509         * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
19511 2018-02-26  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
19513         PR target/84521
19514         * common/config/aarch64/aarch64-common.c
19515         (aarch_option_optimization_table[]): Switch
19516         off fomit-frame-pointer
19518 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
19519             Chung-Ju Wu  <jasonwucj@gmail.com>
19521         * config/nds32/nds32-multiple.md (load_multiple): Disallow
19522         volatile memory.
19523         (store_multiple): Ditto.
19525 2018-02-26  Kito Cheng  <kito.cheng@gmail.com>
19527         * config.gcc: Add --with-cpu support for nds32 target.
19528         * config/nds32/nds32-opts.h (nds32_cpu_type): New.
19529         * config/nds32/nds32.opt: Add -mcpu= option.
19531 2018-02-25  Segher Boessenkool  <segher@kernel.crashing.org>
19533         * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
19534         isel=yes): Warn for these deprecated options.
19536 2018-02-23  David Edelsohn  <dje.gcc@gmail.com>
19538         * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
19539         ISA_2_5_MASKS_EMBEDDED.
19541 2018-02-23  Jakub Jelinek  <jakub@redhat.com>
19543         * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
19544         p->max as pointers rather than using iterative_hash_expr.
19546 2018-02-23  Carl Love  <cel@us.ibm.com>
19548         * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
19549         macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
19550         BU_P8V_OVERLOAD_2.
19551         * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
19552         P8V_BUILTIN_VEC_VSIGNED2.  Change VSX_BUILTIN_VEC_VUNSIGNED2 to
19553         P8V_BUILTIN_VEC_VUNSIGNED2.
19555 2018-02-22  Vladimir Makarov  <vmakarov@redhat.com>
19557         PR target/81572
19558         * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
19559         * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
19560         LRA_UNKNOWN_ALT.
19561         * lra-constraints.c (curr_insn_transform): Set up
19562         LRA_NON_CLOBBERED_ALT for moves processed on the fast path.  Use
19563         LRA_UNKNOWN_ALT.
19564         (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
19565         * lra-eliminations.c (spill_pseudos): Ditto.
19566         (process_insn_for_elimination): Ditto.
19567         * lra-lives.c (reg_early_clobber_p): Use the new macros.
19568         * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
19569         LRA_NON_CLOBBERED_ALT.
19571 2018-02-22  Martin Sebor  <msebor@redhat.com>
19573         PR tree-optimization/84480
19574         * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
19575         to maybe_diag_stxncpy_trunc.  Call it.
19576         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
19577         from gimple_fold_builtin_strcpy.  Print inlining stack.
19578         (handle_builtin_stxncpy): Print inlining stack.
19579         * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
19581 2018-02-22  H.J. Lu  <hongjiu.lu@intel.com>
19583         PR target/84176
19584         * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
19585         error when -mindirect-branch=thunk-extern, -fcf-protection=branch
19586         and -fcheck-pointer-bounds are used together.
19587         (indirect_thunk_prefix): New enum.
19588         (indirect_thunk_need_prefix): New function.
19589         (indirect_thunk_name): Replace need_bnd_p with need_prefix.  Use
19590         "_nt" instead of "_bnd" for NOTRACK prefix.
19591         (output_indirect_thunk): Replace need_bnd_p with need_prefix.
19592         (output_indirect_thunk_function): Likewise.
19593         (): Likewise.
19594         (ix86_code_end): Update output_indirect_thunk_function calls.
19595         (ix86_output_indirect_branch_via_reg): Replace
19596         ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
19597         (ix86_output_indirect_branch_via_push): Likewise.
19598         (ix86_output_function_return): Likewise.
19599         * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
19600         incompatible with -fcf-protection=branch and
19601         -fcheck-pointer-bounds.
19603 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
19605         PR target/83335
19606         * config/aarch64/aarch64.c (aarch64_print_address_internal):
19607         Change gcc_assert call to output_operand_lossage.
19609 2018-02-22  Steve Ellcey  <sellcey@cavium.com>
19611         * doc/extend.texi (__builtin_extend_pointer): Document builtin.
19613 2018-02-22  DJ Delorie  <dj@redhat.com>
19614             Sebastian Perta  <sebastian.perta@renesas.com>
19615             Oleg Endo  <olegendo@gcc.gnu.org>
19617         * config/rx/rx.c (rx_rtx_costs): New function.
19618         (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
19620 2018-02-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
19622         * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
19624 2018-02-22  Martin Liska  <mliska@suse.cz>
19626         PR driver/83193
19627         * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
19628         Add "native" as a possible value.
19630 2018-02-22  Martin Liska  <mliska@suse.cz>
19632         PR driver/83193
19633         * config/i386/i386.c (ix86_option_override_internal):
19634         Add "native" as a possible value for -march and -mtune.
19636 2018-02-22  Jakub Jelinek  <jakub@redhat.com>
19638         PR target/84502
19639         * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
19640         to all type variants.
19642         PR tree-optimization/84503
19643         * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
19644         width as info->bitpos + info->bitsize - start.
19645         (merged_store_group::merge_overlapping): Simplify width computation.
19646         (check_no_overlap): New function.
19647         (imm_store_chain_info::try_coalesce_bswap): Compute expected
19648         start + width and last_order of the group, fail if check_no_overlap
19649         fails.
19650         (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
19651         to group if check_no_overlap fails.
19653 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
19655         * config/rs6000/altivec.md: Delete contraint arguments to
19656         define_expand, define_split, and define_peephole2, and in
19657         define_insn_and_split if always unused.
19658         * config/rs6000/darwin.md: Ditto.
19659         * config/rs6000/dfp.md: Ditto.
19660         * config/rs6000/rs6000.md: Ditto.
19661         * config/rs6000/sync.md: Ditto.
19662         * config/rs6000/vector.md: Ditto.
19663         * config/rs6000/vsx.md: Ditto.
19665 2018-02-21  Segher Boessenkool  <segher@kernel.crashing.org>
19667         * config/rs6000/altivec.md: Write output control strings as braced
19668         blocks instead of double-quoted strings.
19669         * config/rs6000/darwin.md: Ditto.
19670         * config/rs6000/rs6000.md: Ditto.
19671         * config/rs6000/vector.md: Ditto.
19672         * config/rs6000/vsx.md: Ditto.
19674 2018-02-21  Jason Merrill  <jason@redhat.com>
19676         PR c++/84314 - ICE with templates and fastcall attribute.
19677         * attribs.c (build_type_attribute_qual_variant): Remove assert.
19679 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
19681         * ipa-cp.c (determine_versionability): Fix comment typos.
19683 2018-02-21  Jan Hubicka  <hubicka@ucw.cz>
19685         PR c/84229
19686         * ipa-cp.c (determine_versionability): Do not version functions caling
19687         va_arg_pack.
19689 2018-02-21  Martin Liska  <mliska@suse.cz>
19691         PR driver/83193
19692         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
19693         Add "native" as a possible value.
19694         * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT):  Define
19695         the macro when native cpu detection is available.
19697 2018-02-21  Martin Liska  <mliska@suse.cz>
19699         PR driver/83193
19700         * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
19701         Add "native" as a possible value.
19702         * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
19703         when native cpu detection is available.
19705 2018-02-21  Jakub Jelinek  <jakub@redhat.com>
19706             Martin Sebor  <msebor@redhat.com>
19708         PR tree-optimization/84478
19709         * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
19710         false.
19711         * gimple-fold.c (get_range_strlen): Make minlen const and assume it
19712         can't be NULL.  Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
19713         support which is conservatively correct, for 2 only stay conservative
19714         for maxlen.  Formatting and comment capitalization fixes.  Add STRICT
19715         argument to the 2 argument get_range_strlen, adjust 6 arg
19716         get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
19717         false.
19718         (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
19719         (gimple_fold_builtin_strlen): Pass true as last argument to
19720         get_range_strlen.
19722 2018-02-20  Martin Sebor  <msebor@redhat.com>
19724         PR middle-end/84095
19725         * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
19726         (builtin_memref::set_base_and_offset): Same.  Handle inner references.
19727         (builtin_memref::builtin_memref): Factor out parts into
19728         set_base_and_offset and call it.
19730 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
19732         PR middle-end/84406
19733         * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
19734         is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
19735         greater precision.  If to_mode is a MODE_PARTIAL_INT, stop the
19736         search at the associated MODE_INT.
19738 2018-02-20  Jeff Law  <law@redhat.com>
19740         PR middle-end/82123
19741         PR tree-optimization/81592
19742         PR middle-end/79257
19743         * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
19744         for range data rather than using global data.
19745         * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
19746         range data rather than using global data.
19747         * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
19748         pass it to children as needed.
19749         (struct directive::fmtresult): Similarly.
19750         (struct directive::set_width): Similarly.
19751         (struct directive::set_precision): Similarly.
19752         (format_integer, format_directive, parse_directive): Similarly.
19753         (format_none): Accept unnamed vr_values parameter.
19754         (format_percent, format_floating, format_character): Similarly.
19755         (format_string, format_plain): Similarly.
19756         * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
19757         the EVRP range analyzer for range data rather than using global data.
19758         * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
19759         gimple-ssa-evrp-analyze.h
19760         (class sprintf_dom_walker): Add after_dom_children member function.
19761         Add evrp_range_analyzer member.
19762         (sprintf_dom_walker::before_dom_children): Call into the EVRP
19763         range analyzer as needed.
19764         (sprintf_dom_walker::after_dom_children): New member function.
19765         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
19766         if not optimizing.
19767         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
19768         (evrp_range_analyzer::pop_to_marker): Likewise.
19770 2018-02-20  Richard Sandiford  <richard.sandiford@linaro.org>
19772         PR tree-optimization/84419
19773         * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
19774         with the required type if its current type is compatible but
19775         different.
19777 2018-02-20  Jakub Jelinek  <jakub@redhat.com>
19779         PR middle-end/82004
19780         * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
19781         after vectorization.
19783 2018-02-20  Martin Liska  <mliska@suse.cz>
19785         PR driver/83193
19786         * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
19787         possible values if we don't have a hint.
19789 2018-02-20  Martin Liska  <mliska@suse.cz>
19791         PR c/84310
19792         PR target/79747
19793         * final.c (shorten_branches): Build align_tab array with one
19794         more element.
19795         * opts.c (finish_options): Add alignment option limit check.
19796         (MAX_CODE_ALIGN): Likewise.
19797         (MAX_CODE_ALIGN_VALUE): Likewise.
19798         * doc/invoke.texi: Document maximum allowed option value for
19799         all -falign-* options.
19801 2018-02-19  Jakub Jelinek  <jakub@redhat.com>
19803         PR target/84146
19804         * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
19805         * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
19806         * var-tracking.c (emit_note_insn_var_location): Remove all references
19807         to NOTE_INSN_CALL_ARG_LOCATION.
19808         (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
19809         the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
19810         Use copy_rtx_if_shared.
19811         * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
19812         NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
19813         (dwarf2out_var_location): Remove handling of
19814         NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
19815         on call_insn.
19816         * final.c (final_scan_insn): Remove all references to
19817         NOTE_INSN_CALL_ARG_LOCATION.
19818         (rest_of_clean_state): Likewise.  Remove REG_CALL_ARG_LOCATION notes
19819         before dumping final insns.
19820         * except.c (emit_note_eh_region_end): Remove all references to
19821         NOTE_INSN_CALL_ARG_LOCATION.
19822         * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
19823         * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
19824         * config/arc/arc.c (hwloop_optimize): Likewise.
19825         * config/arm/arm.c (create_fix_barrier): Likewise.
19826         * config/s390/s390.c (s390_chunkify_start): Likewise.
19827         * config/sh/sh.c (find_barrier): Likewise.
19828         * config/i386/i386.c (rest_of_insert_endbranch,
19829         ix86_seh_fixup_eh_fallthru): Likewise.
19830         * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
19831         * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
19832         * config/frv/frv.c (frv_function_prologue): Likewise.
19833         * emit-rtl.c (try_split): Likewise.  Copy over REG_CALL_ARG_LOCATION
19834         reg note.
19835         (note_outside_basic_block_p): Remove all references to
19836         NOTE_INSN_CALL_ARG_LOCATION.
19837         * gengtype.c (adjust_field_rtx_def): Likewise.
19838         * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
19839         Likewise.
19840         * jump.c (cleanup_barriers, delete_related_insns): Likewise.
19841         * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
19843         PR c++/84444
19844         * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
19845         is ADDR_EXPR.
19847         PR tree-optimization/84452
19848         * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
19849         expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
19850         is NULL.
19852 2018-02-19  Martin Liska  <mliska@suse.cz>
19854         PR sanitizer/82183
19855         * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
19857 2018-02-19  Martin Liska  <mliska@suse.cz>
19858             Richard Sandiford  <richard.sandiford@linaro.org>
19860         PR tree-optimization/82491
19861         * gimple-fold.c (get_base_constructor): Make earlier bail out
19862         to prevent ubsan.
19864 2018-02-19  Carl Love  <cel@us.ibm.com>
19866         * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
19867         BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
19868         BU_P8V_OVERLOAD_1.
19869         * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
19870         P8V_BUILTIN_VEC_NEG.
19872 2018-02-19  Sebastian Perta  <sebastian.perta@renesas.com>
19874         * config/rl78/rl78.md (movdf): New define expand.
19876 2018-02-19  Martin Liska  <mliska@suse.cz>
19878         PR other/80589
19879         * doc/invoke.texi: Fix typo.
19880         * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
19882 2018-02-18  Segher Boessenkool  <segher@kernel.crashing.org>
19884         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
19885         handle rs6000_single_float and rs6000_double_float specially for
19886         e500 family CPUs.
19888 2018-02-16  Jeff Law  <law@redhat.com>
19890         * config/rx/rx.c (add_pop_cfi_notes): New function.;
19891         (pop_regs): Use it.
19893 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
19895         PR ipa/84425
19896         * ipa-inline.c (inline_small_functions): Fix a typo.
19898 2018-02-16  Nathan Sidwell  <nathan@acm.org>
19900         * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
19902 2018-02-16  Carl Love  <cel@us.ibm.com>
19904         * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
19905         Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
19906         from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
19907         * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
19908         expansion to P8V_BUILTIN_VEC_FLOAT2.
19910 2018-02-16  Vladimir Makarov  <vmakarov@redhat.com>
19912         PR rtl-optimization/70023
19913         * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
19914         src_regno into account.
19916 2018-02-16  Carl Love  <cel@us.ibm.com>
19918         * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
19919         * config/rs6000/rs6000-builtin.def: Remove macro expansion for
19920         VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
19921         * config/rs6000/rs6000.c: Remove case statements for
19922         P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
19923         P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
19924         and P9V_BUILTIN_VEC_VINSERT4B.
19925         * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
19926         P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
19927         * config/rs6000/vsx.md:
19928         * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
19929         vec_insert4b.
19931 2018-02-16  Carl Love  <cel@us.ibm.com>
19933         * config/rs6000/altivec.h: Add builtin names vec_extract4b
19934         vec_insert4b.
19935         * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
19936         definitions.
19937         * config/rs6000/rs6000-c.c: Add the definitions for
19938         P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
19939         * config/rs6000/rs6000.c (altivec_expand_builtin): Add
19940         P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
19941         * config/rs6000/vsx.md: Add define_insn extract4b.  Add define_expand
19942         definition for insert4b and define insn *insert3b_internal.
19943         * doc/extend.texi: Add documentation for vec_extract4b.
19945 2018-02-16  Nathan Sidwell  <nathan@acm.org>
19947         * doc/extend.texi (Backwards Compatibility): Mention friend
19948         injection.  Note for-scope is deprecated.
19949         * doc/invoke.texi (-ffriend-injection): Deprecate.
19951 2018-02-16  Segher Boessenkool  <segher@kernel.crashing.org>
19953         * combine.c (try_combine): When adjusting LOG_LINKS for the destination
19954         that moved to I2, also allow destinations that are a paradoxical
19955         subreg (instead of a normal reg).
19957 2018-02-16  Oleg Endo  <olegendo@gcc.gnu.org>
19959         PR target/83831
19960         * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
19961         to QImode.
19963 2018-02-16  Richard Biener  <rguenther@suse.de>
19965         PR tree-optimization/84037
19966         PR tree-optimization/84016
19967         PR target/82862
19968         * config/i386/i386.c (ix86_builtin_vectorization_cost):
19969         Adjust vec_construct for the fact we need additional higher latency
19970         128bit inserts for AVX256 and AVX512 vector builds.
19971         (ix86_add_stmt_cost): Scale vector construction cost for
19972         elementwise loads.
19974 2018-02-16  Richard Biener  <rguenther@suse.de>
19976         PR tree-optimization/84417
19977         * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
19978         the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
19979         (non_rewritable_lvalue_p): Likewise, use poly-ints.
19981 2018-02-16  Martin Liska  <mliska@suse.cz>
19983         PR sanitizer/84307
19984         * internal-fn.def (ASAN_CHECK): Set proper flags.
19985         (ASAN_MARK): Likewise.
19987 2018-02-16  Julia Koval  <julia.koval@intel.com>
19989         * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
19990         from PTA_CANNONLAKE.
19992 2018-02-16  Jakub Jelinek  <jakub@redhat.com>
19994         PR target/84272
19995         * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
19996         Use ++iter rather than iter++ for std::list iterators.
19997         (func_fma_steering::dfs): Likewise.  Don't delete nodes right away,
19998         defer deleting them until all nodes in the forest are processed.  Do
19999         free even leaf nodes.  Change to_process into auto_vec.
20001         PR bootstrap/84405
20002         * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
20003         * vec.h (vec_default_construct): Use memset instead of placement new
20004         if BROKEN_VALUE_INITIALIZATION is defined.
20005         * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
20006         memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
20007         is defined.
20009         PR rtl-optimization/83723
20010         * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
20011         * lra.c (lra_substitute_pseudo): Likewise.  If true, use
20012         gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG.  Pass DEBUG_P to
20013         recursive calls.
20014         (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
20015         callers.
20016         * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
20018 2018-02-16  Eric Botcazou  <ebotcazou@adacore.com>
20020         PR rtl-optimization/81443
20021         * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
20022         from inner REGs to paradoxical SUBREGs.
20024 2018-02-16  Richard Biener  <rguenther@suse.de>
20026         PR tree-optimization/84399
20027         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
20028         For operands we can analyze at their definition make sure we can
20029         analyze them at each use as well.
20031 2018-02-16  Richard Biener  <rguenther@suse.de>
20033         PR tree-optimization/84190
20034         * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
20035         volatile accesses if the decl isn't volatile.
20037 2018-02-15  Jason Merrill  <jason@redhat.com>
20039         PR c++/84314 - ICE with templates and fastcall attribute.
20040         * attribs.c (build_type_attribute_qual_variant): Don't clobber
20041         TYPE_CANONICAL on an existing type.
20043 2018-02-15  Jakub Jelinek  <jakub@redhat.com>
20045         PR tree-optimization/84383
20046         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
20047         dstoff nor call operand_equal_p if dstbase is NULL.
20049         PR tree-optimization/84334
20050         * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
20051         also a CONSTANT_CLASS_P, punt.
20053 2018-02-14  Jim Wilson  <jimw@sifive.com>
20055         * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
20056         first SMALL_OPERAND check.  New local min_second_step.  Move assert
20057         to where locals are set.  Add TARGET_RVC support.
20058         * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
20060 2018-02-14  Indu Bhagat  <indu.bhagat@oracle.com>
20062         * doc/invoke.texi: Correct -Wformat-overflow code sample.
20064 2018-02-14  Martin Sebor  <msebor@redhat.com>
20066         PR tree-optimization/83698
20067         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
20068         arrays constrain the offset range to their bounds.
20069         (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
20070         (builtin_access::overlap): Avoid setting the size of overlap if it's
20071         already been set.
20072         (maybe_diag_overlap): Also consider arrays when deciding what values
20073         of offsets to include in diagnostics.
20075 2018-02-14  Martin Sebor  <msebor@redhat.com>
20077         PR c/84108
20078         * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
20079         that correspond to the kind of a declaration.
20081 2018-02-14  John David Anglin  <danglin@gcc.gnu.org>
20083         PR target/83984
20084         * config/pa/pa.md: Load address of PIC label using the linkage table
20085         if the label is nonlocal.
20087 2018-02-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
20089         * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
20090         warning message if user requests -maltivec=be.
20091         * doc/invoke.texi: Document deprecation of -maltivec=be.
20093 2018-02-14  Will Schmidt  <will_schmidt@vnet.ibm.com>
20095         PR target/84220
20096         * config/rs6000/rs6000-c.c: Update definitions for
20097         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
20098         VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
20100 2018-02-14  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
20102         PR target/84239
20103         * config/i386/cetintrin.h: Remove _rdssp[d|q] and
20104         add _get_ssp intrinsics. Remove argument from
20105         __builtin_ia32_rdssp[d|q].
20106         * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
20107         * config/i386/i386-builtin.def: Remove argument from
20108         __builtin_ia32_rdssp[d|q].
20109         * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
20110         ix86_expand_special_args_builtin for _rdssp[d|q].
20111         * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
20112         Clear register before usage.
20113         * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
20114         Add documentation for new _get_ssp and _inc_ssp intrinsics.
20116 2018-02-14  Richard Sandiford  <richard.sandiford@linaro.org>
20118         PR tree-optimization/84357
20119         * tree-data-ref.c (object_address_invariant_in_loop_p): Check
20120         operand 1 of an ARRAY_REF too.
20122 2018-02-14  Oleg Endo  <olegendo@gcc.gnu.org>
20124         PR target/83831
20125         * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
20126         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20127         declarations.
20128         (set_of_reg): New struct.
20129         (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
20130         * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
20131         rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
20132         functions.
20133         * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
20134         Split into bitclr, bitset, bitinvert patterns if appropriate.
20135         (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
20136         use rx_fuse_in_memory_bitop.
20137         (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
20138         to named insn, correct maximum insn length.
20140 2018-02-14  Jozef Lawrynowicz  <jozefl.gcc@gmail.com>
20142         PR target/79242
20143         * machmode.def: Define a complex mode for PARTIAL_INT.
20144         * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
20145         MODE_PARTIAL_INT.
20146         * doc/rtl.texi: Document CSPImode.
20147         * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
20148         handling.
20149         (msp430_hard_regno_nregs_with_padding): Likewise.
20151 2018-02-13  Peter Bergner  <bergner@vnet.ibm.com>
20153         PR target/84279
20154         * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
20156 2018-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
20158         PR rtl-optimization/84169
20159         * combine.c (try_combine): New variable split_i2i3.  Set it to true if
20160         we generated a parallel as new i3 and we split that to new i2 and i3
20161         instructions.  Handle split_i2i3 similar to swap_i2i3: scan the
20162         LOG_LINKs of i3 to see which of those need to link to i2 now.  Link
20163         those to i2, not i1.  Partially rewrite this scan code.
20165 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
20167         PR c/82210
20168         * stor-layout.c (place_field): For variable length fields, adjust
20169         offset_align afterwards not just based on the field's alignment,
20170         but also on the size.
20172         PR middle-end/84309
20173         * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
20174         of exps and logs in the use_exp2 case.
20176 2018-02-13  Jeff Law  <law@redhat.com>
20178         * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
20179         entry for "vector".
20181         * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
20182         ARGS as unused.
20184 2018-02-13  Alexandre Oliva  <aoliva@redhat.com>
20186         PR debug/84342
20187         PR debug/84319
20188         * common.opt (gas-loc-support, gas-locview-support): New.
20189         (ginline-points, ginternal-reset-location-views): New.
20190         * doc/invoke.texi: Document them.  Use @itemx where intended.
20191         (gvariable-location-views): Adjust.
20192         * target.def (reset_location_view): New.
20193         * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
20194         (TARGET_RESET_LOCATION_VIEW): New.
20195         * doc/tm.texi: Rebuilt.
20196         * dwarf2out.c (dwarf2out_default_as_loc_support): New.
20197         (dwarf2out_default_as_locview_support): New.
20198         (output_asm_line_debug_info): Use option variables.
20199         (dwarf2out_maybe_output_loclist_view_pair): Likewise.
20200         (output_loc_list): Likewise.
20201         (add_high_low_attributes): Check option variables.
20202         Don't output entry view attribute in strict mode.
20203         (gen_inlined_subroutine_die): Check option variables.
20204         (dwarf2out_inline_entry): Likewise.
20205         (init_sections_and_labels): Likewise.
20206         (dwarf2out_early_finish): Likewise.
20207         (maybe_reset_location_view): New, from...
20208         (dwarf2out_var_location): ... here.  Call it.
20209         * debug.h (dwarf2out_default_as_loc_support): Declare.
20210         (dwarf2out_default_as_locview_support): Declare.
20211         * hooks.c (hook_int_rtx_insn_0): New.
20212         * hooks.h (hook_int_rtx_insn_0): Declare.
20213         * toplev.c (process_options): Take -gas-loc-support and
20214         -gas-locview-support from dwarf2out.  Enable
20215         -gvariable-location-views by default only with locview
20216         assembler support.  Enable -ginternal-reset-location-views by
20217         default only if the target defines the corresponding hook.
20218         Enable -ginline-points by default if location views are
20219         enabled; force it disabled if statement frontiers are
20220         disabled.
20221         * tree-inline.c (expand_call_inline): Check option variables.
20222         * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
20224 2018-02-13  Richard Sandiford  <richard.sandiford@linaro.org>
20226         PR tree-optimization/84321
20227         * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
20228         handling.  Also check whether the anti-range contains any values
20229         that satisfy the mask; switch to a VR_RANGE if not.
20231 2018-02-13  Paolo Bonzini  <bonzini@gnu.org>
20233         PR sanitizer/84340
20234         * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
20236 2018-02-13  Martin Jambor  <mjambor@suse.cz>
20238         PR c++/83990
20239         * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
20240         of call statements, also set location of a load to a temporary.
20242 2018-02-13  Sebastian Perta  <sebastian.perta@renesas.com>
20244         * config/rl78/rl78.c (add_vector_labels): New function.
20245         * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
20246         * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
20247         * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
20248         which checks that no arguments are passed.
20249         * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
20250         * doc/extend.texi: Documentation for the new attribute.
20252 2018-02-13  Andreas Schwab  <schwab@suse.de>
20254         * config/riscv/linux.h (CPP_SPEC): Define.
20256 2018-02-13  Jakub Jelinek  <jakub@redhat.com>
20258         PR target/84335
20259         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
20260         OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
20261         OPTION_MASK_ISA_AES as first argument to def_builtin_const
20262         for AES builtins.  Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
20263         instead of OPTION_MASK_ISA_PCLMUL as first argument to
20264         def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
20265         * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
20266         temporarily for AES and PCLMUL builtins.
20268         PR tree-optimization/84339
20269         * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
20270         ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
20271         Formatting fixes.
20273         PR middle-end/84309
20274         * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
20275         exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
20276         * generic-match-head.c (canonicalize_math_after_vectorization_p): New
20277         inline function.
20278         * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
20279         inline function.
20280         * omp-simd-clone.h: New file.
20281         * omp-simd-clone.c: Include omp-simd-clone.h.
20282         (expand_simd_clones): No longer static.
20283         * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
20284         cgraph.h and omp-simd-clone.h.
20285         (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
20286         (vect_recog_widen_shift_pattern): Formatting fix.
20287         (vect_pattern_recog_1): Don't check optab for calls.
20289         PR target/84336
20290         * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
20291         operands[2] into a REG before using gen_lowpart on it.
20293 2018-02-12  Jeff Law  <law@redhat.com>
20295         PR target/83760
20296         * config/sh/sh.c (find_barrier): Consider a sibling call
20297         a barrier as well.
20299         * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
20300         successfully back substituting a reg.
20302 2018-02-12  Richard Biener  <rguenther@suse.de>
20304         PR tree-optimization/84037
20305         * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
20306         parameter, move visited init to caller.
20307         (vect_slp_analyze_operations): Separate cost from validity
20308         check, initialize visited once for all instances.
20309         (vect_schedule_slp): Analyze map to CSE vectorized nodes once
20310         for all instances.
20311         * tree-vect-stmts.c (vect_model_simple_cost): Make early
20312         out an assert.
20313         (vect_model_promotion_demotion_cost): Likewise.
20314         (vectorizable_bswap): Guard cost modeling with !slp_node
20315         instead of !PURE_SLP_STMT to avoid double-counting on hybrid
20316         SLP stmts.
20317         (vectorizable_call): Likewise.
20318         (vectorizable_conversion): Likewise.
20319         (vectorizable_assignment): Likewise.
20320         (vectorizable_shift): Likewise.
20321         (vectorizable_operation): Likewise.
20322         (vectorizable_store): Likewise.
20323         (vectorizable_load): Likewise.
20324         (vectorizable_condition): Likewise.
20325         (vectorizable_comparison): Likewise.
20327 2018-02-12  Paolo Bonzini  <bonzini@gnu.org>
20329         PR sanitizer/84307
20330         * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
20331         (ASAN_MARK): Fix fnspec to account for return value, change pointer
20332         argument from 'R' to 'W' so that the pointed-to datum is clobbered.
20334 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
20336         PR middle-end/83665
20337         * params.def (inline-min-speedup): Increase from 8 to 15.
20338         (max-inline-insns-auto): Decrease from 40 to 30.
20339         * ipa-split.c (consider_split): Add some buffer for function to
20340         be considered inlining candidate.
20341         * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
20342         default values.
20344 2018-02-12  Richard Biener  <rguenther@suse.de>
20346         PR tree-optimization/84037
20347         * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
20348         matched stmts if we cannot swap the non-matched ones.
20350 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
20352         * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
20353         _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
20354         _mm_maskz_scalef_round_ss): New intrinsics.
20355         (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
20356         * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
20357         __builtin_ia32_scalefss_round): Remove.
20358         (__builtin_ia32_scalefsd_mask_round,
20359         __builtin_ia32_scalefss_mask_round): New intrinsics.
20360         * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
20361         (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
20362         ((match_operand:VF_128 2 "<round_nimm_predicate>"
20363         "<round_constraint>")): Changed to ...
20364         ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
20365         "<round_scalar_constraint>")): ... this.
20366         ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
20367         %0, %1, %2<round_op3>}"): Changed to ...
20368         ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
20369         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
20370         %2<round_scalar_mask_op3>}"): ... this.
20371         * config/i386/subst.md (round_scalar_nimm_predicate): New.
20373 2018-02-12  Olga Makhotina  <olga.makhotina@intel.com>
20375         * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
20376         (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
20377         (_mm_maskz_sqrt_round_ss): New intrinsics.
20378         (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
20379         (__builtin_ia32_sqrtsd_mask_round)
20380         (__builtin_ia32_sqrtss_mask_round): New builtins.
20381         * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
20382         (__builtin_ia32_sqrtss_round): Remove.
20383         (__builtin_ia32_sqrtsd_mask_round)
20384         (__builtin_ia32_sqrtss_mask_round): New builtins.
20385         * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
20386         (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
20387         ((match_operand:VF_128 1 "vector_operand"
20388         "xBm,<round_constraint>")): Changed to ...
20389         ((match_operand:VF_128 1 "vector_operand"
20390         "xBm,<round_scalar_constraint>")): ... this.
20391         (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
20392         %0, %2, %<iptr>1<round_op3>}): Changed to ...
20393         (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
20394         %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
20395         %<iptr>1<round_scalar_mask_op3>}): ... this.
20396         ((set_attr "prefix" "<round_prefix>")): Changed to ...
20397         ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
20399 2018-02-11  Steven Munroe  <munroesj@gcc.gnu.org>
20401         PR target/84266
20402         * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
20403         Cast vec_cmpeq result to correct type.
20404         * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
20405         Cast vec_cmpgt result to correct type.
20407 2018-02-11  Alexandre Oliva  <aoliva@redhat.com>
20409         * final.c (final_scan_insn_1): Renamed from...
20410         (final_scan_insn): ... this.  New wrapper, to recover
20411         seen from the outermost call in recursive ones.
20412         * config/sparc/sparc.c (output_return): Drop seen from call.
20413         (output_sibcall): Likewise.
20414         * config/visium/visium.c (output_branch): Likewise.
20416 2018-02-10  John David Anglin  <danglin@gcc.gnu.org>
20418         * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
20419         function label.
20421 2018-02-10  Alan Modra  <amodra@gmail.com>
20423         PR target/84300
20424         * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
20425         Specify LR as an input.
20427 2018-02-10  Jakub Jelinek  <jakub@redhat.com>
20429         PR sanitizer/83987
20430         * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
20431         remove_member_access_dummy_vars): New functions.
20432         (lower_omp_for, lower_omp_taskreg, lower_omp_target,
20433         lower_omp_1, execute_lower_omp): Use them.
20435         PR rtl-optimization/84308
20436         * shrink-wrap.c (spread_components): Release todo vector.
20438 2018-02-09  Vladimir Makarov  <vmakarov@redhat.com>
20440         PR rtl-optimization/57193
20441         * ira-color.c (struct allocno_color_data): Add member
20442         conflict_allocno_hard_prefs.
20443         (update_conflict_allocno_hard_prefs): New.
20444         (bucket_allocno_compare_func): Add a preference based on
20445         conflict_allocno_hard_prefs.
20446         (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
20447         (color_allocnos): Remove a dead code.  Initiate
20448         conflict_allocno_hard_prefs.  Call update_costs_from_prefs.
20450 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
20452         PR target/84226
20453         * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
20454         constraint from =wa to wa.  Avoid a subreg on the output operand,
20455         instead use a pseudo and subreg it in a move.
20456         (p9_xxbrd_<mode>): Changed to ...
20457         (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
20458         (p9_xxbrd_v2df): New expander.
20459         (p9_xxbrw_<mode>): Changed to ...
20460         (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
20461         (p9_xxbrw_v4sf): New expander.
20463 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
20465         * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
20467 2018-02-09  Peter Bergner  <bergner@vnet.ibm.com>
20469         PR target/83926
20470         * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
20471         multiply in 32-bit mode.
20472         (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
20473         (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
20474         mode.
20476 2018-02-09  Sebastian Perta  <sebastian.perta@renesas.com>
20478         * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
20479         to allow or block "symbol_ref" depending on the value of TARGET_JSR.
20480         * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
20481         * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
20483 2018-02-09  Pierre-Marie de Rodat  <derodat@adacore.com>
20485         PR lto/84213
20486         * dwarf2out.c (is_trivial_indirect_ref): New function.
20487         (dwarf2out_late_global_decl): Do not generate a location
20488         attribute for variables that have a non-trivial DECL_VALUE_EXPR
20489         and that are not defined in the current unit.
20491 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
20493         * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
20494         instead of a libcall for UNORDERED.
20496 2018-02-09  Tamar Christina  <tamar.christina@arm.com>
20498         PR target/82641
20499         * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
20500         __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
20502 2018-02-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
20504         PR target/PR84295
20505         * config/s390/s390.c (s390_set_current_function): Invoke
20506         s390_indirect_branch_settings also if fndecl didn't change.
20508 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
20510         * config/rs6000/rs6000.md (blockage): Set length to zero.
20512 2018-02-09  Eric Botcazou  <ebotcazou@adacore.com>
20514         * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
20516 2018-02-09  Jakub Jelinek  <jakub@redhat.com>
20518         PR sanitizer/84285
20519         * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
20520         STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
20521         -static-lib*san.
20523         PR debug/84252
20524         * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
20525         PARALLEL incoming that failed vt_get_decl_and_offset check.
20527         PR middle-end/84237
20528         * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
20529         * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
20530         TREE_READONLY bit.
20531         (get_variable_section): For decls in named .bss* sections pass true as
20532         second argument to bss_initializer_p.
20534 2018-02-09  Marek Polacek  <polacek@redhat.com>
20535             Jakub Jelinek  <jakub@redhat.com>
20537         PR c++/83659
20538         * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
20539         Formatting fixes.  Verify first that tree_fits_poly_int64_p (op01).
20540         Sync some changes from cxx_fold_indirect_ref.
20542 2018-02-09  Alexandre Oliva  <aoliva@redhat.com>
20544         * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
20545         markers.
20546         * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
20547         (BLOCK_INLINE_ENTRY_LABEL): New.
20548         (dwarf2out_var_location): Disregard inline entry markers.
20549         (inline_entry_data): New struct.
20550         (inline_entry_data_hasher): New hashtable type.
20551         (inline_entry_data_hasher::hash): New.
20552         (inline_entry_data_hasher::equal): New.
20553         (inline_entry_data_table): New variable.
20554         (add_high_low_attributes): Add DW_AT_entry_pc and
20555         DW_AT_GNU_entry_view attributes if a pending entry is found
20556         in inline_entry_data_table.  Add old entry_pc attribute only
20557         if debug nonbinding markers are disabled.
20558         (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
20559         markers are enabled.
20560         (block_within_block_p, dwarf2out_inline_entry): New.
20561         (dwarf2out_finish): Check that no entries remained in
20562         inline_entry_data_table.
20563         * final.c (reemit_insn_block_notes): Handle inline entry notes.
20564         (final_scan_insn, notice_source_line): Likewise.
20565         (rest_of_clean_state): Skip inline entry markers.
20566         * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
20567         markers.
20568         * gimple.c (gimple_build_debug_inline_entry): New.
20569         * gimple.h (enum gimple_debug_subcode): Add
20570         GIMPLE_DEBUG_INLINE_ENTRY.
20571         (gimple_build_debug_inline_entry): Declare.
20572         (gimple_debug_inline_entry_p): New.
20573         (gimple_debug_nonbind_marker_p): Adjust.
20574         * insn-notes.def (INLINE_ENTRY): New.
20575         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
20576         inline entry marker notes.
20577         (print_insn): Likewise.
20578         * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
20579         (INSN_DEBUG_MARKER_KIND): Likewise.
20580         (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
20581         * tree-inline.c (expand_call_inline): Build and insert
20582         debug_inline_entry stmt.
20583         * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
20584         inline entry blocks early, if nonbind markers are enabled.
20585         (dump_scope_block): Dump fragment info.
20586         * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
20587         * doc/gimple.texi (gimple_debug_inline_entry_p): New.
20588         (gimple_build_debug_inline_entry): New.
20589         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
20590         Enable/disable inline entry points too.
20591         * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
20592         (DEBUG_INSN): Describe inline entry markers.
20594         * common.opt (gvariable-location-views): New.
20595         (gvariable-location-views=incompat5): New.
20596         * config.in: Rebuilt.
20597         * configure: Rebuilt.
20598         * configure.ac: Test assembler for view support.
20599         * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
20600         * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
20601         * dwarf2out.c (var_loc_view): New typedef.
20602         (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
20603         (dwarf2out_locviews_in_attribute): New.
20604         (dwarf2out_locviews_in_loclist): New.
20605         (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
20606         (enum dw_line_info_opcode): Add LI_adv_address.
20607         (struct dw_line_info_table): Add view.
20608         (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
20609         (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
20610         (zero_view_p): New variable.
20611         (ZERO_VIEW_P): New macro.
20612         (output_asm_line_debug_info): New.
20613         (struct var_loc_node): Add view.
20614         (add_AT_view_list, AT_loc_list): New.
20615         (add_var_loc_to_decl): Add view param.  Test it against last.
20616         (new_loc_list): Add view params.  Record them.
20617         (AT_loc_list_ptr): Handle loc and view lists.
20618         (view_list_to_loc_list_val_node): New.
20619         (print_dw_val): Handle dw_val_class_view_list.
20620         (size_of_die): Likewise.
20621         (value_format): Likewise.
20622         (loc_list_has_views): New.
20623         (gen_llsym): Set vl_symbol too.
20624         (maybe_gen_llsym, skip_loc_list_entry): New.
20625         (dwarf2out_maybe_output_loclist_view_pair): New.
20626         (output_loc_list): Output view list or entries too.
20627         (output_view_list_offset): New.
20628         (output_die): Handle dw_val_class_view_list.
20629         (output_dwarf_version): New.
20630         (output_compilation_unit_header): Use it.
20631         (output_skeleton_debug_sections): Likewise.
20632         (output_rnglists, output_line_info): Likewise.
20633         (output_pubnames, output_aranges): Update version comments.
20634         (output_one_line_info_table): Output view numbers in asm comments.
20635         (dw_loc_list): Determine current endview, pass it to new_loc_list.
20636         Call maybe_gen_llsym.
20637         (loc_list_from_tree_1): Adjust.
20638         (add_AT_location_description): Create view list attribute if
20639         needed, check it's absent otherwise.
20640         (convert_cfa_to_fb_loc_list): Adjust.
20641         (maybe_emit_file): Call output_asm_line_debug_info for test.
20642         (dwarf2out_var_location): Reset views as needed.  Precompute
20643         add_var_loc_to_decl args.  Call get_attr_min_length only if we have the
20644         attribute.  Set view.
20645         (new_line_info_table): Reset next view.
20646         (set_cur_line_info_table): Call output_asm_line_debug_info for test.
20647         (dwarf2out_source_line): Likewise.  Output view resets and labels to
20648         the assembler, or select appropriate line info opcodes.
20649         (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
20650         (optimize_string_length): Catch it.  Adjust.
20651         (resolve_addr): Copy vl_symbol along with ll_symbol.  Handle
20652         dw_val_class_view_list, and remove it if no longer needed.
20653         (hash_loc_list): Hash view numbers.
20654         (loc_list_hasher::equal): Compare them.
20655         (optimize_location_lists): Check whether a view list symbol is
20656         needed, and whether the locview attribute is present, and
20657         whether they match.  Remove the locview attribute if no longer
20658         needed.
20659         (index_location_lists): Call skip_loc_list_entry for test.
20660         (dwarf2out_finish): Call output_asm_line_debug_info for test.
20661         Use output_dwarf_version.
20662         * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
20663         (struct dw_val_node): Add val_view_list.
20664         * final.c (SEEN_NEXT_VIEW): New.
20665         (set_next_view_needed): New.
20666         (clear_next_view_needed): New.
20667         (maybe_output_next_view): New.
20668         (final_start_function): Rename to...
20669         (final_start_function_1): ... this.  Take pointer to FIRST,
20670         add SEEN parameter.  Emit param bindings in the initial view.
20671         (final_start_function): Reintroduce SEEN-less interface.
20672         (final): Rename to...
20673         (final_1): ... this.  Take SEEN parameter.  Output final pending
20674         next view at the end.
20675         (final): Reintroduce seen-less interface.
20676         (final_scan_insn): Output pending next view before switching
20677         sections or ending a block.  Mark the next view as needed when
20678         outputting variable locations.  Notify debug backend of section
20679         changes, and of location view changes.
20680         (rest_of_handle_final): Adjust.
20681         * toplev.c (process_options): Autodetect value for debug variable
20682         location views option.  Warn on incompat5 without -gdwarf-5.
20683         * doc/invoke.texi (gvariable-location-views): New.
20684         (gvariable-location-views=incompat5): New.
20685         (gno-variable-location-views): New.
20687 2018-02-08  David Malcolm  <dmalcolm@redhat.com>
20689         PR tree-optimization/84136
20690         * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
20691         that the result of find_edge is non-NULL.
20693 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
20695         PR target/83008
20696         * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
20697         storing integer register in SImode.  Fix cost of 256 and 512
20698         byte aligned SSE register store.
20700 2018-02-08  Sergey Shalnov  <sergey.shalnov@intel.com>
20702         * config/i386/i386.c (ix86_multiplication_cost): Fix
20703         multiplication cost for TARGET_AVX512DQ.
20705 2018-02-08  Marek Polacek  <polacek@redhat.com>
20707         PR tree-optimization/84238
20708         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
20709         get_range_strlen.
20711 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
20713         PR tree-optimization/84265
20714         * tree-vect-stmts.c (vectorizable_store): Don't treat
20715         VMAT_CONTIGUOUS accesses as grouped.
20716         (vectorizable_load): Likewise.
20718 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
20720         PR tree-optimization/81635
20721         * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
20722         * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
20723         (test_round_for_mask): New functions.
20724         (wide_int_cc_tests): Call test_round_for_mask.
20725         * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
20726         * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
20727         * tree-data-ref.c (split_constant_offset_1): Use it to refine the
20728         range returned by get_range_info.
20730 2018-02-08  Jan Hubicka  <hubicka@ucw.cz>
20732         PR ipa/81360
20733         * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
20734         * symtab.c: Include builtins.h
20735         (symtab_node::output_to_lto_symbol_table_p): Move here
20736         from lto-streamer-out.c:output_symbol_p.
20737         * lto-streamer-out.c (write_symbol): Turn early exit to assert.
20738         (output_symbol_p): Move all logic to symtab.c
20739         (produce_symtab): Update.
20741 2018-02-08  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
20743         * config/s390/s390-opts.h (enum indirect_branch): Define.
20744         * config/s390/s390-protos.h (s390_return_addr_from_memory)
20745         (s390_indirect_branch_via_thunk)
20746         (s390_indirect_branch_via_inline_thunk): Add function prototypes.
20747         (enum s390_indirect_branch_type): Define.
20748         * config/s390/s390.c (struct s390_frame_layout, struct
20749         machine_function): Remove.
20750         (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
20751         (indirect_branch_table_label_no, indirect_branch_table_name):
20752         Define variables.
20753         (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
20754         (enum s390_indirect_branch_option): Define.
20755         (s390_return_addr_from_memory): New function.
20756         (s390_handle_string_attribute): New function.
20757         (s390_attribute_table): Add new attribute handler.
20758         (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
20759         (s390_indirect_branch_via_thunk): New function.
20760         (s390_indirect_branch_via_inline_thunk): New function.
20761         (s390_function_ok_for_sibcall): When jumping via thunk disallow
20762         sibling call optimization for non z10 compiles.
20763         (s390_emit_call): Force indirect branch target to be a single
20764         register.  Add r1 clobber for non-z10 compiles.
20765         (s390_emit_epilogue): Emit return jump via return_use expander.
20766         (s390_reorg): Handle JUMP_INSNs as execute targets.
20767         (s390_option_override_internal): Perform validity checks for the
20768         new command line options.
20769         (s390_indirect_branch_attrvalue): New function.
20770         (s390_indirect_branch_settings): New function.
20771         (s390_set_current_function): Invoke s390_indirect_branch_settings.
20772         (s390_output_indirect_thunk_function):  New function.
20773         (s390_code_end): Implement target hook.
20774         (s390_case_values_threshold): Implement target hook.
20775         (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
20776         macros.
20777         * config/s390/s390.h (struct s390_frame_layout)
20778         (struct machine_function): Move here from s390.c.
20779         (TARGET_INDIRECT_BRANCH_NOBP_RET)
20780         (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
20781         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
20782         (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
20783         (TARGET_INDIRECT_BRANCH_NOBP_CALL)
20784         (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
20785         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
20786         (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
20787         (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
20788         * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
20789         (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
20790         (mnemonic attribute): Add values which aren't recognized
20791         automatically.
20792         ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
20793         pattern for branch conversion.  Fix mnemonic attribute.
20794         ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
20795         indirect branch via thunk if requested.
20796         ("indirect_jump", "<code>"): Expand patterns for branch conversion.
20797         ("*indirect_jump"): Disable for branch conversion using out of
20798         line thunks.
20799         ("indirect_jump_via_thunk<mode>_z10")
20800         ("indirect_jump_via_thunk<mode>")
20801         ("indirect_jump_via_inlinethunk<mode>_z10")
20802         ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
20803         ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
20804         ("casesi_jump_via_inlinethunk<mode>_z10")
20805         ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
20806         ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
20807         ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
20808         ("*indirect2_jump"): Disable for branch conversion.
20809         ("casesi_jump"): Turn into expander and expand patterns for branch
20810         conversion.
20811         ("return_use"): New expander.
20812         ("*return"): Emit return via thunk and rename it to ...
20813         ("*return<mode>"): ... this one.
20814         * config/s390/s390.opt: Add new options and and enum for the
20815         option values.
20817 2018-02-08  Richard Sandiford  <richard.sandiford@linaro.org>
20819         * lra-constraints.c (match_reload): Unconditionally use
20820         gen_lowpart_SUBREG, rather than selecting between that
20821         and equivalent gen_rtx_SUBREG code.
20823 2018-02-08  Richard Biener  <rguenther@suse.de>
20825         PR tree-optimization/84233
20826         * tree-ssa-phiprop.c (propagate_with_phi): Use separate
20827         changed flag instead of boguously re-using phi_inserted.
20829 2018-02-08  Martin Jambor  <mjambor@suse.cz>
20831         * hsa-gen.c (get_symbol_for_decl): Set program allocation for
20832         static local variables.
20834 2018-02-08  Richard Biener  <rguenther@suse.de>
20836         PR tree-optimization/84278
20837         * tree-vect-stmts.c (vectorizable_store): When looking for
20838         smaller vector types to perform grouped strided loads/stores
20839         make sure the mode is supported by the target.
20840         (vectorizable_load): Likewise.
20842 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
20844         * config/aarch64/aarch64.c (aarch64_components_for_bb):
20845         Increase LDP/STP opportunities by adding adjacent callee-saves.
20847 2018-02-08  Wilco Dijkstra  <wdijkstr@arm.com>
20849         PR rtl-optimization/84068
20850         PR rtl-optimization/83459
20851         * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
20853 2018-02-08  Aldy Hernandez  <aldyh@redhat.com>
20855         PR tree-optimization/84224
20856         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
20857         * calls.c (gimple_alloca_call_p): Only return TRUE when we have
20858         non-zero arguments.
20860 2018-02-07  Iain Sandoe  <iain@codesourcery.com>
20862         PR target/84113
20863         * config/rs6000/altivec.md (*restore_world): Remove LR use.
20864         * config/rs6000/predicates.md (restore_world_operation): Adjust op
20865         count, remove one USE.
20867 2018-02-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
20869         * doc/install.texi (Configuration): Document the
20870         --with-long-double-format={ibm,ieee} PowerPC configuration
20871         options.
20873         PR target/84154
20874         * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
20875         Convert from define_expand to be define_insn_and_split.  Rework
20876         float/double/_Float128 conversions to QI/HI/SImode to work with
20877         both ISA 2.07 (power8) or ISA 3.0 (power9).  Fix regression where
20878         conversions to QI/HImode types did a store and then a load to
20879         truncate the value.  For conversions to VSX registers, don't split
20880         the insn, instead emit the code directly.  Use the code iterator
20881         any_fix to combine signed and unsigned conversions.
20882         (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
20883         (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
20884         (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
20885         (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
20886         (fix_<mode>di2_hw): Likewise.
20887         (fixuns_<mode>di2_hw): Likewise.
20888         (fix_<mode>si2_hw): Likewise.
20889         (fixuns_<mode>si2_hw): Likewise.
20890         (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
20891         (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
20892         (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
20893         fix<uns>_trunc<SFDF:mode>si2_p8.
20894         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
20895         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
20896         (fix<uns>_<mode>_mem): Likewise.
20897         (fctiw<u>z_<mode>_mem): Likewise.
20898         (fix<uns>_<mode>_mem): Likewise.
20899         (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
20900         the register allocator from doing a direct move to the GPRs to do
20901         a store, and instead use the ISA 3.0 store byte/half-word from
20902         vector register instruction.  For IEEE 128-bit floating point,
20903         also optimize stores of 32-bit ints.
20904         (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
20906 2018-02-07  Alan Hayward  <alan.hayward@arm.com>
20908         * genextract.c (push_pathstr_operand): New function to support
20909         [a-zA-Z].
20910         (walk_rtx): Call push_pathstr_operand.
20911         (print_path): Support [a-zA-Z].
20913 2018-02-07  Richard Biener  <rguenther@suse.de>
20915         PR tree-optimization/84037
20916         * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
20917         (cse_and_gimplify_to_preheader): Declare.
20918         (vect_get_place_in_interleaving_chain): Likewise.
20919         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
20920         ivexpr_map.
20921         (_loop_vec_info::~_loop_vec_info): Delete it.
20922         (cse_and_gimplify_to_preheader): New function.
20923         * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
20924         * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
20925         (vectorizable_load): Likewise.  For grouped stores always base
20926         the IV on the first element.
20927         * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
20928         condition before gimplifying.
20930 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
20932         * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
20933         *DIV_EXPR and *MOD_EXPR.
20935 2018-02-07  H.J. Lu  <hongjiu.lu@intel.com>
20937         PR target/84248
20938         * config/i386/i386.c (ix86_option_override_internal): Mask out
20939         the CF_SET bit when checking -fcf-protection.
20941 2018-02-07  Tom de Vries  <tom@codesourcery.com>
20943         PR libgomp/84217
20944         * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
20945         enough.
20947 2018-02-07  Richard Biener  <rguenther@suse.de>
20949         PR tree-optimization/84204
20950         * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
20951         this place.
20953         PR tree-optimization/84205
20954         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
20955         special-case isl_ast_op_zdiv_r.
20957         PR tree-optimization/84223
20958         * graphite-scop-detection.c (gather_bbs::before_dom_children):
20959         Only add conditions from within the region.
20960         (gather_bbs::after_dom_children): Adjust.
20962 2018-02-07  Georg-Johann Lay  <avr@gjlay.de>
20964         PR target/84209
20965         * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
20966         * config/avr/avr.md: Only post-reload split REG-REG moves if
20967         either register is GENERAL_REG_P.
20969 2018-02-07  Jakub Jelinek  <jakub@redhat.com>
20971         PR tree-optimization/84235
20972         * tree-ssa-scopedtables.c
20973         (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
20974         if the subtraction is performed in floating point type where NaNs are
20975         honored.  For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
20976         build 1.  Formatting fix.
20978 2018-02-06  Jakub Jelinek  <jakub@redhat.com>
20980         PR target/84146
20981         * config/i386/i386.c (rest_of_insert_endbranch): Only skip
20982         NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
20983         and skip it regardless of bb boundaries.  Use CALL_P macro,
20984         don't test INSN_P (insn) together with CALL_P or JUMP_P check
20985         unnecessarily, formatting fix.
20987 2018-02-06  Michael Collison  <michael.collison@arm.com>
20989         * config/arm/thumb2.md:
20990         (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
20991         (*thumb_mov_notscc): Ditto.
20993 2018-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
20995         PR target/84154
20996         * config/rs6000/rs6000.md (su code attribute): Use "u" for
20997         unsigned_fix, not "s".
20999 2018-02-06  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
21001         * configure.ac (gcc_fn_eh_frame_ro): New function.
21002         (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
21003         correct .eh_frame permissions.
21004         * configure: Regenerate.
21006 2018-02-06  Andrew Jenner  <andrew@codeourcery.com>
21008         * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
21009         irrelevant options.
21011 2018-02-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21013         * config/rs6000/rs6000.c (rs6000_option_override_internal):
21014         Display warning message for -mno-speculate-indirect-jumps.
21016 2018-02-06  Andrew Jenner  <andrew@codesourcery.com>
21018         * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
21019         Undocumented.
21020         * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
21022 2018-02-06  Aldy Hernandez  <aldyh@redhat.com>
21024         PR tree-optimization/84225
21025         * tree-eh.c (find_trapping_overflow): Only call
21026         operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
21028 2018-02-06  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
21030         PR target/84145
21031         * config/i386/i386.c: Reimplement the check of possible options
21032         -mibt/-mshstk conbination. Change error messages.
21033         * doc/invoke.texi: Fix a typo: remove extra '='.
21035 2018-02-06  Marek Polacek  <polacek@redhat.com>
21037         PR tree-optimization/84228
21038         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
21040 2018-02-06  Tamar Christina  <tamar.christina@arm.com>
21042         PR target/82641
21043         * config/arm/arm.c (arm_print_asm_arch_directives): Record already
21044         emitted arch directives.
21045         * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
21046         __ARM_FEATURE_COPROC before changing architectures.
21048 2018-02-06  Richard Biener  <rguenther@suse.de>
21050         * config/i386/i386.c (print_reg): Fix typo.
21051         (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
21053 2018-02-06  Eric Botcazou  <ebotcazou@adacore.com>
21055         * configure: Regenerate.
21057 2018-02-05  Martin Sebor  <msebor@redhat.com>
21059         PR tree-optimization/83369
21060         * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
21061         inlining context.
21063 2018-02-05  Martin Liska  <mliska@suse.cz>
21065         * doc/invoke.texi: Cherry-pick upstream r323995.
21067 2018-02-05  Richard Sandiford  <richard.sandiford@linaro.org>
21069         * ira.c (ira_init_register_move_cost): Adjust comment.
21071 2018-02-05  Martin Liska  <mliska@suse.cz>
21073         PR gcov-profile/84137
21074         * doc/gcov.texi: Fix typo in documentation.
21076 2018-02-05  Martin Liska  <mliska@suse.cz>
21078         PR gcov-profile/83879
21079         * doc/gcov.texi: Document necessity of --dynamic-list-data when
21080         using dlopen functionality.
21082 2018-02-05  Olga Makhotina  <olga.makhotina@intel.com>
21084         * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
21085         _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
21086         _mm_maskz_range_ss, _mm_mask_range_round_ss,
21087         _mm_maskz_range_round_ss): New intrinsics.
21088         (__builtin_ia32_rangesd128_round)
21089         (__builtin_ia32_rangess128_round): Remove.
21090         (__builtin_ia32_rangesd128_mask_round,
21091         __builtin_ia32_rangess128_mask_round): New builtins.
21092         * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
21093         __builtin_ia32_rangess128_round): Remove.
21094         (__builtin_ia32_rangesd128_mask_round,
21095         __builtin_ia32_rangess128_mask_round): New builtins.
21096         * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
21097         (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
21098         ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
21099         "<round_saeonly_constraint>")): Changed to ...
21100         ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
21101         "<round_saeonly_scalar_constraint>")): ... this.
21102         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
21103         %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
21104         ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
21105         %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
21106         %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
21108 2018-02-02  Andrew Jenner  <andrew@codesourcery.com>
21110         * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
21111         options.
21112         * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
21113         Remove all values except native, 8540 and 8548.
21115 2018-02-02  H.J. Lu  <hongjiu.lu@intel.com>
21117         * config/i386/i386.c (ix86_output_function_return): Pass
21118         INVALID_REGNUM, instead of -1, as invalid register number to
21119         indirect_thunk_name and output_indirect_thunk.
21121 2018-02-02  Julia Koval  <julia.koval@intel.com>
21123         * config.gcc: Add -march=icelake.
21124         * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
21125         * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
21126         * config/i386/i386.c (processor_costs): Add m_ICELAKE.
21127         (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
21128         PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
21129         (processor_target_table): Add icelake.
21130         (ix86_option_override_internal): Handle new PTAs.
21131         (get_builtin_code_for_version): Handle icelake.
21132         (M_INTEL_COREI7_ICELAKE): New.
21133         (fold_builtin_cpu): Handle icelake.
21134         * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
21135         * doc/invoke.texi: Add -march=icelake.
21137 2018-02-02  Julia Koval  <julia.koval@intel.com>
21139         * config/i386/i386.c (ix86_option_override_internal): Change flags type
21140         to wide_int_bitmask.
21141         * wide-int-bitmask.h: New.
21143 2018-02-02  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
21145         PR target/84066
21146         * config/i386/i386.md: Replace Pmode with word_mode in
21147         builtin_setjmp_setup and builtin_longjmp to support x32.
21149 2018-02-01  Peter Bergner  <bergner@vnet.ibm.com>
21151         PR target/56010
21152         PR target/83743
21153         * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
21154         #include "opts.h".
21155         (rs6000_supported_cpu_names): New static variable.
21156         (linux_cpu_translation_table): Likewise.
21157         (elf_platform) <cpu>: Define new static variable and use it.
21158         Translate kernel AT_PLATFORM name to canonical name if needed.
21159         Error if platform name is unknown.
21161 2018-02-01  Aldy Hernandez  <aldyh@redhat.com>
21163         PR target/84089
21164         * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
21166 2018-02-01  Jeff Law  <law@redhat.com>
21168         PR target/84128
21169         * config/i386/i386.c (release_scratch_register_on_entry): Add new
21170         OFFSET and RELEASE_VIA_POP arguments.  Use SP+OFFSET to restore
21171         the scratch if RELEASE_VIA_POP is false.
21172         (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
21173         If we have to save a temporary register, decrement SIZE appropriately.
21174         Pass new arguments to release_scratch_register_on_entry.
21175         (ix86_adjust_stack_and_probe): Likewise.
21176         (ix86_emit_probe_stack_range): Pass new arguments to
21177         release_scratch_register_on_entry.
21179 2018-02-01  Uros Bizjak  <ubizjak@gmail.com>
21181         PR rtl-optimization/84157
21182         * combine.c (change_zero_ext): Use REG_P predicate in
21183         front of HARD_REGISTER_P predicate.
21185 2018-02-01  Georg-Johann Lay  <avr@gjlay.de>
21187         * config/avr/avr.c (avr_option_override): Move disabling of
21188         -fdelete-null-pointer-checks to...
21189         * common/config/avr/avr-common.c (avr_option_optimization_table):
21190         ...here.
21192 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
21194         PR tree-optimization/81635
21195         * tree-data-ref.c (split_constant_offset_1): For types that
21196         wrap on overflow, try to use range info to prove that wrapping
21197         cannot occur.
21199 2018-02-01  Renlin Li  <renlin.li@arm.com>
21201         PR target/83370
21202         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
21203         TAILCALL_ADDR_REGS.
21204         (aarch64_register_move_cost): Likewise.
21205         * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
21206         TAILCALL_ADDR_REGS.
21207         (REG_CLASS_NAMES): Likewise.
21208         (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
21209         TAILCALL_ADDR_REGS. Remove IP registers.
21210         * config/aarch64/aarch64.md (Ucs): Update register constraint.
21212 2018-02-01  Richard Biener  <rguenther@suse.de>
21214         * domwalk.h (dom_walker::dom_walker): Add additional constructor
21215         for specifying RPO order and allow NULL for that.
21216         * domwalk.c (dom_walker::dom_walker): Likewise.
21217         (dom_walker::walk): Handle NULL RPO order.
21218         * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
21219         in RPO order.
21220         (rewrite_update_dom_walker): Likewise.
21221         (mark_def_dom_walker): Likewise.
21223 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
21225         * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
21226         (aarch64_maybe_expand_sve_subreg_move): Declare.
21227         * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
21228         * config/aarch64/predicates.md (aarch64_any_register_operand): New
21229         predicate.
21230         * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
21231         that are semantically a reverse operation.
21232         (*aarch64_sve_mov<mode>_subreg_be): New pattern.
21233         * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
21234         (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
21235         functions.
21236         (aarch64_can_change_mode_class): For big-endian, forbid changes
21237         between two SVE modes if they have different element sizes.
21239 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
21241         * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
21242         the TImode handling for big-endian targets.
21244 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
21246         * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
21247         (*sve_ld1rq<Vesize>): ... this new pattern.  Handle all element sizes,
21248         not just bytes.
21249         * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
21250         Remove BSWAP handing for big-endian targets and use the form of
21251         LD1RQ appropariate for the mode.
21253 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
21255         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
21256         all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
21257         duplicated element.
21259 2018-02-01  Richard Sandiford  <richard.sandiford@linaro.org>
21261         PR tearget/83845
21262         * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
21263         check for operands that need to go through aarch64_sve_reload_be.
21265 2018-02-01  Jakub Jelinek  <jakub@redhat.com>
21267         PR tree-optimization/81661
21268         PR tree-optimization/84117
21269         * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
21270         * tree-eh.c: Include gimplify.h.
21271         (find_trapping_overflow, replace_trapping_overflow,
21272         rewrite_to_non_trapping_overflow): New functions.
21273         * tree-vect-loop.c: Include tree-eh.h.
21274         (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
21275         * tree-data-ref.c: Include tree-eh.h.
21276         (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
21278 2018-01-31  Uros Bizjak  <ubizjak@gmail.com>
21280         PR rtl-optimization/84123
21281         * combine.c (change_zero_ext): Check if hard register satisfies
21282         can_change_dest_mode before calling gen_lowpart_SUBREG.
21284 2018-01-31  Vladimir Makarov  <vmakarov@redhat.com>
21286         PR target/82444
21287         * ira.c (ira_init_register_move_cost): Remove assert.
21289 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
21291         PR rtl-optimization/84071
21292         * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
21293         * doc/tm.texi: Regenerate.
21295 2018-01-31  Richard Biener  <rguenther@suse.de>
21297         PR tree-optimization/84132
21298         * tree-data-ref.c (analyze_miv_subscript): Properly
21299         check whether evolution_function_is_affine_multivariate_p
21300         before calling gcd_of_steps_may_divide_p.
21302 2018-01-31  Julia Koval  <julia.koval@intel.com>
21304         PR target/83618
21305         * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
21306         * config/i386/i386.md (rdpid_rex64) New.
21307         (rdpid): Make 32bit only.
21309 2018-01-29  Aldy Hernandez  <aldyh@redhat.com>
21311         PR lto/84105
21312         * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
21313         an IDENTIFIER_NODE for FUNCTION_TYPE's.
21315 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
21317         Revert
21318         2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
21320         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
21322 2018-01-31  Eric Botcazou  <ebotcazou@adacore.com>
21324         PR rtl-optimization/84071
21325         * combine.c (record_dead_and_set_regs_1): Record the source unmodified
21326         for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
21328 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
21330         * config/arc/arc.c (arc_handle_aux_attribute): New function.
21331         (arc_attribute_table): Add 'aux' attribute.
21332         (arc_in_small_data_p): Consider aux like variables.
21333         (arc_is_aux_reg_p): New function.
21334         (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
21335         (arc_get_aux_arg): New function.
21336         (prepare_move_operands): Handle aux-register access.
21337         (arc_handle_aux_attribute): New function.
21338         * doc/extend.texi (ARC Variable attributes): Add subsection.
21340 2018-01-31  Claudiu Zissulescu  <claziss@synopsys.com>
21342         * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
21343         * config/arc/arc.c (arc_handle_uncached_attribute): New function.
21344         (arc_attribute_table): Add 'uncached' attribute.
21345         (arc_print_operand): Print '.di' flag for uncached memory
21346         accesses.
21347         (arc_in_small_data_p): Do not consider for small data the uncached
21348         types.
21349         (arc_is_uncached_mem_p): New function.
21350         * config/arc/predicates.md (compact_store_memory_operand): Check
21351         for uncached memory accesses.
21352         (nonvol_nonimm_operand): Likewise.
21353         * doc/extend.texi (ARC Type Attribute): New subsection.
21355 2018-01-31  Jakub Jelinek  <jakub@redhat.com>
21357         PR c/84100
21358         * common.opt (falign-functions=, falign-jumps=, falign-labels=,
21359         falign-loops=): Add Optimization flag.
21361 2018-01-30  Jeff Law  <law@redhat.com>
21363         PR target/84064
21364         * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
21365         INT_REGISTERS_SAVED.  Check it prior to calling
21366         get_scratch_register_on_entry.
21367         (ix86_adjust_stack_and_probe): Similarly.
21368         (ix86_emit_probe_stack_range): Similarly.
21369         (ix86_expand_prologue): Corresponding changes.
21371 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
21373         PR target/40411
21374         * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
21375         -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
21377 2018-01-30  Vladimir Makarov  <vmakarov@redhat.com>
21379         PR target/84112
21380         * lra-constraints.c (curr_insn_transform): Process AND in the
21381         address.
21383 2018-01-30  Jakub Jelinek  <jakub@redhat.com>
21385         PR rtl-optimization/83986
21386         * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
21387         dependence against last_pending_memory_flush in addition to
21388         pending_jump_insns.
21390 2018-01-30  Alexandre Oliva  <aoliva@redhat.com>
21392         PR tree-optimization/81611
21393         * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
21394         copies.
21396 2018-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
21398         PR target/83758
21399         * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
21400         a reg rtx.
21402 2018-01-30  Richard Biener  <rguenther@suse.de>
21403             Jakub Jelinek  <jakub@redhat.com>
21405         PR tree-optimization/84111
21406         * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
21407         inner loops added during recursion, as they don't have up-to-date
21408         SSA form.
21410 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
21412         PR ipa/81360
21413         * ipa-inline.c (can_inline_edge_p): Break out late tests to...
21414         (can_inline_edge_by_limits_p): ... here.
21415         (can_early_inline_edge_p, check_callers,
21416         update_caller_keys, update_callee_keys, recursive_inlining,
21417         add_new_edges_to_heap, speculation_useful_p,
21418         inline_small_functions,
21419         inline_small_functions, flatten_function,
21420         inline_to_all_callers_1): Update.
21422 2018-01-30  Jan Hubicka  <hubicka@ucw.cz>
21424         * profile-count.c (profile_count::combine_with_ipa_count): Handle
21425         zeros correctly.
21427 2018-01-30  Richard Biener  <rguenther@suse.de>
21429         PR tree-optimization/83008
21430         * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
21431         invariant and constant vector uses in stmts when they need
21432         more than one stmt.
21434 2018-01-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
21436         PR bootstrap/84017
21437         * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
21438         * configure: Regenerate.
21440 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
21442         * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
21443         pattern.
21444         (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
21445         Use gen_rtx_REG rather than gen_lowpart.
21447 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
21449         * lra-constraints.c (match_reload): Use subreg_lowpart_offset
21450         rather than 0 when creating partial subregs.
21452 2018-01-30  Richard Sandiford  <richard.sandiford@linaro.org>
21454         * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
21455         of usage.
21457 2018-01-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
21459         PR target/81550
21460         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
21461         and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
21462         -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
21463         flags.  This restores the settings used before the 2017-07-24.
21464         Turning off pre increment/decrement/modify allows IVOPTS to
21465         optimize DF/SF loops where the index is an int.
21467 2018-01-29  Richard Biener  <rguenther@suse.de>
21468             Kelvin Nilsen  <kelvin@gcc.gnu.org>
21470         PR bootstrap/80867
21471         * tree-vect-stmts.c (vectorizable_call): Don't call
21472         targetm.vectorize_builtin_md_vectorized_function if callee is
21473         NULL.
21475 2018-01-22  Carl Love  <cel@us.ibm.com>
21477         * doc/extend.tex: Fix typo in second arg in
21478         __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
21480 2018-01-29  Richard Biener  <rguenther@suse.de>
21482         PR tree-optimization/84086
21483         * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
21484         (flush_ssaname_freelist): When SSA names were released reset
21485         the SCEV hash table.
21487 2018-01-29  Richard Biener  <rguenther@suse.de>
21489         PR tree-optimization/84057
21490         * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
21491         removed paths when removing edges.
21493 2018-01-27  H.J. Lu  <hongjiu.lu@intel.com>
21495         * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
21496         -mfunction-return=@var{choice}.
21498 2018-01-27  Bernd Edlinger  <bernd.edlinger@hotmail.de>
21500         PR diagnostic/84034
21501         * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
21502         Handle CR like TAB.
21503         (layout::print_source_line): Likewise.
21504         (test_get_line_width_without_trailing_whitespace): Add test cases.
21506 2018-01-27  Jakub Jelinek  <jakub@redhat.com>
21508         PR middle-end/84040
21509         * sched-deps.c (sched_macro_fuse_insns): Return immediately for
21510         debug insns.
21512 2018-01-26  Jim Wilson  <jimw@sifive.com>
21514         * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
21516         * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
21517         specified.
21519 2018-01-26  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
21521         * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
21522         and CMP + SUB-immediate -> SUBS.
21524 2018-01-26  Martin Sebor  <msebor@redhat.com>
21526         PR tree-optimization/83896
21527         * tree-ssa-strlen.c (get_string_len): Rename...
21528         (get_string_cst_length): ...to this.  Return HOST_WIDE_INT.
21529         Avoid assuming length is constant.
21530         (handle_char_store): Use HOST_WIDE_INT for string length.
21532 2018-01-26  Uros Bizjak  <ubizjak@gmail.com>
21534         PR target/81763
21535         * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
21536         to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
21538 2018-01-26  Richard Biener  <rguenther@suse.de>
21540         PR rtl-optimization/84003
21541         * dse.c (record_store): Only record redundant stores when
21542         the earlier store aliases at least all accesses the later one does.
21544 2018-01-26  Jakub Jelinek  <jakub@redhat.com>
21546         PR rtl-optimization/83985
21547         * dce.c (deletable_insn_p): Return false for separate shrink wrapping
21548         REG_CFA_RESTORE insns.
21549         (delete_unmarked_insns): Don't ignore separate shrink wrapping
21550         REG_CFA_RESTORE insns here.
21552         PR c/83989
21553         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
21554         use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
21556 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
21558         * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
21559         * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
21560         (arc_init): Likewise.
21561         (arc_override_options): Likewise.
21562         (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
21563         value.
21564         (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
21565         support.
21566         * config/arc/arc.h (TARGET_DBNZ): Define.
21567         * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
21568         properly set the tune attribute.
21569         (dbnz): Use TARGET_DBNZ guard.
21570         * config/arc/arc.opt (mtune): Add core3 option.
21572 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
21574         * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
21575         recognize new pic like addresses.
21576         (arc_delegitimize_address): Clean up.
21578 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
21580         * config/arc/arc-arches.def: Option mrf16 valid for all
21581         architectures.
21582         * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
21583         * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
21584         * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
21585         * config/arc/arc-tables.opt: Regenerate.
21586         * config/arc/arc.c (arc_conditional_register_usage): Handle
21587         reduced register file case.
21588         (arc_file_start): Set must have build attributes.
21589         * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
21590         mrf16 option value.
21591         * config/arc/arc.opt (mrf16): Add new option.
21592         * config/arc/elf.h (ATTRIBUTE_PCS): Define.
21593         * config/arc/genmultilib.awk: Handle new mrf16 option.
21594         * config/arc/linux.h (ATTRIBUTE_PCS): Define.
21595         * config/arc/t-multilib: Regenerate.
21596         * doc/invoke.texi (ARC Options): Document mrf16 option.
21598 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
21600         * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
21601         * config/arc/arc.c (arc_handle_secure_attribute): New function.
21602         (arc_attribute_table): Add 'secure_call' attribute.
21603         (arc_print_operand): Print secure call operand.
21604         (arc_function_ok_for_sibcall): Don't optimize tail calls when
21605         secure.
21606         (arc_is_secure_call_p): New function.  * config/arc/arc.md
21607         (call_i): Add support for sjli instruction.
21608         (call_value_i): Likewise.
21609         * config/arc/constraints.md (Csc): New constraint.
21611 2018-01-26  Claudiu Zissulescu  <claziss@synopsys.com>
21612             John Eric Martin  <John.Martin@emmicro-us.com>
21614         * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
21615         * config/arc/arc.c (_arc_jli_section): New struct.
21616         (arc_jli_section): New type.
21617         (rc_jli_sections): New static variable.
21618         (arc_handle_jli_attribute): New function.
21619         (arc_attribute_table): Add jli_always and jli_fixed attribute.
21620         (arc_file_end): New function.
21621         (TARGET_ASM_FILE_END): Define.
21622         (arc_print_operand): Reuse 'S' letter for JLI output instruction.
21623         (arc_add_jli_section): New function.
21624         (jli_call_scan): Likewise.
21625         (arc_reorg): Call jli_call_scan.
21626         (arc_output_addsi): Remove 'S' from printing asm operand.
21627         (arc_is_jli_call_p): New function.
21628         * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
21629         operand.
21630         (movhi_insn): Likewise.
21631         (movsi_insn): Likewise.
21632         (movsi_set_cc_insn): Likewise.
21633         (loadqi_update): Likewise.
21634         (load_zeroextendqisi_update): Likewise.
21635         (load_signextendqisi_update): Likewise.
21636         (loadhi_update): Likewise.
21637         (load_zeroextendhisi_update): Likewise.
21638         (load_signextendhisi_update): Likewise.
21639         (loadsi_update): Likewise.
21640         (loadsf_update): Likewise.
21641         (movsicc_insn): Likewise.
21642         (bset_insn): Likewise.
21643         (bxor_insn): Likewise.
21644         (bclr_insn): Likewise.
21645         (bmsk_insn): Likewise.
21646         (bicsi3_insn): Likewise.
21647         (cmpsi_cc_c_insn): Likewise.
21648         (movsi_ne): Likewise.
21649         (movsi_cond_exec): Likewise.
21650         (clrsbsi2): Likewise.
21651         (norm_f): Likewise.
21652         (normw): Likewise.
21653         (swap): Likewise.
21654         (divaw): Likewise.
21655         (flag): Likewise.
21656         (sr): Likewise.
21657         (kflag): Likewise.
21658         (ffs): Likewise.
21659         (ffs_f): Likewise.
21660         (fls): Likewise.
21661         (call_i): Remove 'S' asm letter, add jli instruction.
21662         (call_value_i): Likewise.
21663         * config/arc/arc.op (mjli-always): New option.
21664         * config/arc/constraints.md (Cji): New constraint.
21665         * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
21666         operand.
21667         (subsf3_fpx): Likewise.
21668         (mulsf3_fpx): Likewise.
21669         * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
21670         asm operand.
21671         * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
21672         function attrbutes.
21673         * doc/invoke.texi (ARC): Document mjli-always option.
21675 2018-01-26  Sebastian Perta  <sebastian.perta@renesas.com>
21677         * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
21678         avoid addition with 0 and use incw and decw where possible.
21680 2018-01-26  Richard Biener  <rguenther@suse.de>
21682         PR tree-optimization/81082
21683         * fold-const.c (fold_plusminus_mult_expr): Do not perform the
21684         association if it requires casting to unsigned.
21685         * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
21686         from fold_plusminus_mult_expr to catch important cases late when
21687         range info is available.
21689 2018-01-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
21691         * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
21692         * configure.ac (hidden_linkonce): New test.
21693         * configure: Regenerate.
21694         * config.in: Regenerate.
21696 2018-01-26  Julia Koval  <julia.koval@intel.com>
21698         * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
21699         _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
21700         _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
21701         _mm_mask_bitshuffle_epi64_mask): Fix type.
21702         * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
21703         USI_FTYPE_V4DI_V4DI_USI): Remove.
21704         * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
21705         __builtin_ia32_vpshufbitqmb256_mask,
21706         __builtin_ia32_vpshufbitqmb128_mask): Fix types.
21707         * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
21708         * config/i386/sse.md (VI1_AVX512VLBW): Change types.
21710 2018-01-26  Alan Modra  <amodra@gmail.com>
21712         PR target/84033
21713         * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
21714         UNSPEC_VBPERMQ.  Sort other unspecs.
21716 2018-01-25  David Edelsohn  <dje.gcc@gmail.com>
21718         * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
21720 2018-01-25  Jan Hubicka  <hubicka@ucw.cz>
21722         PR middle-end/83055
21723         * predict.c (drop_profile): Do not push/pop cfun; update also
21724         node->count.
21725         (handle_missing_profiles): Fix logic looking for zero profiles.
21727 2018-01-25  Jakub Jelinek  <jakub@redhat.com>
21729         PR middle-end/83977
21730         * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
21731         on functions with #pragma omp declare simd or functions with simd
21732         attribute.
21733         * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
21734         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
21735         Remove trailing \n from warning_at calls.
21737 2018-01-25  Tom de Vries  <tom@codesourcery.com>
21739         PR target/84028
21740         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
21741         for neutered workers.
21743 2018-01-24  Joseph Myers  <joseph@codesourcery.com>
21745         PR target/68467
21746         * config/m68k/m68k.c (m68k_promote_function_mode): New function.
21747         (TARGET_PROMOTE_FUNCTION_MODE): New macro.
21749 2018-01-24  Jeff Law  <law@redhat.com>
21751         PR target/83994
21752         * i386.c (get_probe_interval): Move to earlier point.
21753         (ix86_compute_frame_layout): If -fstack-clash-protection and
21754         the frame is larger than the probe interval, then use pushes
21755         to save registers rather than reg->mem moves.
21756         (ix86_expand_prologue): Remove conditional for int_registers_saved
21757         assertion.
21759 2018-01-24  Vladimir Makarov  <vmakarov@redhat.com>
21761         PR target/84014
21762         * ira-build.c (setup_min_max_allocno_live_range_point): Set up
21763         min/max for never referenced object.
21765 2018-01-24  Jakub Jelinek  <jakub@redhat.com>
21767         PR middle-end/83977
21768         * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
21769         here.
21770         * omp-low.c (create_omp_child_function): Remove "omp declare simd"
21771         attributes from DECL_ATTRIBUTES (decl) without affecting
21772         DECL_ATTRIBUTES (current_function_decl).
21773         * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
21774         functions with non-NULL DECL_ABSTRACT_ORIGIN.
21776 2018-01-24  Richard Sandiford  <richard.sandiford@linaro.org>
21778         PR tree-optimization/83979
21779         * fold-const.c (fold_comparison): Use constant_boolean_node
21780         instead of boolean_{true,false}_node.
21782 2018-01-24  Jan Hubicka  <hubicka@ucw.cz>
21784         * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
21785         with zero counts.
21787 2018-01-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21789         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
21790         Simplify the clause that sets the length attribute.
21791         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
21792         (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
21793         clause that sets the length attribute.
21794         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
21796 2018-01-24  Tom de Vries  <tom@codesourcery.com>
21798         PR target/83589
21799         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
21800         (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
21801         Add strict parameter.
21802         (prevent_branch_around_nothing): Insert dummy insn between branch to
21803         label and label with no ptx insn inbetween.
21804         * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
21806 2018-01-24  Tom de Vries  <tom@codesourcery.com>
21808         PR target/81352
21809         * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
21810         for neutered threads in warp.
21811         * config/nvptx/nvptx.md (define_insn "exit"): New insn.
21813 2018-01-24  Richard Biener  <rguenther@suse.de>
21815         PR tree-optimization/83176
21816         * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
21817         operands.
21819 2018-01-24  Richard Biener  <rguenther@suse.de>
21821         PR tree-optimization/82819
21822         * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
21823         code generating pluses that are no-ops in the target precision.
21825 2018-01-24  Richard Biener  <rguenther@suse.de>
21827         PR middle-end/84000
21828         * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
21830 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
21832         * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
21833         to merge probabilities.
21834         * predict.c (probably_never_executed): Also mark as cold functions
21835         with global 0 profile and guessed local profile.
21836         * profile-count.c (profile_probability::combine_with_count): New
21837         member function.
21838         * profile-count.h (profile_probability::operator*,
21839         profile_probability::operator*=, profile_probability::operator/,
21840         profile_probability::operator/=): Reduce precision to adjusted
21841         and set value to guessed on contradictory divisions.
21842         (profile_probability::combine_with_freq): Remove.
21843         (profile_probability::combine_wiht_count): Declare.
21844         (profile_count::force_nonzero):: Set to adjusted.
21845         (profile_count::probability_in):: Set quality to adjusted.
21846         * tree-ssa-tail-merge.c (replace_block_by): Use
21847         combine_with_count.
21849 2018-01-23  Andrew Waterman  <andrew@sifive.com>
21850             Jim Wilson  <jimw@sifive.com>
21852         * config/riscv/riscv.c (riscv_stack_boundary): New.
21853         (riscv_option_override): Set riscv_stack_boundary.  Handle
21854         riscv_preferred_stack_boundary_arg.
21855         * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
21856         (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
21857         (STACK_BOUNDARY): Set to riscv_stack_boundary.
21858         (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
21859         * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
21860         * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
21862 2018-01-23  H.J. Lu  <hongjiu.lu@intel.com>
21864         PR target/83905
21865         * config/i386/i386.c (ix86_expand_prologue): Use cost reference
21866         of struct ix86_frame.
21867         (ix86_expand_epilogue): Likewise.  Add a local variable for
21868         the reg_save_offset field in struct ix86_frame.
21870 2018-01-23  Bin Cheng  <bin.cheng@arm.com>
21872         PR tree-optimization/82604
21873         * tree-loop-distribution.c (enum partition_kind): New enum item
21874         PKIND_PARTIAL_MEMSET.
21875         (partition_builtin_p): Support above new enum item.
21876         (generate_code_for_partition): Ditto.
21877         (compute_access_range): Differentiate cases that equality can be
21878         proven at all loops, the innermost loops or no loops.
21879         (classify_builtin_st, classify_builtin_ldst): Adjust call to above
21880         function.  Set PKIND_PARTIAL_MEMSET for partition appropriately.
21881         (finalize_partitions, distribute_loop): Don't fuse partition of
21882         PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
21883         (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
21884         parloop is enabled.
21886 2018-01-23  Martin Liska  <mliska@suse.cz>
21888         * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
21889         order to ignore the predictor.
21890         (PRED_POLYMORPHIC_CALL): Likewise.
21891         (PRED_RECURSIVE_CALL): Likewise.
21893 2018-01-23  Martin Liska  <mliska@suse.cz>
21895         * tree-profile.c (tree_profiling): Print function header to
21896         aware reader which function we are working on.
21897         * value-prof.c (gimple_find_values_to_profile): Do not print
21898         not interesting value histograms.
21900 2018-01-23  Martin Liska  <mliska@suse.cz>
21902         * profile-count.h (enum profile_quality): Add
21903         profile_uninitialized as the first value. Do not number values
21904         as they are zero based.
21905         (profile_count::verify): Update sanity check.
21906         (profile_probability::verify): Likewise.
21908 2018-01-23  Nathan Sidwell  <nathan@acm.org>
21910         * doc/invoke.texi (ffor-scope): Deprecate.
21912 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
21914         PR tree-optimization/83510
21915         * domwalk.c (set_all_edges_as_executable): New function.
21916         (dom_walker::dom_walker): Convert bool param
21917         "skip_unreachable_blocks" to enum reachability.  Move setup of
21918         edge flags to set_all_edges_as_executable and only do it when
21919         reachability is REACHABLE_BLOCKS.
21920         * domwalk.h (enum dom_walker::reachability): New enum.
21921         (dom_walker::dom_walker): Convert bool param
21922         "skip_unreachable_blocks" to enum reachability.
21923         (set_all_edges_as_executable): New decl.
21924         * graphite-scop-detection.c  (gather_bbs::gather_bbs): Convert
21925         from false for "skip_unreachable_blocks" to ALL_BLOCKS for
21926         "reachability".
21927         * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
21928         but converting true to REACHABLE_BLOCKS.
21929         * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
21930         * tree-vrp.c
21931         (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
21932         Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
21933         (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
21934         REACHABLE_BLOCKS.
21935         (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
21936         if check_all_array_refs will be called.
21938 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
21940         * tree.c (selftest::test_location_wrappers): Add more test
21941         coverage.
21943 2018-01-23  David Malcolm  <dmalcolm@redhat.com>
21945         * sbitmap.c (selftest::test_set_range): Fix memory leaks.
21946         (selftest::test_bit_in_range): Likewise.
21948 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
21950         PR testsuite/83888
21951         * doc/sourcebuild.texi (vect_float): Say that the selector
21952         only describes the situation when -funsafe-math-optimizations is on.
21953         (vect_float_strict): Document.
21955 2018-01-23  Richard Sandiford  <richard.sandiford@linaro.org>
21957         PR tree-optimization/83965
21958         * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
21959         (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
21960         instead of checking only for a reduction.
21961         (vect_recog_widen_sum_pattern): Likewise.
21963 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
21965         * predict.c (probably_never_executed): Only use precise profile info.
21966         (compute_function_frequency): Skip after inlining hack since we now
21967         have quality checking.
21969 2018-01-23  Jan Hubicka  <hubicka@ucw.cz>
21971         * profile-count.h (profile_probability::very_unlikely,
21972         profile_probability::unlikely, profile_probability::even): Set
21973         precision to guessed.
21975 2018-01-23  Richard Biener  <rguenther@suse.de>
21977         PR tree-optimization/83963
21978         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
21979         Properly terminate dominator walk when crossing the exit edge not
21980         when visiting its source block.
21982 2018-01-23  Jakub Jelinek  <jakub@redhat.com>
21984         PR c++/83918
21985         * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
21986         VIEW_CONVERT_EXPR to wrap CONST_DECLs.
21988 2018-01-22  Jakub Jelinek  <jakub@redhat.com>
21990         PR tree-optimization/83957
21991         * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs.  Remove
21992         semicolon after for body surrounded by braces.
21994         PR tree-optimization/83081
21995         * profile-count.h (profile_probability::split): New method.
21996         * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
21997         Use profile_probability::split.
21998         (do_compare_rtx_and_jump): Fix adjustment of probabilities
21999         when splitting a single conditional jump into 2.
22001 2018-01-22  David Malcolm  <dmalcolm@redhat.com>
22003         PR tree-optimization/69452
22004         * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
22005         decl.
22007 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22009         * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
22010         * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
22011         * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
22013 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22015         * config/rl78/rl78-protos.h (rl78_split_movdi): New function
22016         declaration.
22017         * config/rl78/rl78.md (movdi): New define_expand.
22018         * config/rl78/rl78.c (rl78_split_movdi): New function.
22020 2018-01-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
22022         PR target/83862
22023         * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
22024         no longer used.
22025         * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
22026         * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
22027         128-bit to produce an UNSPEC move to get the double word with the
22028         signbit and then a shift directly to do signbit.
22029         (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
22030         implementation with a new version that just does either a direct
22031         move or a regular move.  Move memory interface to separate insns.
22032         Move insns so they are next to the expander.
22033         (signbit<mode>2_dm_mem_be): New combiner insns to combine load
22034         with signbit move.  Split big and little endian case.
22035         (signbit<mode>2_dm_mem_le): Likewise.
22036         (signbit<mode>2_dm_<su>ext): Delete, no longer used.
22037         (signbit<mode>2_dm2): Likewise.
22039 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22041         * config/rl78/rl78.md (anddi3): New define_expand.
22043 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22045         * config/rl78/rl78.md (umindi3): New define_expand.
22047 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22049         * config/rl78/rl78.md (smindi3): New define_expand.
22051 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22053         * config/rl78/rl78.md (smaxdi3): New define_expand.
22055 2018-01-22  Carl Love  <cel@us.ibm.com>
22057         * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
22058         LVX_V1TI): Add macro expansion.
22059         * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
22060         definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
22061         VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
22062         * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
22063         Change check to determine if the instruction is a byte reversing
22064         entry.  Fix typo in comment.
22065         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
22066         for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
22067         Add def_builtin calls for new builtins.
22068         * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
22069         Add define_insn expansion.
22071 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22073         * config/rl78/rl78.md (umaxdi3): New define_expand.
22075 2018-01-22  Sebastian Perta  <sebastian.perta@renesas.com>
22077         * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
22078         for non-QImode registers.
22080 2018-01-22  Richard Biener  <rguenther@suse.de>
22082         PR tree-optimization/83963
22083         * graphite-scop-detection.c (scop_detection::get_sese): Delay
22084         including the loop exit block.
22085         (scop_detection::merge_sese): Likewise.
22086         (scop_detection::add_scop): Do it here instead.
22088 2018-01-22  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22090         * doc/sourcebuild.texi (arm_softfloat): Document.
22092 2018-01-21  John David Anglin  <danglin@gcc.gnu.org>
22094         PR gcc/77734
22095         * config/pa/pa.c (pa_function_ok_for_sibcall): Use
22096         targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
22097         Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
22099 2018-01-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22100             David Edelsohn  <dje.gcc@gmail.com>
22102         PR target/83946
22103         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22104         Change "crset eq" to "crset 2".
22105         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22106         (*call_indirect_aix<mode>_nospec): Likewise.
22107         (*call_value_indirect_aix<mode>_nospec): Likewise.
22108         (*call_indirect_elfv2<mode>_nospec): Likewise.
22109         (*call_value_indirect_elfv2<mode>_nospec): Likewise.
22110         (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
22111         change assembly output from . to $.
22112         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22113         (indirect_jump<mode>_nospec): Change assembly output from . to $.
22114         (*tablejump<mode>_internal1_nospec): Likewise.
22116 2018-01-21  Oleg Endo  <olegendo@gcc.gnu.org>
22118         PR target/80870
22119         * config/sh/sh_optimize_sett_clrt.cc:
22120         Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
22122 2018-01-20  Richard Sandiford  <richard.sandiford@linaro.org>
22124         PR tree-optimization/83940
22125         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
22126         offset_dt to vect_constant_def rather than vect_unknown_def_type.
22127         (vect_check_load_store_mask): Add a mask_dt_out parameter and
22128         use it to pass back the definition type.
22129         (vect_check_store_rhs): Likewise rhs_dt_out.
22130         (vect_build_gather_load_calls): Add a mask_dt argument and use
22131         it instead of a call to vect_is_simple_use.
22132         (vectorizable_store): Update calls to vect_check_load_store_mask
22133         and vect_check_store_rhs.  Use the dt returned by the latter instead
22134         of scatter_src_dt.  Use the cached mask_dt and gs_info.offset_dt
22135         instead of calls to vect_is_simple_use.  Pass the scalar rather
22136         than the vector operand to vect_is_simple_use when handling
22137         second and subsequent copies of an rhs value.
22138         (vectorizable_load): Update calls to vect_check_load_store_mask
22139         and vect_build_gather_load_calls.  Use the cached mask_dt and
22140         gs_info.offset_dt instead of calls to vect_is_simple_use.
22142 2018-01-20  Jakub Jelinek  <jakub@redhat.com>
22144         PR middle-end/83945
22145         * tree-emutls.c: Include gimplify.h.
22146         (lower_emutls_2): New function.
22147         (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
22148         with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
22149         it before further processing.
22151         PR target/83930
22152         * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
22153         UINTVAL (trueop1) instead of INTVAL (op1).
22155 2018-01-19  Jakub Jelinek  <jakub@redhat.com>
22157         PR debug/81570
22158         PR debug/83728
22159         * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
22160         INCOMING_FRAME_SP_OFFSET if not defined.
22161         (scan_trace): Add ENTRY argument.  If true and
22162         DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
22163         emit a note to adjust the CFA offset.
22164         (create_cfi_notes): Adjust scan_trace callers.
22165         (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
22166         INCOMING_FRAME_SP_OFFSET in the CIE.
22167         * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
22168         * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
22169         Likewise.
22170         * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
22171         * doc/tm.texi: Regenerated.
22173 2018-01-19  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22175         PR rtl-optimization/83147
22176         * lra-constraints.c (remove_inheritance_pseudos): Use
22177         lra_substitute_pseudo_within_insn.
22179 2018-01-19  Tom de Vries  <tom@codesourcery.com>
22180             Cesar Philippidis  <cesar@codesourcery.com>
22182         PR target/83920
22183         * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
22185 2018-01-19  Cesar Philippidis  <cesar@codesourcery.com>
22187         PR target/83790
22188         * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
22189         spaces for function labels.
22191 2018-01-19  Martin Liska  <mliska@suse.cz>
22193         * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
22194         (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
22195         (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
22196         (PRED_OPCODE_POSITIVE): Change from 64 to 59.
22197         (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
22198         (PRED_CONST_RETURN): Change from 69 to 65.
22199         (PRED_NULL_RETURN): Change from 91 to 71.
22200         (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
22201         (PRED_LOOP_GUARD): Change from 66 to 73.
22203 2018-01-19  Martin Liska  <mliska@suse.cz>
22205         * predict.c (predict_insn_def): Add new assert.
22206         (struct branch_predictor): Change type to signed integer.
22207         (test_prediction_value_range): Amend test to cover
22208         PROB_UNINITIALIZED.
22209         * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
22210         (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
22211         (PRED_LOOP_ITERATIONS_MAX): Likewise.
22212         (PRED_LOOP_IV_COMPARE): Likewise.
22213         * predict.h (PROB_UNINITIALIZED): Define new constant.
22215 2018-01-19  Martin Liska  <mliska@suse.cz>
22217         * predict.c (dump_prediction): Add new format for
22218         analyze_brprob.py script which is enabled with -details
22219         suboption.
22220         * profile-count.h (precise_p): New function.
22222 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
22224         PR tree-optimization/83922
22225         * tree-vect-loop.c (vect_verify_full_masking): Return false if
22226         there are no statements that need masking.
22227         (vect_active_double_reduction_p): New function.
22228         (vect_analyze_loop_operations): Use it when handling phis that
22229         are not in the loop header.
22231 2018-01-19  Richard Sandiford  <richard.sandiford@linaro.org>
22233         PR tree-optimization/83914
22234         * tree-vect-loop.c (vectorizable_induction): Don't convert
22235         init_expr or apply the peeling adjustment for inductions
22236         that are nested within the vectorized loop.
22238 2018-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22240         * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
22241         instead of NEG.
22243 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
22245         PR sanitizer/81715
22246         PR testsuite/83882
22247         * function.h (gimplify_parameters): Add gimple_seq * argument.
22248         * function.c: Include gimple.h and options.h.
22249         (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
22250         for the added local temporaries if needed.
22251         * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
22252         if there are any parameter cleanups, wrap whole body into a
22253         try/finally with the cleanups.
22255 2018-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
22257         PR target/82964
22258         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
22259         Use GET_MODE_CLASS for scalar floating point.
22261 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
22263         PR ipa/82256
22264         patch by PaX Team
22265         * cgraphclones.c (cgraph_node::create_version_clone_with_body):
22266         Fix call of call_cgraph_insertion_hooks.
22268 2018-01-18  Martin Sebor  <msebor@redhat.com>
22270         * doc/invoke.texi (-Wclass-memaccess): Tweak text.
22272 2018-01-18  Jan Hubicka  <hubicka@ucw.cz>
22274         PR ipa/83619
22275         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
22276         frequencies.
22278 2018-01-18  Boris Kolpackov  <boris@codesynthesis.com>
22280         PR other/70268
22281         * common.opt: (-ffile-prefix-map): New option.
22282         * opts.c (common_handle_option): Defer it.
22283         * opts-global.c (handle_common_deferred_options): Handle it.
22284         * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
22285         * file-prefix-map.h: New file.
22286         (remap_debug_filename, add_debug_prefix_map): ...here.
22287         (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
22288         * final.c (debug_prefix_map, add_debug_prefix_map
22289         remap_debug_filename): Move to...
22290         * file-prefix-map.c: New file.
22291         (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
22292         generalize, get rid of alloca(), use strrchr() instead of strchr().
22293         (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
22294         Implement in terms of add_prefix_map().
22295         (remap_macro_filename, remap_debug_filename): Implement in term of
22296         remap_filename().
22297         * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
22298         * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
22299         * dbxout.c: Include file-prefix-map.h.
22300         * varasm.c: Likewise.
22301         * vmsdbgout.c: Likewise.
22302         * xcoffout.c: Likewise.
22303         * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
22304         * doc/cppopts.texi (-fmacro-prefix-map): Document.
22305         * doc/invoke.texi (-ffile-prefix-map): Document.
22306         (-fdebug-prefix-map): Update description.
22308 2018-01-18  Martin Liska  <mliska@suse.cz>
22310         * config/i386/i386.c (indirect_thunk_name): Document that also
22311         lfence is emitted.
22312         (output_indirect_thunk): Document why both instructions
22313         (pause and lfence) are generated.
22315 2018-01-18  Richard Biener  <rguenther@suse.de>
22317         PR tree-optimization/83887
22318         * graphite-scop-detection.c
22319         (scop_detection::get_nearest_dom_with_single_entry): Remove.
22320         (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
22321         (scop_detection::merge_sese): Re-implement with a flood-fill
22322         algorithm that properly finds a SESE region if it exists.
22324 2018-01-18  Jakub Jelinek  <jakub@redhat.com>
22326         PR c/61240
22327         * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
22328         pointer_diff optimizations use view_convert instead of convert.
22330 2018-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22332         * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
22333         Generate different code for -mno-speculate-indirect-jumps.
22334         (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
22335         (*call_indirect_aix<mode>): Disable for
22336         -mno-speculate-indirect-jumps.
22337         (*call_indirect_aix<mode>_nospec): New define_insn.
22338         (*call_value_indirect_aix<mode>): Disable for
22339         -mno-speculate-indirect-jumps.
22340         (*call_value_indirect_aix<mode>_nospec): New define_insn.
22341         (*sibcall_nonlocal_sysv<mode>): Generate different code for
22342         -mno-speculate-indirect-jumps.
22343         (*sibcall_value_nonlocal_sysv<mode>): Likewise.
22345 2018-01-17  Michael Meissner  <meissner@linux.vnet.ibm.com>
22347         * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
22348         long double type, set the flags for noting the default long double
22349         type, even if we don't pass or return a long double type.
22351 2018-01-17  Jan Hubicka  <hubicka@ucw.cz>
22353         PR ipa/83051
22354         * ipa-inline.c (flatten_function): Do not overwrite final inlining
22355         failure.
22357 2018-01-17  Will Schmidt  <will_schmidt@vnet.ibm.com>
22359         * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
22360         support for merge[hl].
22361         (fold_mergehl_helper): New helper function.
22362         (tree-vector-builder.h): New #include for tree_vector_builder usage.
22363         * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
22364         (altivec_vmrglw_direct): Add xxmrglw insn.
22366 2018-01-17  Andrew Waterman  <andrew@sifive.com>
22368         * config/riscv/riscv.c (riscv_conditional_register_usage): If
22369         UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
22371 2018-01-17  David Malcolm  <dmalcolm@redhat.com>
22373         PR lto/83121
22374         * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
22375         call the lto_location_cache before reading the
22376         DECL_SOURCE_LOCATION of the types.
22378 2018-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
22379             Richard Sandiford  <richard.sandiford@linaro.org>
22381         * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
22382         * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
22383         (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
22384         SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
22385         * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
22386         Add declaration.
22387         * config/aarch64/constraints.md (aarch64_movti_operand):
22388         Limit immediates.
22389         * config/aarch64/predicates.md (Uti): Add new constraint.
22391 2018-01-17  Carl Love  <cel@us.ibm.com>
22393         * config/rs6000/vsx.md (define_expand xl_len_r,
22394         define_expand stxvl, define_expand *stxvl): Add match_dup argument.
22395         (define_insn): Add, match_dup 1 argument to define_insn stxvll and
22396         lxvll.
22397         (define_expand, define_insn): Move the shift left from  the
22398         define_insn to the define_expand for lxvl and stxvl instructions.
22399         * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
22400         and XL_LEN_R definitions to PURE.
22402 2018-01-17  Uros Bizjak  <ubizjak@gmail.com>
22404         * config/i386/i386.c (indirect_thunk_name): Declare regno
22405         as unsigned int.  Compare regno with INVALID_REGNUM.
22406         (output_indirect_thunk): Ditto.
22407         (output_indirect_thunk_function): Ditto.
22408         (ix86_code_end): Declare regno as unsigned int.  Use INVALID_REGNUM
22409         in the call to output_indirect_thunk_function.
22411 2018-01-17  Richard Sandiford  <richard.sandiford@linaro.org>
22413         PR middle-end/83884
22414         * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
22415         rather than the size of inner_type to determine the stack slot size
22416         when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
22418 2018-01-16  Sebastian Peryt  <sebastian.peryt@intel.com>
22420         PR target/83546
22421         * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
22422         to PTA_SILVERMONT.
22424 2018-01-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
22426         * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
22427         endian Linux systems to optionally enable multilibs for selecting
22428         the long double type if the user configured an explicit type.
22429         * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
22430         have no long double multilibs if not defined.
22431         * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
22432         warn if the user used -mabi={ieee,ibm}longdouble and we built
22433         multilibs for long double.
22434         * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
22435         appropriate multilib option.
22436         (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
22437         multilib options.
22438         * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
22439         for building long double multilibs.
22440         * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
22442 2018-01-16  John David Anglin  <danglin@gcc.gnu.org>
22444         * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
22445         copies.
22447         * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
22448         64 bits.
22449         * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
22450         128 bits.
22452         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
22453         variables.
22455         * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
22456         return value.
22458 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
22460         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
22461         ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
22463 2018-01-16  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22465         * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
22466         different rtl trees depending on TARGET_64BIT.
22467         (rs6000_gen_lvx): Likewise.
22469 2018-01-16  Eric Botcazou  <ebotcazou@adacore.com>
22471         * config/visium/visium.md (nop): Tweak comment.
22472         (hazard_nop): Likewise.
22474 2018-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22476         * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
22477         -mspeculate-indirect-jumps.
22478         * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
22479         for -mno-speculate-indirect-jumps.
22480         (*call_indirect_elfv2<mode>_nospec): New define_insn.
22481         (*call_value_indirect_elfv2<mode>): Disable for
22482         -mno-speculate-indirect-jumps.
22483         (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
22484         (indirect_jump): Emit different RTL for
22485         -mno-speculate-indirect-jumps.
22486         (*indirect_jump<mode>): Disable for
22487         -mno-speculate-indirect-jumps.
22488         (*indirect_jump<mode>_nospec): New define_insn.
22489         (tablejump): Emit different RTL for
22490         -mno-speculate-indirect-jumps.
22491         (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
22492         (tablejumpsi_nospec): New define_expand.
22493         (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
22494         (tablejumpdi_nospec): New define_expand.
22495         (*tablejump<mode>_internal1): Disable for
22496         -mno-speculate-indirect-jumps.
22497         (*tablejump<mode>_internal1_nospec): New define_insn.
22498         * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
22499         option.
22501 2018-01-16  Artyom Skrobov tyomitch@gmail.com
22503         * caller-save.c (insert_save): Drop unnecessary parameter.  All
22504         callers updated.
22506 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
22507             Richard Biener  <rguenth@suse.de>
22509         PR libgomp/83590
22510         * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
22511         return early, inline manually is_gimple_sizepos.  Make sure if we
22512         call gimplify_expr we don't end up with a gimple constant.
22513         * tree.c (variably_modified_type_p): Don't return true for
22514         is_gimple_constant (_t).  Inline manually is_gimple_sizepos.
22515         * gimplify.h (is_gimple_sizepos): Remove.
22517 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
22519         PR tree-optimization/83857
22520         * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
22521         vectorizable_live_operation for pure SLP statements.
22522         (vectorizable_live_operation): Handle PHIs.
22524 2018-01-16  Richard Biener  <rguenther@suse.de>
22526         PR tree-optimization/83867
22527         * tree-vect-stmts.c (vect_transform_stmt): Precompute
22528         nested_in_vect_loop_p since the scalar stmt may get invalidated.
22530 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
22532         PR c/83844
22533         * stor-layout.c (handle_warn_if_not_align): Use byte_position and
22534         multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
22535         If off is not INTEGER_CST, issue a may not be aligned warning
22536         rather than isn't aligned.  Use isn%'t rather than isn't.
22537         * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
22538         into MULT_EXPR.
22539         <case MULT_EXPR>: Improve the case when bottom and one of the
22540         MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
22541         operand, in that case check if the other operand is multiple of
22542         bottom divided by the INTEGER_CST operand.
22544 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
22546         PR target/83858
22547         * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
22548         * config/pa/pa-protos.h (pa_function_arg_size): Declare.
22549         * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
22550         pa_function_arg_size instead of FUNCTION_ARG_SIZE.
22551         * config/pa/pa.c (pa_function_arg_advance): Likewise.
22552         (pa_function_arg, pa_arg_partial_bytes): Likewise.
22553         (pa_function_arg_size): New function.
22555 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
22557         * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
22558         in a separate statement.
22560 2018-01-16  Richard Sandiford  <richard.sandiford@linaro.org>
22562         PR tree-optimization/83847
22563         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
22564         group gathers and scatters.
22566 2018-01-16  Jakub Jelinek  <jakub@redhat.com>
22568         PR rtl-optimization/86620
22569         * params.def (max-sched-ready-insns): Bump minimum value to 1.
22571         PR rtl-optimization/83213
22572         * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
22573         to last if both are JUMP_INSNs.
22575         PR tree-optimization/83843
22576         * gimple-ssa-store-merging.c
22577         (imm_store_chain_info::output_merged_store): Handle bit_not_p on
22578         store_immediate_info for bswap/nop orig_stores.
22580 2018-01-15  Andrew Waterman  <andrew@sifive.com>
22582         * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
22583         !TARGET_MUL.
22584         <UDIV>: Increase cost if !TARGET_DIV.
22586 2018-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
22588         * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
22589         (define_attr "cr_logical_3op"): New.
22590         (cceq_ior_compare): Adjust.
22591         (cceq_ior_compare_complement): Adjust.
22592         (*cceq_rev_compare): Adjust.
22593         * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
22594         (is_cracked_insn): Adjust.
22595         (insn_must_be_first_in_group): Adjust.
22596         * config/rs6000/40x.md: Adjust.
22597         * config/rs6000/440.md: Adjust.
22598         * config/rs6000/476.md: Adjust.
22599         * config/rs6000/601.md: Adjust.
22600         * config/rs6000/603.md: Adjust.
22601         * config/rs6000/6xx.md: Adjust.
22602         * config/rs6000/7450.md: Adjust.
22603         * config/rs6000/7xx.md: Adjust.
22604         * config/rs6000/8540.md: Adjust.
22605         * config/rs6000/cell.md: Adjust.
22606         * config/rs6000/e300c2c3.md: Adjust.
22607         * config/rs6000/e500mc.md: Adjust.
22608         * config/rs6000/e500mc64.md: Adjust.
22609         * config/rs6000/e5500.md: Adjust.
22610         * config/rs6000/e6500.md: Adjust.
22611         * config/rs6000/mpc.md: Adjust.
22612         * config/rs6000/power4.md: Adjust.
22613         * config/rs6000/power5.md: Adjust.
22614         * config/rs6000/power6.md: Adjust.
22615         * config/rs6000/power7.md: Adjust.
22616         * config/rs6000/power8.md: Adjust.
22617         * config/rs6000/power9.md: Adjust.
22618         * config/rs6000/rs64.md: Adjust.
22619         * config/rs6000/titan.md: Adjust.
22621 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
22623         * config/i386/predicates.md (indirect_branch_operand): Rewrite
22624         ix86_indirect_branch_register logic.
22626 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
22628         * config/i386/constraints.md (Bs): Update
22629         ix86_indirect_branch_register check.  Don't check
22630         ix86_indirect_branch_register with GOT_memory_operand.
22631         (Bw): Likewise.
22632         * config/i386/predicates.md (GOT_memory_operand): Don't check
22633         ix86_indirect_branch_register here.
22634         (GOT32_symbol_operand): Likewise.
22636 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
22638         * config/i386/predicates.md (constant_call_address_operand):
22639         Rewrite ix86_indirect_branch_register logic.
22640         (sibcall_insn_operand): Likewise.
22642 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
22644         * config/i386/constraints.md (Bs): Replace
22645         ix86_indirect_branch_thunk_register with
22646         ix86_indirect_branch_register.
22647         (Bw): Likewise.
22648         * config/i386/i386.md (indirect_jump): Likewise.
22649         (tablejump): Likewise.
22650         (*sibcall_memory): Likewise.
22651         (*sibcall_value_memory): Likewise.
22652         Peepholes of indirect call and jump via memory: Likewise.
22653         * config/i386/i386.opt: Likewise.
22654         * config/i386/predicates.md (indirect_branch_operand): Likewise.
22655         (GOT_memory_operand): Likewise.
22656         (call_insn_operand): Likewise.
22657         (sibcall_insn_operand): Likewise.
22658         (GOT32_symbol_operand): Likewise.
22660 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
22662         PR middle-end/83837
22663         * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
22664         type rather than type addr's type points to.
22665         (expand_omp_atomic_mutex): Likewise.
22666         (expand_omp_atomic): Likewise.
22668 2018-01-15  H.J. Lu  <hongjiu.lu@intel.com>
22670         PR target/83839
22671         * config/i386/i386.c (output_indirect_thunk_function): Use
22672         ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
22673         for  __x86_return_thunk.
22675 2018-01-15  Richard Biener  <rguenther@suse.de>
22677         PR middle-end/83850
22678         * expmed.c (extract_bit_field_1): Fix typo.
22680 2018-01-15  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
22682         PR target/83687
22683         * config/arm/iterators.md (VF): New mode iterator.
22684         * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
22685         Remove integer-related logic from pattern.
22686         (neon_vabd<mode>_3): Likewise.
22688 2018-01-15  Jakub Jelinek  <jakub@redhat.com>
22690         PR middle-end/82694
22691         * common.opt (fstrict-overflow): No longer an alias.
22692         (fwrapv-pointer): New option.
22693         * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
22694         also for pointer types based on flag_wrapv_pointer.
22695         * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
22696         opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
22697         opts->x_flag_wrapv got set.
22698         * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
22699         changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
22700         POINTER_TYPE_OVERFLOW_UNDEFINED.
22701         * match.pd: Likewise in address comparison pattern.
22702         * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
22704 2018-01-15  Richard Biener  <rguenther@suse.de>
22706         PR lto/83804
22707         * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
22708         from TYPE_FIELDS.  Free TYPE_BINFO if not used by devirtualization.
22709         Reset type names to their identifier if their TYPE_DECL doesn't
22710         have linkage (and thus is used for ODR and devirt).
22711         (save_debug_info_for_decl): Remove.
22712         (save_debug_info_for_type): Likewise.
22713         (add_tree_to_fld_list): Adjust.
22714         * tree-pretty-print.c (dump_generic_node): Make dumping of
22715         type names more robust.
22717 2018-01-15  Richard Biener  <rguenther@suse.de>
22719         * BASE-VER: Bump to 8.0.1.
22721 2018-01-14  Martin Sebor  <msebor@redhat.com>
22723         PR other/83508
22724         * builtins.c (check_access): Avoid warning when the no-warning bit
22725         is set.
22727 2018-01-14  Cory Fields  <cory-nospam-@coryfields.com>
22729         * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
22730         * ira-color (allocno_hard_regs_compare): Likewise.
22732 2018-01-14  Nathan Rossi  <nathan@nathanrossi.com>
22734         PR target/83013
22735         * config/microblaze/microblaze.c (microblaze_asm_output_ident):
22736         Use .pushsection/.popsection.
22738 2018-01-14  Martin Sebor  <msebor@redhat.com>
22740         PR c++/81327
22741         * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
22743 2018-01-14  Jakub Jelinek  <jakub@redhat.com>
22745         * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
22746         entry from extra_headers.
22747         (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
22748         extra_headers, make the list bitwise identical to the i?86-*-* one.
22750 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
22752         * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
22753         -mcmodel=large with -mindirect-branch=thunk,
22754         -mindirect-branch=thunk-extern, -mfunction-return=thunk and
22755         -mfunction-return=thunk-extern.
22756         * doc/invoke.texi: Document -mcmodel=large is incompatible with
22757         -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
22758         -mfunction-return=thunk and -mfunction-return=thunk-extern.
22760 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
22762         * config/i386/i386.c (print_reg): Print the name of the full
22763         integer register without '%'.
22764         (ix86_print_operand): Handle 'V'.
22765         * doc/extend.texi: Document 'V' modifier.
22767 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
22769         * config/i386/constraints.md (Bs): Disallow memory operand for
22770         -mindirect-branch-register.
22771         (Bw): Likewise.
22772         * config/i386/predicates.md (indirect_branch_operand): Likewise.
22773         (GOT_memory_operand): Likewise.
22774         (call_insn_operand): Likewise.
22775         (sibcall_insn_operand): Likewise.
22776         (GOT32_symbol_operand): Likewise.
22777         * config/i386/i386.md (indirect_jump): Call convert_memory_address
22778         for -mindirect-branch-register.
22779         (tablejump): Likewise.
22780         (*sibcall_memory): Likewise.
22781         (*sibcall_value_memory): Likewise.
22782         Disallow peepholes of indirect call and jump via memory for
22783         -mindirect-branch-register.
22784         (*call_pop): Replace m with Bw.
22785         (*call_value_pop): Likewise.
22786         (*sibcall_pop_memory): Replace m with Bs.
22787         * config/i386/i386.opt (mindirect-branch-register): New option.
22788         * doc/invoke.texi: Document -mindirect-branch-register option.
22790 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
22792         * config/i386/i386-protos.h (ix86_output_function_return): New.
22793         * config/i386/i386.c (ix86_set_indirect_branch_type): Also
22794         set function_return_type.
22795         (indirect_thunk_name): Add ret_p to indicate thunk for function
22796         return.
22797         (output_indirect_thunk_function): Pass false to
22798         indirect_thunk_name.
22799         (ix86_output_indirect_branch_via_reg): Likewise.
22800         (ix86_output_indirect_branch_via_push): Likewise.
22801         (output_indirect_thunk_function): Create alias for function
22802         return thunk if regno < 0.
22803         (ix86_output_function_return): New function.
22804         (ix86_handle_fndecl_attribute): Handle function_return.
22805         (ix86_attribute_table): Add function_return.
22806         * config/i386/i386.h (machine_function): Add
22807         function_return_type.
22808         * config/i386/i386.md (simple_return_internal): Use
22809         ix86_output_function_return.
22810         (simple_return_internal_long): Likewise.
22811         * config/i386/i386.opt (mfunction-return=): New option.
22812         (indirect_branch): Mention -mfunction-return=.
22813         * doc/extend.texi: Document function_return function attribute.
22814         * doc/invoke.texi: Document -mfunction-return= option.
22816 2018-01-14  H.J. Lu  <hongjiu.lu@intel.com>
22818         * config/i386/i386-opts.h (indirect_branch): New.
22819         * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
22820         * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
22821         with local indirect jump when converting indirect call and jump.
22822         (ix86_set_indirect_branch_type): New.
22823         (ix86_set_current_function): Call ix86_set_indirect_branch_type.
22824         (indirectlabelno): New.
22825         (indirect_thunk_needed): Likewise.
22826         (indirect_thunk_bnd_needed): Likewise.
22827         (indirect_thunks_used): Likewise.
22828         (indirect_thunks_bnd_used): Likewise.
22829         (INDIRECT_LABEL): Likewise.
22830         (indirect_thunk_name): Likewise.
22831         (output_indirect_thunk): Likewise.
22832         (output_indirect_thunk_function): Likewise.
22833         (ix86_output_indirect_branch_via_reg): Likewise.
22834         (ix86_output_indirect_branch_via_push): Likewise.
22835         (ix86_output_indirect_branch): Likewise.
22836         (ix86_output_indirect_jmp): Likewise.
22837         (ix86_code_end): Call output_indirect_thunk_function if needed.
22838         (ix86_output_call_insn): Call ix86_output_indirect_branch if
22839         needed.
22840         (ix86_handle_fndecl_attribute): Handle indirect_branch.
22841         (ix86_attribute_table): Add indirect_branch.
22842         * config/i386/i386.h (machine_function): Add indirect_branch_type
22843         and has_local_indirect_jump.
22844         * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
22845         to true.
22846         (tablejump): Likewise.
22847         (*indirect_jump): Use ix86_output_indirect_jmp.
22848         (*tablejump_1): Likewise.
22849         (simple_return_indirect_internal): Likewise.
22850         * config/i386/i386.opt (mindirect-branch=): New option.
22851         (indirect_branch): New.
22852         (keep): Likewise.
22853         (thunk): Likewise.
22854         (thunk-inline): Likewise.
22855         (thunk-extern): Likewise.
22856         * doc/extend.texi: Document indirect_branch function attribute.
22857         * doc/invoke.texi: Document -mindirect-branch= option.
22859 2018-01-14  Jan Hubicka  <hubicka@ucw.cz>
22861         PR ipa/83051
22862         * ipa-inline.c (edge_badness): Tolerate roundoff errors.
22864 2018-01-14  Richard Sandiford  <richard.sandiford@linaro.org>
22866         * ipa-inline.c (want_inline_small_function_p): Return false if
22867         inlining has already failed with CIF_FINAL_ERROR.
22868         (update_caller_keys): Call want_inline_small_function_p before
22869         can_inline_edge_p.
22870         (update_callee_keys): Likewise.
22872 2018-01-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
22874         * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
22875         New function.
22876         (rs6000_quadword_masked_address_p): Likewise.
22877         (quad_aligned_load_p): Likewise.
22878         (quad_aligned_store_p): Likewise.
22879         (const_load_sequence_p): Add comment to describe the outer-most loop.
22880         (mimic_memory_attributes_and_flags): New function.
22881         (rs6000_gen_stvx): Likewise.
22882         (replace_swapped_aligned_store): Likewise.
22883         (rs6000_gen_lvx): Likewise.
22884         (replace_swapped_aligned_load): Likewise.
22885         (replace_swapped_load_constant): Capitalize argument name in
22886         comment describing this function.
22887         (rs6000_analyze_swaps): Add a third pass to search for vector loads
22888         and stores that access quad-word aligned addresses and replace
22889         with stvx or lvx instructions when appropriate.
22890         * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
22891         New function prototype.
22892         (rs6000_quadword_masked_address_p): Likewise.
22893         (rs6000_gen_lvx): Likewise.
22894         (rs6000_gen_stvx): Likewise.
22895         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
22896         VSX_D (V2DF, V2DI), modify this split to select lvx instruction
22897         when memory address is aligned.
22898         (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
22899         this split to select lvx instruction when memory address is aligned.
22900         (*vsx_le_perm_load_v8hi): Modify this split to select lvx
22901         instruction when memory address is aligned.
22902         (*vsx_le_perm_load_v16qi): Likewise.
22903         (four unnamed splitters): Modify to select the stvx instruction
22904         when memory is aligned.
22906 2018-01-13  Jan Hubicka  <hubicka@ucw.cz>
22908         * predict.c (determine_unlikely_bbs): Handle correctly BBs
22909         which appears in the queue multiple times.
22911 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22912             Alan Hayward  <alan.hayward@arm.com>
22913             David Sherwood  <david.sherwood@arm.com>
22915         * tree-vectorizer.h (vec_lower_bound): New structure.
22916         (_loop_vec_info): Add check_nonzero and lower_bounds.
22917         (LOOP_VINFO_CHECK_NONZERO): New macro.
22918         (LOOP_VINFO_LOWER_BOUNDS): Likewise.
22919         (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
22920         * tree-data-ref.h (dr_with_seg_len): Add access_size and align
22921         fields.  Make seg_len the distance travelled, not including the
22922         access size.
22923         (dr_direction_indicator): Declare.
22924         (dr_zero_step_indicator): Likewise.
22925         (dr_known_forward_stride_p): Likewise.
22926         * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
22927         tree-ssanames.h.
22928         (runtime_alias_check_p): Allow runtime alias checks with
22929         variable strides.
22930         (operator ==): Compare access_size and align.
22931         (prune_runtime_alias_test_list): Rework for new distinction between
22932         the access_size and seg_len.
22933         (create_intersect_range_checks_index): Likewise.  Cope with polynomial
22934         segment lengths.
22935         (get_segment_min_max): New function.
22936         (create_intersect_range_checks): Use it.
22937         (dr_step_indicator): New function.
22938         (dr_direction_indicator): Likewise.
22939         (dr_zero_step_indicator): Likewise.
22940         (dr_known_forward_stride_p): Likewise.
22941         * tree-loop-distribution.c (data_ref_segment_size): Return
22942         DR_STEP * (niters - 1).
22943         (compute_alias_check_pairs): Update call to the dr_with_seg_len
22944         constructor.
22945         * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
22946         (vect_preserves_scalar_order_p): New function, split out from...
22947         (vect_analyze_data_ref_dependence): ...here.  Check for zero steps.
22948         (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
22949         (vect_vfa_access_size): New function.
22950         (vect_vfa_align): Likewise.
22951         (vect_compile_time_alias): Take access_size_a and access_b arguments.
22952         (dump_lower_bound): New function.
22953         (vect_check_lower_bound): Likewise.
22954         (vect_small_gap_p): Likewise.
22955         (vectorizable_with_step_bound_p): Likewise.
22956         (vect_prune_runtime_alias_test_list): Ignore cross-iteration
22957         depencies if the vectorization factor is 1.  Convert the checks
22958         for nonzero steps into checks on the bounds of DR_STEP.  Try using
22959         a bunds check for variable steps if the minimum required step is
22960         relatively small. Update calls to the dr_with_seg_len
22961         constructor and to vect_compile_time_alias.
22962         * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
22963         function.
22964         (vect_loop_versioning): Call it.
22965         * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
22966         when retrying.
22967         (vect_estimate_min_profitable_iters): Account for any bounds checks.
22969 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
22970             Alan Hayward  <alan.hayward@arm.com>
22971             David Sherwood  <david.sherwood@arm.com>
22973         * doc/sourcebuild.texi (vect_scatter_store): Document.
22974         * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
22975         optabs.
22976         * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
22977         Document.
22978         * genopinit.c (main): Add supports_vec_scatter_store and
22979         supports_vec_scatter_store_cached to target_optabs.
22980         * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
22981         IFN_MASK_SCATTER_STORE.
22982         * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
22983         functions.
22984         * internal-fn.h (internal_store_fn_p): Declare.
22985         (internal_fn_stored_value_index): Likewise.
22986         * internal-fn.c (scatter_store_direct): New macro.
22987         (expand_scatter_store_optab_fn): New function.
22988         (direct_scatter_store_optab_supported_p): New macro.
22989         (internal_store_fn_p): New function.
22990         (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
22991         IFN_MASK_SCATTER_STORE.
22992         (internal_fn_mask_index): Likewise.
22993         (internal_fn_stored_value_index): New function.
22994         (internal_gather_scatter_fn_supported_p): Adjust operand numbers
22995         for scatter stores.
22996         * optabs-query.h (supports_vec_scatter_store_p): Declare.
22997         * optabs-query.c (supports_vec_scatter_store_p): New function.
22998         * tree-vectorizer.h (vect_get_store_rhs): Declare.
22999         * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
23000         true for scatter stores.
23001         (vect_gather_scatter_fn_p): Handle scatter stores too.
23002         (vect_check_gather_scatter): Consider using scatter stores if
23003         supports_vec_scatter_store_p.
23004         * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
23005         scatter stores too.
23006         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23007         internal_fn_stored_value_index.
23008         (check_load_store_masking): Handle scatter stores too.
23009         (vect_get_store_rhs): Make public.
23010         (vectorizable_call): Use internal_store_fn_p.
23011         (vectorizable_store): Handle scatter store internal functions.
23012         (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
23013         when deciding whether the end of the group has been reached.
23014         * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
23015         * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
23016         (mask_scatter_store<mode>): New insns.
23018 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23019             Alan Hayward  <alan.hayward@arm.com>
23020             David Sherwood  <david.sherwood@arm.com>
23022         * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
23023         * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
23024         * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
23025         function.
23026         (vect_use_strided_gather_scatters_p): Take a masked_p argument.
23027         Use vect_truncate_gather_scatter_offset if we can't treat the
23028         operation as a normal gather load or scatter store.
23029         (get_group_load_store_type): Take the gather_scatter_info
23030         as argument.  Try using a gather load or scatter store for
23031         single-element groups.
23032         (get_load_store_type): Update calls to get_group_load_store_type
23033         and vect_use_strided_gather_scatters_p.
23035 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23036             Alan Hayward  <alan.hayward@arm.com>
23037             David Sherwood  <david.sherwood@arm.com>
23039         * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
23040         optional tree argument.
23041         * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
23042         null target hooks.
23043         (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
23044         but continue to use the current value as a fallback.
23045         (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
23046         to compare the updates.
23047         * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
23048         (get_load_store_type): Use it when handling a strided access.
23049         (vect_get_strided_load_store_ops): New function.
23050         (vect_get_data_ptr_increment): Likewise.
23051         (vectorizable_load): Handle strided gather loads.  Always pass
23052         a step to vect_create_data_ref_ptr and bump_vector_ptr.
23054 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23055             Alan Hayward  <alan.hayward@arm.com>
23056             David Sherwood  <david.sherwood@arm.com>
23058         * doc/md.texi (gather_load@var{m}): Document.
23059         (mask_gather_load@var{m}): Likewise.
23060         * genopinit.c (main): Add supports_vec_gather_load and
23061         supports_vec_gather_load_cached to target_optabs.
23062         * optabs-tree.c (init_tree_optimization_optabs): Use
23063         ggc_cleared_alloc to allocate target_optabs.
23064         * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
23065         * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
23066         functions.
23067         * internal-fn.h (internal_load_fn_p): Declare.
23068         (internal_gather_scatter_fn_p): Likewise.
23069         (internal_fn_mask_index): Likewise.
23070         (internal_gather_scatter_fn_supported_p): Likewise.
23071         * internal-fn.c (gather_load_direct): New macro.
23072         (expand_gather_load_optab_fn): New function.
23073         (direct_gather_load_optab_supported_p): New macro.
23074         (direct_internal_fn_optab): New function.
23075         (internal_load_fn_p): Likewise.
23076         (internal_gather_scatter_fn_p): Likewise.
23077         (internal_fn_mask_index): Likewise.
23078         (internal_gather_scatter_fn_supported_p): Likewise.
23079         * optabs-query.c (supports_at_least_one_mode_p): New function.
23080         (supports_vec_gather_load_p): Likewise.
23081         * optabs-query.h (supports_vec_gather_load_p): Declare.
23082         * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
23083         and memory_type field.
23084         (NUM_PATTERNS): Bump to 15.
23085         * tree-vect-data-refs.c: Include internal-fn.h.
23086         (vect_gather_scatter_fn_p): New function.
23087         (vect_describe_gather_scatter_call): Likewise.
23088         (vect_check_gather_scatter): Try using internal functions for
23089         gather loads.  Recognize existing calls to a gather load function.
23090         (vect_analyze_data_refs): Consider using gather loads if
23091         supports_vec_gather_load_p.
23092         * tree-vect-patterns.c (vect_get_load_store_mask): New function.
23093         (vect_get_gather_scatter_offset_type): Likewise.
23094         (vect_convert_mask_for_vectype): Likewise.
23095         (vect_add_conversion_to_patterm): Likewise.
23096         (vect_try_gather_scatter_pattern): Likewise.
23097         (vect_recog_gather_scatter_pattern): New pattern recognizer.
23098         (vect_vect_recog_func_ptrs): Add it.
23099         * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
23100         internal_fn_mask_index and internal_gather_scatter_fn_p.
23101         (check_load_store_masking): Take the gather_scatter_info as an
23102         argument and handle gather loads.
23103         (vect_get_gather_scatter_ops): New function.
23104         (vectorizable_call): Check internal_load_fn_p.
23105         (vectorizable_load): Likewise.  Handle gather load internal
23106         functions.
23107         (vectorizable_store): Update call to check_load_store_masking.
23108         * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
23109         * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
23110         * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
23111         (aarch64_gather_scale_operand_d): New predicates.
23112         * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
23113         (mask_gather_load<mode>): New insns.
23115 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23116             Alan Hayward  <alan.hayward@arm.com>
23117             David Sherwood  <david.sherwood@arm.com>
23119         * optabs.def (fold_left_plus_optab): New optab.
23120         * doc/md.texi (fold_left_plus_@var{m}): Document.
23121         * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
23122         * internal-fn.c (fold_left_direct): Define.
23123         (expand_fold_left_optab_fn): Likewise.
23124         (direct_fold_left_optab_supported_p): Likewise.
23125         * fold-const-call.c (fold_const_fold_left): New function.
23126         (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
23127         * tree-parloops.c (valid_reduction_p): New function.
23128         (gather_scalar_reductions): Use it.
23129         * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
23130         (vect_finish_replace_stmt): Declare.
23131         * tree-vect-loop.c (fold_left_reduction_fn): New function.
23132         (needs_fold_left_reduction_p): New function, split out from...
23133         (vect_is_simple_reduction): ...here.  Accept reductions that
23134         forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
23135         (vect_force_simple_reduction): Also store the reduction type in
23136         the assignment's STMT_VINFO_REDUC_TYPE.
23137         (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
23138         (merge_with_identity): New function.
23139         (vect_expand_fold_left): Likewise.
23140         (vectorize_fold_left_reduction): Likewise.
23141         (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION.  Leave the
23142         scalar phi in place for it.  Check for target support and reject
23143         cases that would reassociate the operation.  Defer the transform
23144         phase to vectorize_fold_left_reduction.
23145         * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
23146         * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
23147         (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
23149 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23151         * tree-if-conv.c (predicate_mem_writes): Remove redundant
23152         call to ifc_temp_var.
23154 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23155             Alan Hayward  <alan.hayward@arm.com>
23156             David Sherwood  <david.sherwood@arm.com>
23158         * target.def (legitimize_address_displacement): Take the original
23159         offset as a poly_int.
23160         * targhooks.h (default_legitimize_address_displacement): Update
23161         accordingly.
23162         * targhooks.c (default_legitimize_address_displacement): Likewise.
23163         * doc/tm.texi: Regenerate.
23164         * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
23165         as an argument, moving assert of ad->disp == ad->disp_term to...
23166         (process_address_1): ...here.  Update calls to base_plus_disp_to_reg.
23167         Try calling targetm.legitimize_address_displacement before expanding
23168         the address rather than afterwards, and adjust for the new interface.
23169         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
23170         Match the new hook interface.  Handle SVE addresses.
23171         * config/sh/sh.c (sh_legitimize_address_displacement): Make the
23172         new hook interface.
23174 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23176         * Makefile.in (OBJS): Add early-remat.o.
23177         * target.def (select_early_remat_modes): New hook.
23178         * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
23179         * doc/tm.texi: Regenerate.
23180         * targhooks.h (default_select_early_remat_modes): Declare.
23181         * targhooks.c (default_select_early_remat_modes): New function.
23182         * timevar.def (TV_EARLY_REMAT): New timevar.
23183         * passes.def (pass_early_remat): New pass.
23184         * tree-pass.h (make_pass_early_remat): Declare.
23185         * early-remat.c: New file.
23186         * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
23187         function.
23188         (TARGET_SELECT_EARLY_REMAT_MODES): Define.
23190 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23191             Alan Hayward  <alan.hayward@arm.com>
23192             David Sherwood  <david.sherwood@arm.com>
23194         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
23195         vfm1 with a bound_epilog parameter.
23196         (vect_do_peeling): Update calls accordingly, and move the prologue
23197         call earlier in the function.  Treat the base bound_epilog as 0 for
23198         fully-masked loops and retain vf - 1 for other loops.  Add 1 to
23199         this base when peeling for gaps.
23200         * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
23201         with fully-masked loops.
23202         (vect_estimate_min_profitable_iters): Handle the single peeled
23203         iteration in that case.
23205 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23206             Alan Hayward  <alan.hayward@arm.com>
23207             David Sherwood  <david.sherwood@arm.com>
23209         * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
23210         single-element interleaving even if the size is not a power of 2.
23211         * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
23212         accesses for single-element interleaving if the group size is
23213         not a power of 2.
23215 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23216             Alan Hayward  <alan.hayward@arm.com>
23217             David Sherwood  <david.sherwood@arm.com>
23219         * doc/md.texi (fold_extract_last_@var{m}): Document.
23220         * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
23221         * optabs.def (fold_extract_last_optab): New optab.
23222         * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
23223         * internal-fn.c (fold_extract_direct): New macro.
23224         (expand_fold_extract_optab_fn): Likewise.
23225         (direct_fold_extract_optab_supported_p): Likewise.
23226         * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
23227         * tree-vect-loop.c (vect_model_reduction_cost): Handle
23228         EXTRACT_LAST_REDUCTION.
23229         (get_initial_def_for_reduction): Do not create an initial vector
23230         for EXTRACT_LAST_REDUCTION reductions.
23231         (vectorizable_reduction): Leave the scalar phi in place for
23232         EXTRACT_LAST_REDUCTIONs.  Try using EXTRACT_LAST_REDUCTION
23233         ahead of INTEGER_INDUC_COND_REDUCTION.  Do not check for an
23234         epilogue code for EXTRACT_LAST_REDUCTION and defer the
23235         transform phase to vectorizable_condition.
23236         * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
23237         split out from...
23238         (vect_finish_stmt_generation): ...here.
23239         (vect_finish_replace_stmt): New function.
23240         (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
23241         * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
23242         pattern.
23243         * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
23245 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23246             Alan Hayward  <alan.hayward@arm.com>
23247             David Sherwood  <david.sherwood@arm.com>
23249         * doc/md.texi (extract_last_@var{m}): Document.
23250         * optabs.def (extract_last_optab): New optab.
23251         * internal-fn.def (EXTRACT_LAST): New internal function.
23252         * internal-fn.c (cond_unary_direct): New macro.
23253         (expand_cond_unary_optab_fn): Likewise.
23254         (direct_cond_unary_optab_supported_p): Likewise.
23255         * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
23256         loops using EXTRACT_LAST.
23257         * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
23258         (extract_last_<mode>): ...this optab.
23259         (vec_extract<mode><Vel>): Update accordingly.
23261 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23262             Alan Hayward  <alan.hayward@arm.com>
23263             David Sherwood  <david.sherwood@arm.com>
23265         * target.def (empty_mask_is_expensive): New hook.
23266         * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
23267         * doc/tm.texi: Regenerate.
23268         * targhooks.h (default_empty_mask_is_expensive): Declare.
23269         * targhooks.c (default_empty_mask_is_expensive): New function.
23270         * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
23271         if the target says that empty masks are expensive.
23272         * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
23273         New function.
23274         (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
23276 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23277             Alan Hayward  <alan.hayward@arm.com>
23278             David Sherwood  <david.sherwood@arm.com>
23280         * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
23281         (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
23282         (vect_use_loop_mask_for_alignment_p): New function.
23283         (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
23284         * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
23285         niters_skip argument.  Make sure that the first niters_skip elements
23286         of the first iteration are inactive.
23287         (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
23288         Update call to vect_set_loop_masks_directly.
23289         (get_misalign_in_elems): New function, split out from...
23290         (vect_gen_prolog_loop_niters): ...here.
23291         (vect_update_init_of_dr): Take a code argument that specifies whether
23292         the adjustment should be added or subtracted.
23293         (vect_update_init_of_drs): Likewise.
23294         (vect_prepare_for_masked_peels): New function.
23295         (vect_do_peeling): Skip prologue peeling if we're using a mask
23296         instead.  Update call to vect_update_inits_of_drs.
23297         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
23298         mask_skip_niters.
23299         (vect_analyze_loop_2): Allow fully-masked loops with peeling for
23300         alignment.  Do not include the number of peeled iterations in
23301         the minimum threshold in that case.
23302         (vectorizable_induction): Adjust the start value down by
23303         LOOP_VINFO_MASK_SKIP_NITERS iterations.
23304         (vect_transform_loop): Call vect_prepare_for_masked_peels.
23305         Take the number of skipped iterations into account when calculating
23306         the loop bounds.
23307         * tree-vect-stmts.c (vect_gen_while_not): New function.
23309 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23310             Alan Hayward  <alan.hayward@arm.com>
23311             David Sherwood  <david.sherwood@arm.com>
23313         * doc/sourcebuild.texi (vect_fully_masked): Document.
23314         * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
23315         default value to 0.
23316         * tree-vect-loop.c (vect_analyze_loop_costing): New function,
23317         split out from...
23318         (vect_analyze_loop_2): ...here. Don't check the vectorization
23319         factor against the number of loop iterations if the loop is
23320         fully-masked.
23322 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23323             Alan Hayward  <alan.hayward@arm.com>
23324             David Sherwood  <david.sherwood@arm.com>
23326         * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
23327         (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
23328         (dump_groups): Update accordingly.
23329         (iv_use::mem_type): New member variable.
23330         (address_p): New function.
23331         (record_use): Add a mem_type argument and initialize the new
23332         mem_type field.
23333         (record_group_use): Add a mem_type argument.  Use address_p.
23334         Remove obsolete null checks of base_object.  Update call to record_use.
23335         (find_interesting_uses_op): Update call to record_group_use.
23336         (find_interesting_uses_cond): Likewise.
23337         (find_interesting_uses_address): Likewise.
23338         (get_mem_type_for_internal_fn): New function.
23339         (find_address_like_use): Likewise.
23340         (find_interesting_uses_stmt): Try find_address_like_use before
23341         calling find_interesting_uses_op.
23342         (addr_offset_valid_p): Use the iv mem_type field as the type
23343         of the addressed memory.
23344         (add_autoinc_candidates): Likewise.
23345         (get_address_cost): Likewise.
23346         (split_small_address_groups_p): Use address_p.
23347         (split_address_groups): Likewise.
23348         (add_iv_candidate_for_use): Likewise.
23349         (autoinc_possible_for_pair): Likewise.
23350         (rewrite_groups): Likewise.
23351         (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
23352         (determine_group_iv_cost): Update after split of USE_ADDRESS.
23353         (get_alias_ptr_type_for_ptr_address): New function.
23354         (rewrite_use_address): Rewrite address uses in calls that were
23355         identified by find_address_like_use.
23357 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23358             Alan Hayward  <alan.hayward@arm.com>
23359             David Sherwood  <david.sherwood@arm.com>
23361         * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
23362         TARGET_MEM_REFs.
23363         * gimple-expr.h (is_gimple_addressable: Likewise.
23364         * gimple-expr.c (is_gimple_address): Likewise.
23365         * internal-fn.c (expand_call_mem_ref): New function.
23366         (expand_mask_load_optab_fn): Use it.
23367         (expand_mask_store_optab_fn): Likewise.
23369 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23370             Alan Hayward  <alan.hayward@arm.com>
23371             David Sherwood  <david.sherwood@arm.com>
23373         * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
23374         (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
23375         (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
23376         (cond_umax@var{mode}): Document.
23377         * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
23378         (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
23379         (cond_umin_optab, cond_umax_optab): New optabs.
23380         * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
23381         (COND_IOR, COND_XOR): New internal functions.
23382         * internal-fn.h (get_conditional_internal_fn): Declare.
23383         * internal-fn.c (cond_binary_direct): New macro.
23384         (expand_cond_binary_optab_fn): Likewise.
23385         (direct_cond_binary_optab_supported_p): Likewise.
23386         (get_conditional_internal_fn): New function.
23387         * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
23388         Cope with reduction statements that are vectorized as calls rather
23389         than assignments.
23390         * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
23391         * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
23392         (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
23393         (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
23394         (UNSPEC_COND_EOR): New unspecs.
23395         (optab): Add mappings for them.
23396         (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
23397         (sve_int_op, sve_fp_op): New int attributes.
23399 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23400             Alan Hayward  <alan.hayward@arm.com>
23401             David Sherwood  <david.sherwood@arm.com>
23403         * optabs.def (while_ult_optab): New optab.
23404         * doc/md.texi (while_ult@var{m}@var{n}): Document.
23405         * internal-fn.def (WHILE_ULT): New internal function.
23406         * internal-fn.h (direct_internal_fn_supported_p): New override
23407         that takes two types as argument.
23408         * internal-fn.c (while_direct): New macro.
23409         (expand_while_optab_fn): New function.
23410         (convert_optab_supported_p): Likewise.
23411         (direct_while_optab_supported_p): New macro.
23412         * wide-int.h (wi::udiv_ceil): New function.
23413         * tree-vectorizer.h (rgroup_masks): New structure.
23414         (vec_loop_masks): New typedef.
23415         (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
23416         and fully_masked_p.
23417         (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
23418         (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
23419         (vect_max_vf): New function.
23420         (slpeel_make_loop_iterate_ntimes): Delete.
23421         (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
23422         (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
23423         (vect_record_loop_mask, vect_get_loop_mask): Likewise.
23424         * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
23425         internal-fn.h, stor-layout.h and optabs-query.h.
23426         (vect_set_loop_mask): New function.
23427         (add_preheader_seq): Likewise.
23428         (add_header_seq): Likewise.
23429         (interleave_supported_p): Likewise.
23430         (vect_maybe_permute_loop_masks): Likewise.
23431         (vect_set_loop_masks_directly): Likewise.
23432         (vect_set_loop_condition_masked): Likewise.
23433         (vect_set_loop_condition_unmasked): New function, split out from
23434         slpeel_make_loop_iterate_ntimes.
23435         (slpeel_make_loop_iterate_ntimes): Rename to..
23436         (vect_set_loop_condition): ...this.  Use vect_set_loop_condition_masked
23437         for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
23438         (vect_do_peeling): Update call accordingly.
23439         (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
23440         loops.
23441         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
23442         mask_compare_type, can_fully_mask_p and fully_masked_p.
23443         (release_vec_loop_masks): New function.
23444         (_loop_vec_info): Use it to free the loop masks.
23445         (can_produce_all_loop_masks_p): New function.
23446         (vect_get_max_nscalars_per_iter): Likewise.
23447         (vect_verify_full_masking): Likewise.
23448         (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
23449         retries, and free the mask rgroups before retrying.  Check loop-wide
23450         reasons for disallowing fully-masked loops.  Make the final decision
23451         about whether use a fully-masked loop or not.
23452         (vect_estimate_min_profitable_iters): Do not assume that peeling
23453         for the number of iterations will be needed for fully-masked loops.
23454         (vectorizable_reduction): Disable fully-masked loops.
23455         (vectorizable_live_operation): Likewise.
23456         (vect_halve_mask_nunits): New function.
23457         (vect_double_mask_nunits): Likewise.
23458         (vect_record_loop_mask): Likewise.
23459         (vect_get_loop_mask): Likewise.
23460         (vect_transform_loop): Handle the case in which the final loop
23461         iteration might handle a partial vector.  Call vect_set_loop_condition
23462         instead of slpeel_make_loop_iterate_ntimes.
23463         * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
23464         (check_load_store_masking): New function.
23465         (prepare_load_store_mask): Likewise.
23466         (vectorizable_store): Handle fully-masked loops.
23467         (vectorizable_load): Likewise.
23468         (supportable_widening_operation): Use vect_halve_mask_nunits for
23469         booleans.
23470         (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
23471         (vect_gen_while): New function.
23472         * config/aarch64/aarch64.md (umax<mode>3): New expander.
23473         (aarch64_uqdec<mode>): New insn.
23475 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23476             Alan Hayward  <alan.hayward@arm.com>
23477             David Sherwood  <david.sherwood@arm.com>
23479         * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
23480         (reduc_xor_scal_optab): New optabs.
23481         * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
23482         (reduc_xor_scal_@var{m}): Document.
23483         * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
23484         * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
23485         internal functions.
23486         * fold-const-call.c (fold_const_call): Handle them.
23487         * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
23488         internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
23489         * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
23490         (*reduc_<bit_reduc>_scal_<mode>): New patterns.
23491         * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
23492         (UNSPEC_XORV): New unspecs.
23493         (optab): Add entries for them.
23494         (BITWISEV): New int iterator.
23495         (bit_reduc_op): New int attributes.
23497 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23498             Alan Hayward  <alan.hayward@arm.com>
23499             David Sherwood  <david.sherwood@arm.com>
23501         * doc/md.texi (vec_shl_insert_@var{m}): New optab.
23502         * internal-fn.def (VEC_SHL_INSERT): New internal function.
23503         * optabs.def (vec_shl_insert_optab): New optab.
23504         * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
23505         (duplicate_and_interleave): Likewise.
23506         * tree-vect-loop.c: Include internal-fn.h.
23507         (neutral_op_for_slp_reduction): New function, split out from
23508         get_initial_defs_for_reduction.
23509         (get_initial_def_for_reduction): Handle option 2 for variable-length
23510         vectors by loading the neutral value into a vector and then shifting
23511         the initial value into element 0.
23512         (get_initial_defs_for_reduction): Replace the code argument with
23513         the neutral value calculated by neutral_op_for_slp_reduction.
23514         Use gimple_build_vector for constant-length vectors.
23515         Use IFN_VEC_SHL_INSERT for variable-length vectors if all
23516         but the first group_size elements have a neutral value.
23517         Use duplicate_and_interleave otherwise.
23518         (vect_create_epilog_for_reduction): Take a neutral_op parameter.
23519         Update call to get_initial_defs_for_reduction.  Handle SLP
23520         reductions for variable-length vectors by creating one vector
23521         result for each scalar result, with the elements associated
23522         with other scalar results stubbed out with the neutral value.
23523         (vectorizable_reduction): Call neutral_op_for_slp_reduction.
23524         Require IFN_VEC_SHL_INSERT for double reductions on
23525         variable-length vectors, or SLP reductions that have
23526         a neutral value.  Require can_duplicate_and_interleave_p
23527         support for variable-length unchained SLP reductions if there
23528         is no neutral value, such as for MIN/MAX reductions.  Also require
23529         the number of vector elements to be a multiple of the number of
23530         SLP statements when doing variable-length unchained SLP reductions.
23531         Update call to vect_create_epilog_for_reduction.
23532         * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
23533         and remove initial values.
23534         (duplicate_and_interleave): Make public.
23535         * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
23536         * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
23538 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23539             Alan Hayward  <alan.hayward@arm.com>
23540             David Sherwood  <david.sherwood@arm.com>
23542         * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
23543         (can_duplicate_and_interleave_p): New function.
23544         (vect_get_and_check_slp_defs): Take the vector of statements
23545         rather than just the current one.  Remove excess parentheses.
23546         Restriction rejectinon of vect_constant_def and vect_external_def
23547         for variable-length vectors to boolean types, or types for which
23548         can_duplicate_and_interleave_p is false.
23549         (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
23550         (duplicate_and_interleave): New function.
23551         (vect_get_constant_vectors): Use gimple_build_vector for
23552         constant-length vectors and suitable variable-length constant
23553         vectors.  Use duplicate_and_interleave for other variable-length
23554         vectors.  Don't defer the update when inserting new statements.
23556 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23557             Alan Hayward  <alan.hayward@arm.com>
23558             David Sherwood  <david.sherwood@arm.com>
23560         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
23561         min_profitable_iters doesn't go negative.
23563 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23564             Alan Hayward  <alan.hayward@arm.com>
23565             David Sherwood  <david.sherwood@arm.com>
23567         * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
23568         (vec_mask_store_lanes@var{m}@var{n}): Likewise.
23569         * optabs.def (vec_mask_load_lanes_optab): New optab.
23570         (vec_mask_store_lanes_optab): Likewise.
23571         * internal-fn.def (MASK_LOAD_LANES): New internal function.
23572         (MASK_STORE_LANES): Likewise.
23573         * internal-fn.c (mask_load_lanes_direct): New macro.
23574         (mask_store_lanes_direct): Likewise.
23575         (expand_mask_load_optab_fn): Handle masked operations.
23576         (expand_mask_load_lanes_optab_fn): New macro.
23577         (expand_mask_store_optab_fn): Handle masked operations.
23578         (expand_mask_store_lanes_optab_fn): New macro.
23579         (direct_mask_load_lanes_optab_supported_p): Likewise.
23580         (direct_mask_store_lanes_optab_supported_p): Likewise.
23581         * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
23582         parameter.
23583         (vect_load_lanes_supported): Likewise.
23584         * tree-vect-data-refs.c (strip_conversion): New function.
23585         (can_group_stmts_p): Likewise.
23586         (vect_analyze_data_ref_accesses): Use it instead of checking
23587         for a pair of assignments.
23588         (vect_store_lanes_supported): Take a masked_p parameter.
23589         (vect_load_lanes_supported): Likewise.
23590         * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
23591         vect_store_lanes_supported and vect_load_lanes_supported.
23592         * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
23593         * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
23594         parameter.  Don't allow gaps for masked accesses.
23595         Use vect_get_store_rhs.  Update calls to vect_store_lanes_supported
23596         and vect_load_lanes_supported.
23597         (get_load_store_type): Take a masked_p parameter and update
23598         call to get_group_load_store_type.
23599         (vectorizable_store): Update call to get_load_store_type.
23600         Handle IFN_MASK_STORE_LANES.
23601         (vectorizable_load): Update call to get_load_store_type.
23602         Handle IFN_MASK_LOAD_LANES.
23604 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23605             Alan Hayward  <alan.hayward@arm.com>
23606             David Sherwood  <david.sherwood@arm.com>
23608         * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
23609         modes for SVE.
23610         * config/aarch64/aarch64-protos.h
23611         (aarch64_sve_struct_memory_operand_p): Declare.
23612         * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
23613         (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
23614         (VPRED, vpred): Handle SVE structure modes.
23615         * config/aarch64/constraints.md (Utx): New constraint.
23616         * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
23617         (aarch64_sve_struct_nonimmediate_operand): New predicates.
23618         * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
23619         * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
23620         (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
23621         structure modes.  Split into pieces after RA.
23622         (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
23623         (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
23624         New patterns.
23625         * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
23626         SVE structure modes.
23627         (aarch64_classify_address): Likewise.
23628         (sizetochar): Move earlier in file.
23629         (aarch64_print_operand): Handle SVE register lists.
23630         (aarch64_array_mode): New function.
23631         (aarch64_sve_struct_memory_operand_p): Likewise.
23632         (TARGET_ARRAY_MODE): Redefine.
23634 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23635             Alan Hayward  <alan.hayward@arm.com>
23636             David Sherwood  <david.sherwood@arm.com>
23638         * target.def (array_mode): New target hook.
23639         * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
23640         * doc/tm.texi: Regenerate.
23641         * hooks.h (hook_optmode_mode_uhwi_none): Declare.
23642         * hooks.c (hook_optmode_mode_uhwi_none): New function.
23643         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
23644         targetm.array_mode.
23645         * stor-layout.c (mode_for_array): Likewise.  Support polynomial
23646         type sizes.
23648 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23649             Alan Hayward  <alan.hayward@arm.com>
23650             David Sherwood  <david.sherwood@arm.com>
23652         * fold-const.c (fold_binary_loc): Check the argument types
23653         rather than the result type when testing for a vector operation.
23655 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23657         * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
23658         * doc/tm.texi: Regenerate.
23660 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23661             Alan Hayward  <alan.hayward@arm.com>
23662             David Sherwood  <david.sherwood@arm.com>
23664         * doc/invoke.texi (-msve-vector-bits=): Document new option.
23665         (sve): Document new AArch64 extension.
23666         * doc/md.texi (w): Extend the description of the AArch64
23667         constraint to include SVE vectors.
23668         (Upl, Upa): Document new AArch64 predicate constraints.
23669         * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
23670         enum.
23671         * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
23672         (msve-vector-bits=): New option.
23673         * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
23674         SVE when these are disabled.
23675         (sve): New extension.
23676         * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
23677         modes.  Adjust their number of units based on aarch64_sve_vg.
23678         (MAX_BITSIZE_MODE_ANY_MODE): Define.
23679         * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
23680         aarch64_addr_query_type.
23681         (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
23682         (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
23683         (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
23684         (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
23685         (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
23686         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
23687         (aarch64_simd_imm_zero_p): Delete.
23688         (aarch64_check_zero_based_sve_index_immediate): Declare.
23689         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
23690         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
23691         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
23692         (aarch64_sve_float_mul_immediate_p): Likewise.
23693         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
23694         rather than an rtx.
23695         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
23696         (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
23697         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
23698         (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
23699         (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
23700         (aarch64_regmode_natural_size): Likewise.
23701         * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
23702         (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
23703         left one place.
23704         (AARCH64_ISA_SVE, TARGET_SVE): New macros.
23705         (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
23706         for VG and the SVE predicate registers.
23707         (V_ALIASES): Add a "z"-prefixed alias.
23708         (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
23709         (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
23710         (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
23711         (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
23712         (REG_CLASS_NAMES): Add entries for them.
23713         (REG_CLASS_CONTENTS): Likewise.  Update ALL_REGS to include VG
23714         and the predicate registers.
23715         (aarch64_sve_vg): Declare.
23716         (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
23717         (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
23718         (REGMODE_NATURAL_SIZE): Define.
23719         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
23720         SVE macros.
23721         * config/aarch64/aarch64.c: Include cfgrtl.h.
23722         (simd_immediate_info): Add a constructor for series vectors,
23723         and an associated step field.
23724         (aarch64_sve_vg): New variable.
23725         (aarch64_dbx_register_number): Handle VG and the predicate registers.
23726         (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
23727         (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
23728         (VEC_ANY_DATA, VEC_STRUCT): New constants.
23729         (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
23730         (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
23731         (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
23732         (aarch64_get_mask_mode): New functions.
23733         (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
23734         and FP_LO_REGS.  Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
23735         (aarch64_hard_regno_mode_ok): Handle VG.  Also handle the SVE
23736         predicate modes and predicate registers.  Explicitly restrict
23737         GPRs to modes of 16 bytes or smaller.  Only allow FP registers
23738         to store a vector mode if it is recognized by
23739         aarch64_classify_vector_mode.
23740         (aarch64_regmode_natural_size): New function.
23741         (aarch64_hard_regno_caller_save_mode): Return the original mode
23742         for predicates.
23743         (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
23744         (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
23745         (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
23746         (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
23747         functions.
23748         (aarch64_add_offset): Add a temp2 parameter.  Assert that temp1
23749         does not overlap dest if the function is frame-related.  Handle
23750         SVE constants.
23751         (aarch64_split_add_offset): New function.
23752         (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
23753         them aarch64_add_offset.
23754         (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
23755         and update call to aarch64_sub_sp.
23756         (aarch64_add_cfa_expression): New function.
23757         (aarch64_expand_prologue): Pass extra temporary registers to the
23758         functions above.  Handle the case in which we need to emit new
23759         DW_CFA_expressions for registers that were originally saved
23760         relative to the stack pointer, but now have to be expressed
23761         relative to the frame pointer.
23762         (aarch64_output_mi_thunk): Pass extra temporary registers to the
23763         functions above.
23764         (aarch64_expand_epilogue): Likewise.  Prevent inheritance of
23765         IP0 and IP1 values for SVE frames.
23766         (aarch64_expand_vec_series): New function.
23767         (aarch64_expand_sve_widened_duplicate): Likewise.
23768         (aarch64_expand_sve_const_vector): Likewise.
23769         (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
23770         Handle SVE constants.  Use emit_move_insn to move a force_const_mem
23771         into the register, rather than emitting a SET directly.
23772         (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
23773         (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
23774         (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
23775         (offset_9bit_signed_scaled_p): New functions.
23776         (aarch64_replicate_bitmask_imm): New function.
23777         (aarch64_bitmask_imm): Use it.
23778         (aarch64_cannot_force_const_mem): Reject expressions involving
23779         a CONST_POLY_INT.  Update call to aarch64_classify_symbol.
23780         (aarch64_classify_index): Handle SVE indices, by requiring
23781         a plain register index with a scale that matches the element size.
23782         (aarch64_classify_address): Handle SVE addresses.  Assert that
23783         the mode of the address is VOIDmode or an integer mode.
23784         Update call to aarch64_classify_symbol.
23785         (aarch64_classify_symbolic_expression): Update call to
23786         aarch64_classify_symbol.
23787         (aarch64_const_vec_all_in_range_p): New function.
23788         (aarch64_print_vector_float_operand): Likewise.
23789         (aarch64_print_operand): Handle 'N' and 'C'.  Use "zN" rather than
23790         "vN" for FP registers with SVE modes.  Handle (const ...) vectors
23791         and the FP immediates 1.0 and 0.5.
23792         (aarch64_print_address_internal): Handle SVE addresses.
23793         (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
23794         (aarch64_regno_regclass): Handle predicate registers.
23795         (aarch64_secondary_reload): Handle big-endian reloads of SVE
23796         data modes.
23797         (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
23798         (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
23799         (aarch64_convert_sve_vector_bits): New function.
23800         (aarch64_override_options): Use it to handle -msve-vector-bits=.
23801         (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
23802         rather than an rtx.
23803         (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
23804         Handle SVE vector and predicate modes.  Accept VL-based constants
23805         that need only one temporary register, and VL offsets that require
23806         no temporary registers.
23807         (aarch64_conditional_register_usage): Mark the predicate registers
23808         as fixed if SVE isn't available.
23809         (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
23810         Return true for SVE vector and predicate modes.
23811         (aarch64_simd_container_mode): Take the number of bits as a poly_int64
23812         rather than an unsigned int.  Handle SVE modes.
23813         (aarch64_preferred_simd_mode): Update call accordingly.  Handle
23814         SVE modes.
23815         (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
23816         if SVE is enabled.
23817         (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
23818         (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
23819         (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
23820         (aarch64_sve_float_mul_immediate_p): New functions.
23821         (aarch64_sve_valid_immediate): New function.
23822         (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
23823         Explicitly reject structure modes.  Check for INDEX constants.
23824         Handle PTRUE and PFALSE constants.
23825         (aarch64_check_zero_based_sve_index_immediate): New function.
23826         (aarch64_simd_imm_zero_p): Delete.
23827         (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
23828         vector modes.  Accept constants in the range of CNT[BHWD].
23829         (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
23830         ask for an Advanced SIMD mode.
23831         (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
23832         (aarch64_simd_vector_alignment): Handle SVE predicates.
23833         (aarch64_vectorize_preferred_vector_alignment): New function.
23834         (aarch64_simd_vector_alignment_reachable): Use it instead of
23835         the vector size.
23836         (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
23837         (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
23838         functions.
23839         (MAX_VECT_LEN): Delete.
23840         (expand_vec_perm_d): Add a vec_flags field.
23841         (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
23842         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
23843         (aarch64_evpc_ext): Don't apply a big-endian lane correction
23844         for SVE modes.
23845         (aarch64_evpc_rev): Rename to...
23846         (aarch64_evpc_rev_local): ...this.  Use a predicated operation for SVE.
23847         (aarch64_evpc_rev_global): New function.
23848         (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
23849         (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
23850         MAX_VECT_LEN.
23851         (aarch64_evpc_sve_tbl): New function.
23852         (aarch64_expand_vec_perm_const_1): Update after rename of
23853         aarch64_evpc_rev.  Handle SVE permutes too, trying
23854         aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
23855         than aarch64_evpc_tbl.
23856         (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
23857         (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
23858         (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
23859         (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
23860         (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
23861         (aarch64_expand_sve_vcond): New functions.
23862         (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
23863         of aarch64_vector_mode_p.
23864         (aarch64_dwarf_poly_indeterminate_value): New function.
23865         (aarch64_compute_pressure_classes): Likewise.
23866         (aarch64_can_change_mode_class): Likewise.
23867         (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
23868         (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
23869         (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
23870         (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
23871         (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
23872         (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
23873         * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
23874         (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
23875         constraints.
23876         (Dn, Dl, Dr): Accept const as well as const_vector.
23877         (Dz): Likewise.  Compare against CONST0_RTX.
23878         * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
23879         of "vector" where appropriate.
23880         (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
23881         (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
23882         (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
23883         (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
23884         (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
23885         (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
23886         (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
23887         (v_int_equiv): Extend to SVE modes.
23888         (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
23889         mode attributes.
23890         (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
23891         (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
23892         (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
23893         (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
23894         (SVE_COND_FP_CMP): New int iterators.
23895         (perm_hilo): Handle the new unpack unspecs.
23896         (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
23897         attributes.
23898         * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
23899         (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
23900         (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
23901         (aarch64_equality_operator, aarch64_constant_vector_operand)
23902         (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
23903         (aarch64_sve_nonimmediate_operand): Likewise.
23904         (aarch64_sve_general_operand): Likewise.
23905         (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
23906         (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
23907         (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
23908         (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
23909         (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
23910         (aarch64_sve_float_arith_immediate): Likewise.
23911         (aarch64_sve_float_arith_with_sub_immediate): Likewise.
23912         (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
23913         (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
23914         (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
23915         (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
23916         (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
23917         (aarch64_sve_float_arith_operand): Likewise.
23918         (aarch64_sve_float_arith_with_sub_operand): Likewise.
23919         (aarch64_sve_float_mul_operand): Likewise.
23920         (aarch64_sve_vec_perm_operand): Likewise.
23921         (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
23922         (aarch64_mov_operand): Accept const_poly_int and const_vector.
23923         (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
23924         as well as const_vector.
23925         (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
23926         in file.  Use CONST0_RTX and CONSTM1_RTX.
23927         (aarch64_simd_or_scalar_imm_zero): Likewise.  Add match_codes.
23928         (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
23929         Use aarch64_simd_imm_zero.
23930         * config/aarch64/aarch64-sve.md: New file.
23931         * config/aarch64/aarch64.md: Include it.
23932         (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
23933         (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
23934         (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
23935         (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
23936         (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
23937         (sve): New attribute.
23938         (enabled): Disable instructions with the sve attribute unless
23939         TARGET_SVE.
23940         (movqi, movhi): Pass CONST_POLY_INT operaneds through
23941         aarch64_expand_mov_immediate.
23942         (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
23943         CNT[BHSD] immediates.
23944         (movti): Split CONST_POLY_INT moves into two halves.
23945         (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
23946         Split additions that need a temporary here if the destination
23947         is the stack pointer.
23948         (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
23949         (*add<mode>3_poly_1): New instruction.
23950         (set_clobber_cc): New expander.
23952 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23954         * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
23955         parameter and use it instead of GET_MODE_SIZE (innermode).  Use
23956         inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
23957         Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
23958         GET_MODE_NUNITS (innermode).  Also add a first_elem parameter.
23959         Change innermode from fixed_mode_size to machine_mode.
23960         (simplify_subreg): Update call accordingly.  Handle a constant-sized
23961         subreg of a variable-length CONST_VECTOR.
23963 2018-01-13  Richard Sandiford  <richard.sandiford@linaro.org>
23964             Alan Hayward  <alan.hayward@arm.com>
23965             David Sherwood  <david.sherwood@arm.com>
23967         * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
23968         (add_offset_to_base): New function, split out from...
23969         (create_mem_ref): ...here.  When handling a scale other than 1,
23970         check first whether the address is valid without the offset.
23971         Add it into the base if so, leaving the index and scale as-is.
23973 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
23975         PR c++/83778
23976         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
23977         fold_for_warn before checking if arg2 is INTEGER_CST.
23979 2018-01-12  Segher Boessenkool  <segher@kernel.crashing.org>
23981         * config/rs6000/predicates.md (load_multiple_operation): Delete.
23982         (store_multiple_operation): Delete.
23983         * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
23984         * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
23985         * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
23986         guarded by TARGET_STRING.
23987         (rs6000_output_load_multiple): Delete.
23988         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
23989         OPTION_MASK_STRING / TARGET_STRING handling.
23990         (print_operand) <'N', 'O'>: Add comment that these are unused now.
23991         (const rs6000_opt_masks) <"string">: Change mask to 0.
23992         * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
23993         (MASK_STRING): Delete.
23994         * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
23995         parts.  Simplify.
23996         (load_multiple): Delete.
23997         (*ldmsi8): Delete.
23998         (*ldmsi7): Delete.
23999         (*ldmsi6): Delete.
24000         (*ldmsi5): Delete.
24001         (*ldmsi4): Delete.
24002         (*ldmsi3): Delete.
24003         (store_multiple): Delete.
24004         (*stmsi8): Delete.
24005         (*stmsi7): Delete.
24006         (*stmsi6): Delete.
24007         (*stmsi5): Delete.
24008         (*stmsi4): Delete.
24009         (*stmsi3): Delete.
24010         (movmemsi_8reg): Delete.
24011         (corresponding unnamed define_insn): Delete.
24012         (movmemsi_6reg): Delete.
24013         (corresponding unnamed define_insn): Delete.
24014         (movmemsi_4reg): Delete.
24015         (corresponding unnamed define_insn): Delete.
24016         (movmemsi_2reg): Delete.
24017         (corresponding unnamed define_insn): Delete.
24018         (movmemsi_1reg): Delete.
24019         (corresponding unnamed define_insn): Delete.
24020         * config/rs6000/rs6000.opt (mno-string): New.
24021         (mstring): Replace by deprecation warning stub.
24022         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
24024 2018-01-12  Jakub Jelinek  <jakub@redhat.com>
24026         * regrename.c (regrename_do_replace): If replacing the same
24027         reg multiple times, try to reuse last created gen_raw_REG.
24029         PR debug/81155
24030         * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
24031         main to workaround a bug in GDB.
24033 2018-01-12  Tom de Vries  <tom@codesourcery.com>
24035         PR target/83737
24036         * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
24038 2018-01-12  Vladimir Makarov  <vmakarov@redhat.com>
24040         PR rtl-optimization/80481
24041         * ira-color.c (get_cap_member): New function.
24042         (allocnos_conflict_by_live_ranges_p): Use it.
24043         (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
24044         (setup_slot_coalesced_allocno_live_ranges): Ditto.
24046 2018-01-12  Uros Bizjak  <ubizjak@gmail.com>
24048         PR target/83628
24049         * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
24050         (*saddl_se_1): Ditto.
24051         (*ssubsi_1): Ditto.
24052         (*ssubl_se_1): Ditto.
24054 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
24056         * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
24057         rather than wi::to_widest for DR_INITs.
24058         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
24059         wi::to_poly_offset rather than wi::to_offset for DR_INIT.
24060         (vect_analyze_data_ref_accesses): Require both DR_INITs to be
24061         INTEGER_CSTs.
24062         (vect_analyze_group_access_1): Note that here.
24064 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
24066         * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
24067         polynomial type sizes.
24069 2018-01-12  Richard Sandiford  <richard.sandiford@linaro.org>
24071         * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
24072         poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
24073         (gimple_add_tmp_var): Likewise.
24075 2018-01-12  Martin Liska  <mliska@suse.cz>
24077         * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
24078         (gimple_alloc_sizes): Likewise.
24079         (dump_gimple_statistics): Use PRIu64 in printf format.
24080         * gimple.h: Change uint64_t to int.
24082 2018-01-12  Martin Liska  <mliska@suse.cz>
24084         * tree-core.h: Use uint64_t instead of int.
24085         * tree.c (tree_node_counts): Likewise.
24086         (tree_node_sizes): Likewise.
24087         (dump_tree_statistics): Use PRIu64 in printf format.
24089 2018-01-12  Martin Liska  <mliska@suse.cz>
24091         * Makefile.in: As qsort_chk is implemented in vec.c, add
24092         vec.o to linkage of gencfn-macros.
24093         * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
24094         passing the info to record_node_allocation_statistics.
24095         (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
24096         and pass the info.
24097         * ggc-common.c (struct ggc_usage): Add operator== and use
24098         it in operator< and compare function.
24099         * mem-stats.h (struct mem_usage): Likewise.
24100         * vec.c (struct vec_usage): Remove operator< and compare
24101         function. Can be simply inherited.
24103 2018-01-12  Martin Jambor  <mjambor@suse.cz>
24105         PR target/81616
24106         * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
24107         * tree-ssa-math-opts.c: Include domwalk.h.
24108         (convert_mult_to_fma_1): New function.
24109         (fma_transformation_info): New type.
24110         (fma_deferring_state): Likewise.
24111         (cancel_fma_deferring): New function.
24112         (result_of_phi): Likewise.
24113         (last_fma_candidate_feeds_initial_phi): Likewise.
24114         (convert_mult_to_fma): Added deferring logic, split actual
24115         transformation to convert_mult_to_fma_1.
24116         (math_opts_dom_walker): New type.
24117         (math_opts_dom_walker::after_dom_children): New method, body moved
24118         here from pass_optimize_widening_mul::execute, added deferring logic
24119         bits.
24120         (pass_optimize_widening_mul::execute): Moved most of code to
24121         math_opts_dom_walker::after_dom_children.
24122         * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
24123         * config/i386/i386.c (ix86_option_override_internal): Added
24124         maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
24126 2018-01-12  Richard Biener  <rguenther@suse.de>
24128         PR debug/83157
24129         * dwarf2out.c (gen_variable_die): Do not reset old_die for
24130         inline instance vars.
24132 2018-01-12  Oleg Endo  <olegendo@gcc.gnu.org>
24134         PR target/81819
24135         * config/rx/rx.c (rx_is_restricted_memory_address):
24136         Handle SUBREG case.
24138 2018-01-12  Richard Biener  <rguenther@suse.de>
24140         PR tree-optimization/80846
24141         * target.def (split_reduction): New target hook.
24142         * targhooks.c (default_split_reduction): New function.
24143         * targhooks.h (default_split_reduction): Declare.
24144         * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
24145         target requests first reduce vectors by combining low and high
24146         parts.
24147         * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
24148         (get_vectype_for_scalar_type_and_size): Export.
24149         * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
24150         * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
24151         * doc/tm.texi: Regenerate.
24152         * config/i386/i386.c (ix86_split_reduction): Implement
24153         TARGET_VECTORIZE_SPLIT_REDUCTION.
24155 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
24157         PR target/83368
24158         * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
24159         in PIC mode except for TARGET_VXWORKS_RTP.
24160         * config/sparc/sparc.c: Include cfgrtl.h.
24161         (TARGET_INIT_PIC_REG): Define.
24162         (TARGET_USE_PSEUDO_PIC_REG): Likewise.
24163         (sparc_pic_register_p): New predicate.
24164         (sparc_legitimate_address_p): Use it.
24165         (sparc_legitimize_pic_address): Likewise.
24166         (sparc_delegitimize_address): Likewise.
24167         (sparc_mode_dependent_address_p): Likewise.
24168         (gen_load_pcrel_sym): Remove 4th parameter.
24169         (load_got_register): Adjust call to above.  Remove obsolete stuff.
24170         (sparc_expand_prologue): Do not call load_got_register here.
24171         (sparc_flat_expand_prologue): Likewise.
24172         (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
24173         (sparc_use_pseudo_pic_reg): New function.
24174         (sparc_init_pic_reg): Likewise.
24175         * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
24176         (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
24178 2018-01-12  Christophe Lyon  <christophe.lyon@linaro.org>
24180         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
24181         Add item for branch_cost.
24183 2018-01-12  Eric Botcazou  <ebotcazou@adacore.com>
24185         PR rtl-optimization/83565
24186         * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
24187         not extend the result to a larger mode for rotate operations.
24188         (num_sign_bit_copies1): Likewise.
24190 2018-01-12  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
24192         PR target/40411
24193         * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
24194         -symbolic.
24195         Use values-Xc.o for -pedantic.
24196         Link with values-xpg4.o for C90, values-xpg6.o otherwise.
24198 2018-01-12  Martin Liska  <mliska@suse.cz>
24200         PR ipa/83054
24201         * ipa-devirt.c (final_warning_record::grow_type_warnings):
24202         New function.
24203         (possible_polymorphic_call_targets): Use it.
24204         (ipa_devirt): Likewise.
24206 2018-01-12  Martin Liska  <mliska@suse.cz>
24208         * profile-count.h (enum profile_quality): Use 0 as invalid
24209         enum value of profile_quality.
24211 2018-01-12  Chung-Ju Wu  <jasonwucj@gmail.com>
24213         * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
24214         -mext-string options.
24216 2018-01-12  Richard Biener  <rguenther@suse.de>
24218         * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
24219         DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
24220         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
24221         Likewise.
24222         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
24224 2018-01-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
24226         * configure.ac (--with-long-double-format): Add support for the
24227         configuration option to change the default long double format on
24228         PowerPC systems.
24229         * config.gcc (powerpc*-linux*-*): Likewise.
24230         * configure: Regenerate.
24231         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
24232         double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
24233         used without modification.
24235 2018-01-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24237         * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
24238         (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
24239         * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
24240         MISC_BUILTIN_SPEC_BARRIER.
24241         (rs6000_init_builtins): Likewise.
24242         * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
24243         enum value.
24244         (speculation_barrier): New define_insn.
24245         * doc/extend.texi: Document __builtin_speculation_barrier.
24247 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
24249         PR target/83203
24250         * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
24251         is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
24252         * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
24253         iterators.
24254         (ssescalarmodesuffix): Add 512-bit vectors.  Use "d" or "q" for
24255         integral modes instead of "ss" and "sd".
24256         (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
24257         vectors with 32-bit and 64-bit elements.
24258         (vecdupssescalarmodesuffix): New mode attribute.
24259         (vec_dup<mode>): Use it.
24261 2018-01-11  H.J. Lu  <hongjiu.lu@intel.com>
24263         PR target/83330
24264         * config/i386/i386.c (ix86_compute_frame_layout): Align stack
24265         frame if argument is passed on stack.
24267 2018-01-11  Jakub Jelinek  <jakub@redhat.com>
24269         PR target/82682
24270         * ree.c (combine_reaching_defs): Optimize also
24271         reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
24272         reg2=any_extend(exp); reg1=reg2;, formatting fix.
24274 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
24276         PR middle-end/83189
24277         * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
24279 2018-01-11  Jan Hubicka  <hubicka@ucw.cz>
24281         PR middle-end/83718
24282         * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
24283         after they are computed.
24285 2018-01-11  Bin Cheng  <bin.cheng@arm.com>
24287         PR tree-optimization/83695
24288         * gimple-loop-linterchange.cc
24289         (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
24290         reset cached scev information after interchange.
24291         (pass_linterchange::execute): Remove call to scev_reset_htab.
24293 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24295         * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
24296         vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
24297         vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
24298         vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
24299         vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
24300         vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
24301         * config/arm/arm_neon_builtins.def (vfmal_lane_low,
24302         vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
24303         vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
24304         vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
24305         vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
24306         * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
24307         (V_lane_reg): Likewise.
24308         * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
24309         New define_expand.
24310         (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
24311         (vfmal_lane_low<mode>_intrinsic,
24312         vfmal_lane_low<vfmlsel2><mode>_intrinsic,
24313         vfmal_lane_high<vfmlsel2><mode>_intrinsic,
24314         vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
24315         vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
24316         vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
24317         vfmsl_lane_high<mode>_intrinsic): New define_insns.
24319 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24321         * config/arm/arm-cpus.in (fp16fml): New feature.
24322         (ALL_SIMD): Add fp16fml.
24323         (armv8.2-a): Add fp16fml as an option.
24324         (armv8.3-a): Likewise.
24325         (armv8.4-a): Add fp16fml as part of fp16.
24326         * config/arm/arm.h (TARGET_FP16FML): Define.
24327         * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
24328         when appropriate.
24329         * config/arm/arm-modes.def (V2HF): Define.
24330         * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
24331         vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
24332         vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
24333         * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
24334         vfmsl_low, vfmsl_high): New set of builtins.
24335         * config/arm/iterators.md (PLUSMINUS): New code iterator.
24336         (vfml_op): New code attribute.
24337         (VFMLHALVES): New int iterator.
24338         (VFML, VFMLSEL): New mode attributes.
24339         (V_reg): Define mapping for V2HF.
24340         (V_hi, V_lo): New mode attributes.
24341         (VF_constraint): Likewise.
24342         (vfml_half, vfml_half_selector): New int attributes.
24343         * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
24344         define_expand.
24345         (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
24346         vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
24347         New define_insn.
24348         * config/arm/t-arm-elf (v8_fps): Add fp16fml.
24349         * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
24350         * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
24351         * doc/invoke.texi (ARM Options): Document fp16fml.  Update armv8.4-a
24352         documentation.
24353         * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
24354         Document new effective target and option set.
24356 2018-01-11  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24358         * config/arm/arm-cpus.in (armv8_4): New feature.
24359         (ARMv8_4a): New fgroup.
24360         (armv8.4-a): New arch.
24361         * config/arm/arm-tables.opt: Regenerate.
24362         * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
24363         * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
24364         * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
24365         Add matching rules for -march=armv8.4-a and extensions.
24366         * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
24368 2018-01-11  Oleg Endo  <olegendo@gcc.gnu.org>
24370         PR target/81821
24371         * config/rx/rx.md (BW): New mode attribute.
24372         (sync_lock_test_and_setsi): Add mode suffix to insn output.
24374 2018-01-11  Richard Biener  <rguenther@suse.de>
24376         PR tree-optimization/83435
24377         * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
24378         * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
24379         * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
24381 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
24382             Alan Hayward  <alan.hayward@arm.com>
24383             David Sherwood  <david.sherwood@arm.com>
24385         * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
24386         field.
24387         (aarch64_classify_address): Initialize it.  Track polynomial offsets.
24388         (aarch64_print_address_internal): Use it to check for a zero offset.
24390 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
24391             Alan Hayward  <alan.hayward@arm.com>
24392             David Sherwood  <david.sherwood@arm.com>
24394         * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
24395         * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
24396         Return a poly_int64 rather than a HOST_WIDE_INT.
24397         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
24398         rather than a HOST_WIDE_INT.
24399         * config/aarch64/aarch64.h (aarch64_frame): Protect with
24400         HAVE_POLY_INT_H rather than HOST_WIDE_INT.  Change locals_offset,
24401         hard_fp_offset, frame_size, initial_adjust, callee_offset and
24402         final_offset from HOST_WIDE_INT to poly_int64.
24403         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
24404         to_constant when getting the number of units in an Advanced SIMD
24405         mode.
24406         (aarch64_builtin_vectorized_function): Check for a constant number
24407         of units.
24408         * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
24409         GET_MODE_SIZE.
24410         (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
24411         attribute instead of GET_MODE_NUNITS.
24412         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
24413         (aarch64_class_max_nregs): Use the constant_lowest_bound of the
24414         GET_MODE_SIZE for fixed-size registers.
24415         (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
24416         (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
24417         (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
24418         (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
24419         (aarch64_print_operand, aarch64_print_address_internal)
24420         (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
24421         (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
24422         (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
24423         Handle polynomial GET_MODE_SIZE.
24424         (aarch64_hard_regno_caller_save_mode): Likewise.  Return modes
24425         wider than SImode without modification.
24426         (tls_symbolic_operand_type): Use strip_offset instead of split_const.
24427         (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
24428         (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
24429         passing and returning SVE modes.
24430         (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
24431         rather than GEN_INT.
24432         (aarch64_emit_probe_stack_range): Take the size as a poly_int64
24433         rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
24434         (aarch64_allocate_and_probe_stack_space): Likewise.
24435         (aarch64_layout_frame): Cope with polynomial offsets.
24436         (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
24437         start_offset as a poly_int64 rather than a HOST_WIDE_INT.  Track
24438         polynomial offsets.
24439         (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
24440         (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
24441         poly_int64 rather than a HOST_WIDE_INT.
24442         (aarch64_get_separate_components, aarch64_process_components)
24443         (aarch64_expand_prologue, aarch64_expand_epilogue)
24444         (aarch64_use_return_insn_p): Handle polynomial frame offsets.
24445         (aarch64_anchor_offset): New function, split out from...
24446         (aarch64_legitimize_address): ...here.
24447         (aarch64_builtin_vectorization_cost): Handle polynomial
24448         TYPE_VECTOR_SUBPARTS.
24449         (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
24450         GET_MODE_NUNITS.
24451         (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
24452         number of elements from the PARALLEL rather than the mode.
24453         (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
24454         rather than GET_MODE_BITSIZE.
24455         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
24456         (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
24457         (aarch64_expand_vec_perm_const_1): Handle polynomial
24458         d->perm.length () and d->perm elements.
24459         (aarch64_evpc_tbl): Likewise.  Use nelt rather than GET_MODE_NUNITS.
24460         Apply to_constant to d->perm elements.
24461         (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
24462         polynomial CONST_VECTOR_NUNITS.
24463         (aarch64_move_pointer): Take amount as a poly_int64 rather
24464         than an int.
24465         (aarch64_progress_pointer): Avoid temporary variable.
24466         * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
24467         the mode attribute instead of GET_MODE.
24469 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
24470             Alan Hayward  <alan.hayward@arm.com>
24471             David Sherwood  <david.sherwood@arm.com>
24473         * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
24474         x exists before using it.
24475         (aarch64_add_constant_internal): Rename to...
24476         (aarch64_add_offset_1): ...this.  Replace regnum with separate
24477         src and dest rtxes.  Handle the case in which they're different,
24478         including when the offset is zero.  Replace scratchreg with an rtx.
24479         Use 2 additions if there is no spare register into which we can
24480         move a 16-bit constant.
24481         (aarch64_add_constant): Delete.
24482         (aarch64_add_offset): Replace reg with separate src and dest
24483         rtxes.  Take a poly_int64 offset instead of a HOST_WIDE_INT.
24484         Use aarch64_add_offset_1.
24485         (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
24486         an rtx rather than an int.  Take the delta as a poly_int64
24487         rather than a HOST_WIDE_INT.  Use aarch64_add_offset.
24488         (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
24489         (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
24490         aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
24491         (aarch64_expand_epilogue): Update calls to aarch64_add_offset
24492         and aarch64_add_sp.
24493         (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
24494         aarch64_add_constant.
24496 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
24498         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
24499         Use scalar_float_mode.
24501 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
24503         * config/aarch64/aarch64-simd.md
24504         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
24505         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
24506         (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
24507         (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
24508         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
24509         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
24510         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
24511         (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
24512         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
24513         (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
24515 2018-01-11  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
24517         PR target/83514
24518         * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
24519         targ_options->x_arm_arch_string is non NULL.
24521 2018-01-11  Tamar Christina  <tamar.christina@arm.com>
24523         * config/aarch64/aarch64.h
24524         (AARCH64_FL_FOR_ARCH8_4): Add  AARCH64_FL_DOTPROD.
24526 2018-01-11  Sudakshina Das  <sudi.das@arm.com>
24528         PR target/82096
24529         * expmed.c (emit_store_flag_force): Swap if const op0
24530         and change VOIDmode to mode of op0.
24532 2018-01-11  Richard Sandiford  <richard.sandiford@linaro.org>
24534         PR rtl-optimization/83761
24535         * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
24536         than bytes to mode_for_size.
24538 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
24540         PR middle-end/83189
24541         * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
24542         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
24543         profile.
24545 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
24547         PR middle-end/83575
24548         * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
24549         when in layout mode.
24550         (cfg_layout_finalize): Do not verify cfg before we are out of layout.
24551         * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
24552         partition fixup.
24554 2018-01-10  Michael Collison  <michael.collison@arm.com>
24556         * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
24557         * config/aarch64/aarch64-option-extension.def: Add
24558         AARCH64_OPT_EXTENSION of 'fp16fml'.
24559         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24560         (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
24561         * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
24562         * config/aarch64/constraints.md (Ui7): New constraint.
24563         * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
24564         (VFMLA_SEL_W): Ditto.
24565         (f16quad): Ditto.
24566         (f16mac1): Ditto.
24567         (VFMLA16_LOW): New int iterator.
24568         (VFMLA16_HIGH): Ditto.
24569         (UNSPEC_FMLAL): New unspec.
24570         (UNSPEC_FMLSL): Ditto.
24571         (UNSPEC_FMLAL2): Ditto.
24572         (UNSPEC_FMLSL2): Ditto.
24573         (f16mac): New code attribute.
24574         * config/aarch64/aarch64-simd-builtins.def
24575         (aarch64_fmlal_lowv2sf): Ditto.
24576         (aarch64_fmlsl_lowv2sf): Ditto.
24577         (aarch64_fmlalq_lowv4sf): Ditto.
24578         (aarch64_fmlslq_lowv4sf): Ditto.
24579         (aarch64_fmlal_highv2sf): Ditto.
24580         (aarch64_fmlsl_highv2sf): Ditto.
24581         (aarch64_fmlalq_highv4sf): Ditto.
24582         (aarch64_fmlslq_highv4sf): Ditto.
24583         (aarch64_fmlal_lane_lowv2sf): Ditto.
24584         (aarch64_fmlsl_lane_lowv2sf): Ditto.
24585         (aarch64_fmlal_laneq_lowv2sf): Ditto.
24586         (aarch64_fmlsl_laneq_lowv2sf): Ditto.
24587         (aarch64_fmlalq_lane_lowv4sf): Ditto.
24588         (aarch64_fmlsl_lane_lowv4sf): Ditto.
24589         (aarch64_fmlalq_laneq_lowv4sf): Ditto.
24590         (aarch64_fmlsl_laneq_lowv4sf): Ditto.
24591         (aarch64_fmlal_lane_highv2sf): Ditto.
24592         (aarch64_fmlsl_lane_highv2sf): Ditto.
24593         (aarch64_fmlal_laneq_highv2sf): Ditto.
24594         (aarch64_fmlsl_laneq_highv2sf): Ditto.
24595         (aarch64_fmlalq_lane_highv4sf): Ditto.
24596         (aarch64_fmlsl_lane_highv4sf): Ditto.
24597         (aarch64_fmlalq_laneq_highv4sf): Ditto.
24598         (aarch64_fmlsl_laneq_highv4sf): Ditto.
24599         * config/aarch64/aarch64-simd.md:
24600         (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
24601         (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
24602         (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
24603         (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
24604         (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
24605         (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
24606         (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
24607         (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
24608         (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
24609         (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
24610         (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
24611         (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
24612         (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
24613         (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
24614         (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
24615         (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
24616         (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
24617         (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
24618         (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
24619         (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
24620         * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
24621         (vfmlsl_low_u32): Ditto.
24622         (vfmlalq_low_u32): Ditto.
24623         (vfmlslq_low_u32): Ditto.
24624         (vfmlal_high_u32): Ditto.
24625         (vfmlsl_high_u32): Ditto.
24626         (vfmlalq_high_u32): Ditto.
24627         (vfmlslq_high_u32): Ditto.
24628         (vfmlal_lane_low_u32): Ditto.
24629         (vfmlsl_lane_low_u32): Ditto.
24630         (vfmlal_laneq_low_u32): Ditto.
24631         (vfmlsl_laneq_low_u32): Ditto.
24632         (vfmlalq_lane_low_u32): Ditto.
24633         (vfmlslq_lane_low_u32): Ditto.
24634         (vfmlalq_laneq_low_u32): Ditto.
24635         (vfmlslq_laneq_low_u32): Ditto.
24636         (vfmlal_lane_high_u32): Ditto.
24637         (vfmlsl_lane_high_u32): Ditto.
24638         (vfmlal_laneq_high_u32): Ditto.
24639         (vfmlsl_laneq_high_u32): Ditto.
24640         (vfmlalq_lane_high_u32): Ditto.
24641         (vfmlslq_lane_high_u32): Ditto.
24642         (vfmlalq_laneq_high_u32): Ditto.
24643         (vfmlslq_laneq_high_u32): Ditto.
24644         * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
24645         (AARCH64_FL_FOR_ARCH8_4): New.
24646         (AARCH64_ISA_F16FML): New ISA flag.
24647         (TARGET_F16FML): New feature flag for fp16fml.
24648         (doc/invoke.texi): Document new fp16fml option.
24650 2018-01-10  Michael Collison  <michael.collison@arm.com>
24652         * config/aarch64/aarch64-builtins.c:
24653         (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
24654         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24655         (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
24656         * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
24657         (AARCH64_ISA_SHA3): New ISA flag.
24658         (TARGET_SHA3): New feature flag for sha3.
24659         * config/aarch64/iterators.md (sha512_op): New int attribute.
24660         (CRYPTO_SHA512): New int iterator.
24661         (UNSPEC_SHA512H): New unspec.
24662         (UNSPEC_SHA512H2): Ditto.
24663         (UNSPEC_SHA512SU0): Ditto.
24664         (UNSPEC_SHA512SU1): Ditto.
24665         * config/aarch64/aarch64-simd-builtins.def
24666         (aarch64_crypto_sha512hqv2di): New builtin.
24667         (aarch64_crypto_sha512h2qv2di): Ditto.
24668         (aarch64_crypto_sha512su0qv2di): Ditto.
24669         (aarch64_crypto_sha512su1qv2di): Ditto.
24670         (aarch64_eor3qv8hi): Ditto.
24671         (aarch64_rax1qv2di): Ditto.
24672         (aarch64_xarqv2di): Ditto.
24673         (aarch64_bcaxqv8hi): Ditto.
24674         * config/aarch64/aarch64-simd.md:
24675         (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
24676         (aarch64_crypto_sha512su0qv2di): Ditto.
24677         (aarch64_crypto_sha512su1qv2di): Ditto.
24678         (aarch64_eor3qv8hi): Ditto.
24679         (aarch64_rax1qv2di): Ditto.
24680         (aarch64_xarqv2di): Ditto.
24681         (aarch64_bcaxqv8hi): Ditto.
24682         * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
24683         (vsha512h2q_u64): Ditto.
24684         (vsha512su0q_u64): Ditto.
24685         (vsha512su1q_u64): Ditto.
24686         (veor3q_u16): Ditto.
24687         (vrax1q_u64): Ditto.
24688         (vxarq_u64): Ditto.
24689         (vbcaxq_u16): Ditto.
24690         * config/arm/types.md (crypto_sha512): New type attribute.
24691         (crypto_sha3): Ditto.
24692         (doc/invoke.texi): Document new sha3 option.
24694 2018-01-10  Michael Collison  <michael.collison@arm.com>
24696         * config/aarch64/aarch64-builtins.c:
24697         (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
24698         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24699         (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
24700         (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
24701         * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
24702         (AARCH64_ISA_SM4): New ISA flag.
24703         (TARGET_SM4): New feature flag for sm4.
24704         * config/aarch64/aarch64-simd-builtins.def
24705         (aarch64_sm3ss1qv4si): Ditto.
24706         (aarch64_sm3tt1aq4si): Ditto.
24707         (aarch64_sm3tt1bq4si): Ditto.
24708         (aarch64_sm3tt2aq4si): Ditto.
24709         (aarch64_sm3tt2bq4si): Ditto.
24710         (aarch64_sm3partw1qv4si): Ditto.
24711         (aarch64_sm3partw2qv4si): Ditto.
24712         (aarch64_sm4eqv4si): Ditto.
24713         (aarch64_sm4ekeyqv4si): Ditto.
24714         * config/aarch64/aarch64-simd.md:
24715         (aarch64_sm3ss1qv4si): Ditto.
24716         (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
24717         (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
24718         (aarch64_sm4eqv4si): Ditto.
24719         (aarch64_sm4ekeyqv4si): Ditto.
24720         * config/aarch64/iterators.md (sm3tt_op): New int iterator.
24721         (sm3part_op): Ditto.
24722         (CRYPTO_SM3TT): Ditto.
24723         (CRYPTO_SM3PART): Ditto.
24724         (UNSPEC_SM3SS1): New unspec.
24725         (UNSPEC_SM3TT1A): Ditto.
24726         (UNSPEC_SM3TT1B): Ditto.
24727         (UNSPEC_SM3TT2A): Ditto.
24728         (UNSPEC_SM3TT2B): Ditto.
24729         (UNSPEC_SM3PARTW1): Ditto.
24730         (UNSPEC_SM3PARTW2): Ditto.
24731         (UNSPEC_SM4E): Ditto.
24732         (UNSPEC_SM4EKEY): Ditto.
24733         * config/aarch64/constraints.md (Ui2): New constraint.
24734         * config/aarch64/predicates.md (aarch64_imm2): New predicate.
24735         * config/arm/types.md (crypto_sm3): New type attribute.
24736         (crypto_sm4): Ditto.
24737         * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
24738         (vsm3tt1aq_u32): Ditto.
24739         (vsm3tt1bq_u32): Ditto.
24740         (vsm3tt2aq_u32): Ditto.
24741         (vsm3tt2bq_u32): Ditto.
24742         (vsm3partw1q_u32): Ditto.
24743         (vsm3partw2q_u32): Ditto.
24744         (vsm4eq_u32): Ditto.
24745         (vsm4ekeyq_u32): Ditto.
24746         (doc/invoke.texi): Document new sm4 option.
24748 2018-01-10  Michael Collison  <michael.collison@arm.com>
24750         * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
24751         * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
24752         (AARCH64_FL_FOR_ARCH8_4): New.
24753         (AARCH64_FL_V8_4): New flag.
24754         (doc/invoke.texi): Document new armv8.4-a option.
24756 2018-01-10  Michael Collison  <michael.collison@arm.com>
24758         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
24759         (__ARM_FEATURE_AES): Define if TARGET_AES is true.
24760         (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
24761         * config/aarch64/aarch64-option-extension.def: Add
24762         AARCH64_OPT_EXTENSION of 'sha2'.
24763         (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
24764         (crypto): Disable sha2 and aes if crypto disabled.
24765         (crypto): Enable aes and sha2 if enabled.
24766         (simd): Disable sha2 and aes if simd disabled.
24767         * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
24768         New flags.
24769         (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
24770         (TARGET_SHA2): New feature flag for sha2.
24771         (TARGET_AES): New feature flag for aes.
24772         * config/aarch64/aarch64-simd.md:
24773         (aarch64_crypto_aes<aes_op>v16qi): Make pattern
24774         conditional on TARGET_AES.
24775         (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
24776         (aarch64_crypto_sha1hsi): Make pattern conditional
24777         on TARGET_SHA2.
24778         (aarch64_crypto_sha1hv4si): Ditto.
24779         (aarch64_be_crypto_sha1hv4si): Ditto.
24780         (aarch64_crypto_sha1su1v4si): Ditto.
24781         (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
24782         (aarch64_crypto_sha1su0v4si): Ditto.
24783         (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
24784         (aarch64_crypto_sha256su0v4si): Ditto.
24785         (aarch64_crypto_sha256su1v4si): Ditto.
24786         (doc/invoke.texi): Document new aes and sha2 options.
24788 2018-01-10  Martin Sebor  <msebor@redhat.com>
24790         PR tree-optimization/83781
24791         * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
24792         as string arrays.
24794 2018-01-11  Martin Sebor  <msebor@gmail.com>
24795             Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
24797         PR tree-optimization/83501
24798         PR tree-optimization/81703
24800         * tree-ssa-strlen.c (get_string_cst): Rename...
24801         (get_string_len): ...to this.  Handle global constants.
24802         (handle_char_store): Adjust.
24804 2018-01-10  Kito Cheng  <kito.cheng@gmail.com>
24805             Jim Wilson  <jimw@sifive.com>
24807         * config/riscv/riscv-protos.h (riscv_output_return): New.
24808         * config/riscv/riscv.c (struct machine_function): New naked_p field.
24809         (riscv_attribute_table, riscv_output_return),
24810         (riscv_handle_fndecl_attribute, riscv_naked_function_p),
24811         (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
24812         (riscv_compute_frame_info): Only compute frame->mask if not a naked
24813         function.
24814         (riscv_expand_prologue): Add early return for naked function.
24815         (riscv_expand_epilogue): Likewise.
24816         (riscv_function_ok_for_sibcall): Return false for naked function.
24817         (riscv_set_current_function): New.
24818         (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
24819         (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
24820         * config/riscv/riscv.md (simple_return): Call riscv_output_return.
24821         * doc/extend.texi (RISC-V Function Attributes): New.
24823 2018-01-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
24825         * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
24826         check for 128-bit long double before checking TCmode.
24827         * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
24828         128-bit long doubles before checking TFmode or TCmode.
24829         (FLOAT128_IBM_P): Likewise.
24831 2018-01-10  Martin Sebor  <msebor@redhat.com>
24833         PR tree-optimization/83671
24834         * builtins.c (c_strlen): Unconditionally return zero for the empty
24835         string.
24836         Use -Warray-bounds for warnings.
24837         * gimple-fold.c (get_range_strlen): Handle non-constant lengths
24838         for non-constant array indices with COMPONENT_REF, arrays of
24839         arrays, and pointers to arrays.
24840         (gimple_fold_builtin_strlen): Determine and set length range for
24841         non-constant character arrays.
24843 2018-01-10  Aldy Hernandez  <aldyh@redhat.com>
24845         PR middle-end/81897
24846         * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
24847         empty blocks.
24849 2018-01-10  Eric Botcazou  <ebotcazou@adacore.com>
24851         * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
24853 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
24855         PR target/83399
24856         * config/rs6000/rs6000.c (print_operand) <'y'>: Use
24857         VECTOR_MEM_ALTIVEC_OR_VSX_P.
24858         * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
24859         indexed_or_indirect_operand predicate.
24860         (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
24861         (*vsx_le_perm_load_v8hi): Likewise.
24862         (*vsx_le_perm_load_v16qi): Likewise.
24863         (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
24864         (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
24865         (*vsx_le_perm_store_v8hi): Likewise.
24866         (*vsx_le_perm_store_v16qi): Likewise.
24867         (eight unnamed splitters): Likewise.
24869 2018-01-10  Peter Bergner  <bergner@vnet.ibm.com>
24871         * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
24872         * config/rs6000/emmintrin.h: Likewise.
24873         * config/rs6000/mmintrin.h: Likewise.
24874         * config/rs6000/xmmintrin.h: Likewise.
24876 2018-01-10  David Malcolm  <dmalcolm@redhat.com>
24878         PR c++/43486
24879         * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
24880         "public_flag".
24881         * tree.c (tree_nop_conversion): Return true for location wrapper
24882         nodes.
24883         (maybe_wrap_with_location): New function.
24884         (selftest::check_strip_nops): New function.
24885         (selftest::test_location_wrappers): New function.
24886         (selftest::tree_c_tests): Call it.
24887         * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
24888         (maybe_wrap_with_location): New decl.
24889         (EXPR_LOCATION_WRAPPER_P): New macro.
24890         (location_wrapper_p): New inline function.
24891         (tree_strip_any_location_wrapper): New inline function.
24893 2018-01-10  H.J. Lu  <hongjiu.lu@intel.com>
24895         PR target/83735
24896         * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
24897         stack_realign_offset for the largest alignment of stack slot
24898         actually used.
24899         (ix86_find_max_used_stack_alignment): New function.
24900         (ix86_finalize_stack_frame_flags): Use it.  Set
24901         max_used_stack_alignment if we don't realign stack.
24902         * config/i386/i386.h (machine_function): Add
24903         max_used_stack_alignment.
24905 2018-01-10  Christophe Lyon  <christophe.lyon@linaro.org>
24907         * config/arm/arm.opt (-mbranch-cost): New option.
24908         * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
24909         account.
24911 2018-01-10  Segher Boessenkool  <segher@kernel.crashing.org>
24913         PR target/83629
24914         * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
24915         load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
24917 2018-01-10  Richard Biener  <rguenther@suse.de>
24919         PR debug/83765
24920         * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
24921         early out so it also covers the case where we have a non-NULL
24922         origin.
24924 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
24926         PR tree-optimization/83753
24927         * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
24928         for non-strided grouped accesses if the number of elements is 1.
24930 2018-01-10  Jan Hubicka  <hubicka@ucw.cz>
24932         PR target/81616
24933         * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
24934         * i386.h (TARGET_USE_GATHER): Define.
24935         * x86-tune.def (X86_TUNE_USE_GATHER): New.
24937 2018-01-10  Martin Liska  <mliska@suse.cz>
24939         PR bootstrap/82831
24940         * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
24941         * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
24942         partitioning.
24943         * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
24944         CLEANUP_NO_PARTITIONING is not set.
24946 2018-01-10  Richard Sandiford  <richard.sandiford@linaro.org>
24948         * doc/rtl.texi: Remove documentation of (const ...) wrappers
24949         for vectors, as a partial revert of r254296.
24950         * rtl.h (const_vec_p): Delete.
24951         (const_vec_duplicate_p): Don't test for vector CONSTs.
24952         (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
24953         * expmed.c (make_tree): Likewise.
24955         Revert:
24956         * common.md (E, F): Use CONSTANT_P instead of checking for
24957         CONST_VECTOR.
24958         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
24959         checking for CONST_VECTOR.
24961 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
24963         PR middle-end/83575
24964         * predict.c (force_edge_cold): Handle in more sane way edges
24965         with no prediction.
24967 2018-01-09  Carl Love  <cel@us.ibm.com>
24969         * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
24970         V4SI, V4SF types.
24971         (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
24972         * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
24973         VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
24974         VMRGOW_V2DI, VMRGOW_V2DF.  Remove definition for VMRGOW.
24975         * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
24976         P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW):  Add definitions.
24977         * config/rs6000/rs6000-protos.h: Add extern defition for
24978         rs6000_generate_float2_double_code.
24979         * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
24980         function.
24981         * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
24982         (float2_v2df): Add define_expand.
24984 2018-01-09  Uros Bizjak  <ubizjak@gmail.com>
24986         PR target/83628
24987         * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
24988         op_mode in the force_to_mode call.
24990 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
24992         * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
24993         instead of checking each element individually.
24994         (aarch64_evpc_uzp): Likewise.
24995         (aarch64_evpc_zip): Likewise.
24996         (aarch64_evpc_ext): Likewise.
24997         (aarch64_evpc_rev): Likewise.
24998         (aarch64_evpc_dup): Test the encoding for a single duplicated element,
24999         instead of checking each element individually.  Return true without
25000         generating rtl if
25001         (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
25002         whether all selected elements come from the same input, instead of
25003         checking each element individually.  Remove calls to gen_rtx_REG,
25004         start_sequence and end_sequence and instead assert that no rtl is
25005         generated.
25007 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
25009         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
25010         order of HIGH and CONST checks.
25012 2018-01-09  Richard Sandiford  <richard.sandiford@linaro.org>
25014         * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
25015         if the destination isn't an SSA_NAME.
25017 2018-01-09  Richard Biener  <rguenther@suse.de>
25019         PR tree-optimization/83668
25020         * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
25021         move prologue...
25022         (canonicalize_loop_form): ... here, renamed from ...
25023         (canonicalize_loop_closed_ssa_form): ... this and amended to
25024         swap successor edges for loop exit blocks to make us use
25025         the RPO order we need for initial schedule generation.
25027 2018-01-09  Joseph Myers  <joseph@codesourcery.com>
25029         PR tree-optimization/64811
25030         * match.pd: When optimizing comparisons with Inf, avoid
25031         introducing or losing exceptions from comparisons with NaN.
25033 2018-01-09  Martin Liska  <mliska@suse.cz>
25035         PR sanitizer/82517
25036         * asan.c (shadow_mem_size): Add gcc_assert.
25038 2018-01-09  Georg-Johann Lay  <avr@gjlay.de>
25040         Don't save registers in main().
25042         PR target/83738
25043         * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
25044         * config/avr/avr.opt (-mmain-is-OS_task): New target option.
25045         * config/avr/avr.c (avr_set_current_function): Don't error if
25046         naked, OS_task or OS_main are specified at the same time.
25047         (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
25048         OS_main.
25049         (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
25050         attribute.
25051         * common/config/avr/avr-common.c (avr_option_optimization_table):
25052         Switch on -mmain-is-OS_task for optimizing compilations.
25054 2018-01-09  Richard Biener  <rguenther@suse.de>
25056         PR tree-optimization/83572
25057         * graphite.c: Include cfganal.h.
25058         (graphite_transform_loops): Connect infinite loops to exit
25059         and remove fake edges at the end.
25061 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
25063         * ipa-inline.c (edge_badness): Revert accidental checkin.
25065 2018-01-09  Jan Hubicka  <hubicka@ucw.cz>
25067         PR ipa/80763
25068         * ipa-comdats.c (set_comdat_group): Only set comdat group of real
25069         symbols; not inline clones.
25071 2018-01-09  Jakub Jelinek  <jakub@redhat.com>
25073         PR target/83507
25074         * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
25075         hard registers.  Formatting fixes.
25077         PR preprocessor/83722
25078         * gcc.c (try_generate_repro): Pass
25079         &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
25080         &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
25081         do_report_bug.
25083 2018-01-08  Monk Chiang  <sh.chiang04@gmail.com>
25084             Kito Cheng  <kito.cheng@gmail.com>
25086         * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
25087         (riscv_leaf_function_p): Delete.
25088         (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
25090 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25092         * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
25093         function.
25094         (do_ifelse): New function.
25095         (do_isel): New function.
25096         (do_sub3): New function.
25097         (do_add3): New function.
25098         (do_load_mask_compare): New function.
25099         (do_overlap_load_compare): New function.
25100         (expand_compare_loop): New function.
25101         (expand_block_compare): Call expand_compare_loop() when appropriate.
25102         * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
25103         option description.
25104         (-mblock-compare-inline-loop-limit): New option.
25106 2018-01-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
25108         PR target/83677
25109         * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
25110         Reverse order of second and third operands in first alternative.
25111         * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
25112         of first and second elements in UNSPEC_VPERMR vector.
25113         (altivec_expand_vec_perm_le): Likewise.
25115 2018-01-08  Jeff Law  <law@redhat.com>
25117         PR rtl-optimizatin/81308
25118         * tree-switch-conversion.c (cfg_altered): New file scoped static.
25119         (process_switch): If group_case_labels makes a change, then set
25120         cfg_altered.
25121         (pass_convert_switch::execute): If a switch is converted, then
25122         set cfg_altered.  Return TODO_cfg_cleanup if cfg_altered is true.
25124         PR rtl-optimization/81308
25125         * recog.c (split_all_insns): Conditionally cleanup the CFG after
25126         splitting insns.
25128 2018-01-08  Vidya Praveen  <vidyapraveen@arm.com>
25130         PR target/83663 - Revert r255946
25131         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
25132         generation for cases where splatting a value is not useful.
25133         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
25134         across a vec_duplicate and a paradoxical subreg forming a vector
25135         mode to a vec_concat.
25137 2018-01-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
25139         * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
25140         -march=armv8.3-a variants.
25141         * config/arm/t-multilib: Likewise.
25142         * config/arm/t-arm-elf: Likewise.  Handle dotprod extension.
25144 2018-01-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25146         * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
25147         to generate rtl.
25148         (cceq_ior_compare_complement): Give it a name so I can use it, and
25149         change boolean_or_operator predicate to boolean_operator so it can
25150         be used to generate a crand.
25151         (eqne): New code iterator.
25152         (bd/bd_neg): New code_attrs.
25153         (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
25154         a single define_insn.
25155         (<bd>tf_<mode>): A new insn pattern for the conditional form branch
25156         decrement (bdnzt/bdnzf/bdzt/bdzf).
25157         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
25158         with the new names of the branch decrement patterns, and added the
25159         names of the branch decrement conditional patterns.
25161 2018-01-08  Richard Biener  <rguenther@suse.de>
25163         PR tree-optimization/83563
25164         * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
25165         cache.
25167 2018-01-08  Richard Biener  <rguenther@suse.de>
25169         PR middle-end/83713
25170         * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
25172 2018-01-08  Richard Biener  <rguenther@suse.de>
25174         PR tree-optimization/83685
25175         * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
25176         references to abnormals.
25178 2018-01-08  Richard Biener  <rguenther@suse.de>
25180         PR lto/83719
25181         * dwarf2out.c (output_indirect_strings): Handle empty
25182         skeleton_debug_str_hash.
25183         (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
25185 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
25187         * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
25188         (emit_store_direct): Likewise.
25189         (arc_trampoline_adjust_address): Likewise.
25190         (arc_asm_trampoline_template): New function.
25191         (arc_initialize_trampoline): Use asm_trampoline_template.
25192         (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
25193         * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
25194         * config/arc/arc.md (flush_icache): Delete pattern.
25196 2018-01-08  Claudiu Zissulescu  <claziss@synopsys.com>
25198         * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
25199         * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
25200         munaligned-access.
25202 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
25204         PR target/83681
25205         * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
25206         by not USED_FOR_TARGET.
25207         (make_pass_resolve_sw_modes): Likewise.
25209 2018-01-08  Sebastian Huber  <sebastian.huber@embedded-brains.de>
25211         * config/nios2/nios2.h (nios2_section_threshold): Guard by not
25212         USED_FOR_TARGET.
25214 2018-01-08  Richard Biener  <rguenther@suse.de>
25216         PR middle-end/83580
25217         * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
25219 2018-01-08  Richard Biener  <rguenther@suse.de>
25221         PR middle-end/83517
25222         * match.pd ((t * 2) / 2) -> t): Add missing :c.
25224 2018-01-06  Aldy Hernandez  <aldyh@redhat.com>
25226         PR middle-end/81897
25227         * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
25228         basic blocks with a small number of successors.
25229         (convert_control_dep_chain_into_preds): Improve handling of
25230         forwarder blocks.
25231         (dump_predicates): Split apart into...
25232         (dump_pred_chain): ...here...
25233         (dump_pred_info): ...and here.
25234         (can_one_predicate_be_invalidated_p): Add debugging printfs.
25235         (can_chain_union_be_invalidated_p): Improve check for invalidation
25236         of paths.
25237         (uninit_uses_cannot_happen): Avoid unnecessary if
25238         convert_control_dep_chain_into_preds yielded nothing.
25240 2018-01-06  Martin Sebor  <msebor@redhat.com>
25242         PR tree-optimization/83640
25243         * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
25244         subtracting negative offset from size.
25245         (builtin_access::overlap): Adjust offset bounds of the access to fall
25246         within the size of the object if possible.
25248 2018-01-06  Richard Sandiford  <richard.sandiford@linaro.org>
25250         PR rtl-optimization/83699
25251         * expmed.c (extract_bit_field_1): Restrict the vector usage of
25252         extract_bit_field_as_subreg to cases in which the extracted
25253         value is also a vector.
25255         * lra-constraints.c (process_alt_operands): Test for the equivalence
25256         substitutions when detecting a possible reload cycle.
25258 2018-01-06  Jakub Jelinek  <jakub@redhat.com>
25260         PR debug/83480
25261         * toplev.c (process_options): Don't enable debug_nonbind_markers_p
25262         by default if flag_selective_schedling{,2}.  Formatting fixes.
25264         PR rtl-optimization/83682
25265         * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
25266         if it has non-VECTOR_MODE element mode.
25267         (vec_duplicate_p): Likewise.
25269         PR middle-end/83694
25270         * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
25271         and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
25273 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
25275         PR target/83604
25276         * config/i386/i386-builtin.def
25277         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
25278         __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
25279         Require also OPTION_MASK_ISA_AVX512F in addition to
25280         OPTION_MASK_ISA_GFNI.
25281         (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
25282         __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
25283         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
25284         to OPTION_MASK_ISA_GFNI.
25285         (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
25286         OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
25287         OPTION_MASK_ISA_AVX512BW.
25288         (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
25289         OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
25290         addition to OPTION_MASK_ISA_GFNI.
25291         (__builtin_ia32_vgf2p8affineinvqb_v16qi,
25292         __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
25293         Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
25294         to OPTION_MASK_ISA_GFNI.
25295         * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
25296         a requirement for all ISAs rather than any of them with a few
25297         exceptions.
25298         (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
25299         processing.
25300         (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
25301         bitmasks to be enabled with 3 exceptions, instead of requiring any
25302         enabled ISA with lots of exceptions.
25303         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
25304         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
25305         Change avx512bw in isa attribute to avx512f.
25306         * config/i386/sgxintrin.h: Add license boilerplate.
25307         * config/i386/vaesintrin.h: Likewise.  Fix macro spelling __AVX512F
25308         to __AVX512F__ and __AVX512VL to __AVX512VL__.
25309         (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
25310         _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
25311         defined.
25312         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
25313         _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
25314         temporarily sse2 rather than sse if not enabled already.
25316         PR target/83604
25317         * config/i386/sse.md (VI248_VLBW): Rename to ...
25318         (VI248_AVX512VL): ... this.  Don't guard V32HI with TARGET_AVX512BW.
25319         (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
25320         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
25321         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
25322         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
25323         mode iterator instead of VI248_VLBW.
25325 2018-01-05  Jan Hubicka  <hubicka@ucw.cz>
25327         * ipa-fnsummary.c (record_modified_bb_info): Add OP.
25328         (record_modified): Skip clobbers; add debug output.
25329         (param_change_prob): Use sreal frequencies.
25331 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
25333         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
25334         punt for user-aligned variables.
25336 2018-01-05  Richard Sandiford  <richard.sandiford@linaro.org>
25338         * tree-chrec.c (chrec_contains_symbols): Return true for
25339         POLY_INT_CST.
25341 2018-01-05  Sudakshina Das  <sudi.das@arm.com>
25343         PR target/82439
25344         * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
25345         of (x|y) == x for BICS pattern.
25347 2018-01-05  Jakub Jelinek  <jakub@redhat.com>
25349         PR tree-optimization/83605
25350         * gimple-ssa-strength-reduction.c: Include tree-eh.h.
25351         (find_candidates_dom_walker::before_dom_children): Ignore stmts that
25352         can throw.
25354 2018-01-05  Sebastian Huber  <sebastian.huber@embedded-brains.de>
25356         * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
25357         * config/epiphany/rtems.h: New file.
25359 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
25360             Uros Bizjak  <ubizjak@gmail.com>
25362         PR target/83554
25363         * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
25364         QIreg_operand instead of register_operand predicate.
25365         * config/i386/i386.c (ix86_rop_should_change_byte_p,
25366         set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
25367         comments instead of -fmitigate[-_]rop.
25369 2018-01-04  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
25371         PR bootstrap/81926
25372         * cgraphunit.c (symbol_table::compile): Switch to text_section
25373         before calling assembly_start debug hook.
25374         * run-rtl-passes.c (run_rtl_passes): Likewise.
25375         Include output.h.
25377 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
25379         * tree-vrp.c (extract_range_from_binary_expr_1): Check
25380         range_int_cst_p rather than !symbolic_range_p before calling
25381         extract_range_from_multiplicative_op_1.
25383 2018-01-04  Jeff Law  <law@redhat.com>
25385         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
25386         redundant test in assertion.
25388 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
25390         * doc/rtl.texi: Document machine_mode wrapper classes.
25392 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
25394         * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
25395         using tree_to_uhwi.
25397 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
25399         * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
25400         the VEC_PERM_EXPR fold to fail.
25402 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
25404         PR debug/83585
25405         * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
25406         to switched_sections.
25408 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
25410         PR target/83680
25411         * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
25412         test for d.testing.
25414 2018-01-04  Peter Bergner  <bergner@vnet.ibm.com>
25416         PR target/83387
25417         * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
25418         allow arguments in FP registers if TARGET_HARD_FLOAT is false.
25420 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
25422         PR debug/83666
25423         * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
25424         is BLKmode and bitpos not zero or mode change is needed.
25426 2018-01-04  Richard Sandiford  <richard.sandiford@linaro.org>
25428         PR target/83675
25429         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
25430         TARGET_VIS2.
25432 2018-01-04  Uros Bizjak  <ubizjak@gmail.com>
25434         PR target/83628
25435         * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
25436         instead of MULT rtx.  Update all corresponding splitters.
25437         (*saddl_se): Ditto.
25438         (*ssub<modesuffix>): Ditto.
25439         (*ssubl_se): Ditto.
25440         (*cmp_sadd_di): Update split patterns.
25441         (*cmp_sadd_si): Ditto.
25442         (*cmp_sadd_sidi): Ditto.
25443         (*cmp_ssub_di): Ditto.
25444         (*cmp_ssub_si): Ditto.
25445         (*cmp_ssub_sidi): Ditto.
25446         * config/alpha/predicates.md (const23_operand): New predicate.
25447         * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
25448         Look for ASHIFT, not MULT inner operand.
25449         (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
25451 2018-01-04  Martin Liska  <mliska@suse.cz>
25453         PR gcov-profile/83669
25454         * gcov.c (output_intermediate_file): Add version to intermediate
25455         gcov file.
25456         * doc/gcov.texi: Document new field 'version' in intermediate
25457         file format. Fix location of '-k' option of gcov command.
25459 2018-01-04  Martin Liska  <mliska@suse.cz>
25461         PR ipa/82352
25462         * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
25464 2018-01-04  Jakub Jelinek  <jakub@redhat.com>
25466         * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
25468 2018-01-03  Martin Sebor  <msebor@redhat.com>
25470         PR tree-optimization/83655
25471         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
25472         checking calls with invalid arguments.
25474 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25476         * tree-vect-stmts.c (vect_get_store_rhs): New function.
25477         (vectorizable_mask_load_store): Delete.
25478         (vectorizable_call): Return false for masked loads and stores.
25479         (vectorizable_store): Handle IFN_MASK_STORE.  Use vect_get_store_rhs
25480         instead of gimple_assign_rhs1.
25481         (vectorizable_load): Handle IFN_MASK_LOAD.
25482         (vect_transform_stmt): Don't set is_store for call_vec_info_type.
25484 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25486         * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
25487         split out from..,
25488         (vectorizable_mask_load_store): ...here.
25489         (vectorizable_load): ...and here.
25491 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25493         * tree-vect-stmts.c (vect_build_all_ones_mask)
25494         (vect_build_zero_merge_argument): New functions, split out from...
25495         (vectorizable_load): ...here.
25497 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25499         * tree-vect-stmts.c (vect_check_store_rhs): New function,
25500         split out from...
25501         (vectorizable_mask_load_store): ...here.
25502         (vectorizable_store): ...and here.
25504 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25506         * tree-vect-stmts.c (vect_check_load_store_mask): New function,
25507         split out from...
25508         (vectorizable_mask_load_store): ...here.
25510 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25512         * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
25513         (vect_model_store_cost): Take a vec_load_store_type instead of a
25514         vect_def_type.
25515         * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
25516         (vect_model_store_cost): Take a vec_load_store_type instead of a
25517         vect_def_type.
25518         (vectorizable_mask_load_store): Update accordingly.
25519         (vectorizable_store): Likewise.
25520         * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
25522 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25524         * tree-vect-loop.c (vect_transform_loop): Stub out scalar
25525         IFN_MASK_LOAD calls here rather than...
25526         * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
25528 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25529             Alan Hayward  <alan.hayward@arm.com>
25530             David Sherwood  <david.sherwood@arm.com>
25532         * expmed.c (extract_bit_field_1): For vector extracts,
25533         fall back to extract_bit_field_as_subreg if vec_extract
25534         isn't available.
25536 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25537             Alan Hayward  <alan.hayward@arm.com>
25538             David Sherwood  <david.sherwood@arm.com>
25540         * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
25541         they are variable or constant sized.
25542         (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
25543         slots for constant-sized data.
25545 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25546             Alan Hayward  <alan.hayward@arm.com>
25547             David Sherwood  <david.sherwood@arm.com>
25549         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
25550         handling COND_EXPRs with boolean comparisons, try to find a better
25551         basis for the mask type than the boolean itself.
25553 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25555         * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
25556         is calculated and how it can be overridden.
25557         * genmodes.c (max_bitsize_mode_any_mode): New variable.
25558         (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
25559         if defined.
25560         (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
25561         if nonzero.
25563 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25564             Alan Hayward  <alan.hayward@arm.com>
25565             David Sherwood  <david.sherwood@arm.com>
25567         * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
25568         Remove the mode argument.
25569         (aarch64_simd_valid_immediate): Remove the mode and inverse
25570         arguments.
25571         * config/aarch64/iterators.md (bitsize): New iterator.
25572         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
25573         (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
25574         * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
25575         aarch64_simd_valid_immediate.
25576         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
25577         (aarch64_reg_or_bic_imm): Likewise.
25578         * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
25579         with an insn_type enum and msl with a modifier_type enum.
25580         Replace element_width with a scalar_mode.  Change the shift
25581         to unsigned int.  Add constructors for scalar_float_mode and
25582         scalar_int_mode elements.
25583         (aarch64_vect_float_const_representable_p): Delete.
25584         (aarch64_can_const_movi_rtx_p)
25585         (aarch64_simd_scalar_immediate_valid_for_move)
25586         (aarch64_simd_make_constant): Update call to
25587         aarch64_simd_valid_immediate.
25588         (aarch64_advsimd_valid_immediate_hs): New function.
25589         (aarch64_advsimd_valid_immediate): Likewise.
25590         (aarch64_simd_valid_immediate): Remove mode and inverse
25591         arguments.  Rewrite to use the above.  Use const_vec_duplicate_p
25592         to detect duplicated constants and use aarch64_float_const_zero_rtx_p
25593         and aarch64_float_const_representable_p on the result.
25594         (aarch64_output_simd_mov_immediate): Remove mode argument.
25595         Update call to aarch64_simd_valid_immediate and use of
25596         simd_immediate_info.
25597         (aarch64_output_scalar_simd_mov_immediate): Update call
25598         accordingly.
25600 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25601             Alan Hayward  <alan.hayward@arm.com>
25602             David Sherwood  <david.sherwood@arm.com>
25604         * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
25605         (mode_nunits): Likewise CONST_MODE_NUNITS.
25606         * machmode.def (ADJUST_NUNITS): Document.
25607         * genmodes.c (mode_data::need_nunits_adj): New field.
25608         (blank_mode): Update accordingly.
25609         (adj_nunits): New variable.
25610         (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
25611         parameter.
25612         (emit_mode_size_inline): Set need_bytesize_adj for all modes
25613         listed in adj_nunits.
25614         (emit_mode_nunits_inline): Set need_nunits_adj for all modes
25615         listed in adj_nunits.  Don't emit case statements for such modes.
25616         (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
25617         and CONST_MODE_PRECISION.  Make CONST_MODE_SIZE expand to
25618         nothing if adj_nunits is nonnull.
25619         (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
25620         (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
25621         (emit_mode_fbit): Update use of print_maybe_const_decl.
25622         (emit_move_size): Likewise.  Treat the array as non-const
25623         if adj_nunits.
25624         (emit_mode_adjustments): Handle adj_nunits.
25626 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25628         * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
25629         * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
25630         (VECTOR_MODES): Use it.
25631         (make_vector_modes): Take the prefix as an argument.
25633 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25634             Alan Hayward  <alan.hayward@arm.com>
25635             David Sherwood  <david.sherwood@arm.com>
25637         * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
25638         * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
25639         for MODE_VECTOR_BOOL.
25640         * machmode.def (VECTOR_BOOL_MODE): Document.
25641         * genmodes.c (VECTOR_BOOL_MODE): New macro.
25642         (make_vector_bool_mode): New function.
25643         (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
25644         MODE_VECTOR_BOOL.
25645         * lto-streamer-in.c (lto_input_mode_table): Likewise.
25646         * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
25647         Likewise.
25648         * stor-layout.c (int_mode_for_mode): Likewise.
25649         * tree.c (build_vector_type_for_mode): Likewise.
25650         * varasm.c (output_constant_pool_2): Likewise.
25651         * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
25652         CONSTM1_RTX (BImode) are the same thing.  Initialize const_tiny_rtx
25653         for MODE_VECTOR_BOOL.
25654         * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
25655         of mode class checks.
25656         * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
25657         instead of a list of mode class checks.
25658         (expand_vector_scalar_condition): Likewise.
25659         (type_for_widest_vector_mode): Handle BImode as an inner mode.
25661 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25662             Alan Hayward  <alan.hayward@arm.com>
25663             David Sherwood  <david.sherwood@arm.com>
25665         * machmode.h (mode_size): Change from unsigned short to
25666         poly_uint16_pod.
25667         (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
25668         (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
25669         or if measurement_type is not polynomial.
25670         (fixed_size_mode::includes_p): Check for constant-sized modes.
25671         * genmodes.c (emit_mode_size_inline): Make mode_size_inline
25672         return a poly_uint16 rather than an unsigned short.
25673         (emit_mode_size): Change the type of mode_size from unsigned short
25674         to poly_uint16_pod.  Use ZERO_COEFFS for the initializer.
25675         (emit_mode_adjustments): Cope with polynomial vector sizes.
25676         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
25677         for GET_MODE_SIZE.
25678         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
25679         for GET_MODE_SIZE.
25680         * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
25681         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
25682         * caller-save.c (setup_save_areas): Likewise.
25683         (replace_reg_with_saved_mem): Likewise.
25684         * calls.c (emit_library_call_value_1): Likewise.
25685         * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
25686         * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
25687         (gen_lowpart_for_combine): Likewise.
25688         * convert.c (convert_to_integer_1): Likewise.
25689         * cse.c (equiv_constant, cse_insn): Likewise.
25690         * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
25691         (cselib_subst_to_values): Likewise.
25692         * dce.c (word_dce_process_block): Likewise.
25693         * df-problems.c (df_word_lr_mark_ref): Likewise.
25694         * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
25695         * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
25696         (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
25697         (rtl_for_decl_location): Likewise.
25698         * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
25699         * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
25700         * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
25701         (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
25702         (expand_expr_real_1): Likewise.
25703         * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
25704         (pad_below): Likewise.
25705         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
25706         * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
25707         * ira.c (get_subreg_tracking_sizes): Likewise.
25708         * ira-build.c (ira_create_allocno_objects): Likewise.
25709         * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
25710         (ira_sort_regnos_for_alter_reg): Likewise.
25711         * ira-costs.c (record_operand_costs): Likewise.
25712         * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
25713         (resolve_simple_move): Likewise.
25714         * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
25715         (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
25716         (lra_constraints): Likewise.
25717         (CONST_POOL_OK_P): Reject variable-sized modes.
25718         * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
25719         (add_pseudo_to_slot, lra_spill): Likewise.
25720         * omp-low.c (omp_clause_aligned_alignment): Likewise.
25721         * optabs-query.c (get_best_extraction_insn): Likewise.
25722         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
25723         * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
25724         (expand_mult_highpart, valid_multiword_target_p): Likewise.
25725         * recog.c (offsettable_address_addr_space_p): Likewise.
25726         * regcprop.c (maybe_mode_change): Likewise.
25727         * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
25728         * regrename.c (build_def_use): Likewise.
25729         * regstat.c (dump_reg_info): Likewise.
25730         * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
25731         (find_reloads, find_reloads_subreg_address): Likewise.
25732         * reload1.c (eliminate_regs_1): Likewise.
25733         * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
25734         * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
25735         (simplify_binary_operation_1, simplify_subreg): Likewise.
25736         * targhooks.c (default_function_arg_padding): Likewise.
25737         (default_hard_regno_nregs, default_class_max_nregs): Likewise.
25738         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
25739         (verify_gimple_assign_ternary): Likewise.
25740         * tree-inline.c (estimate_move_cost): Likewise.
25741         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25742         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
25743         (get_address_cost_ainc): Likewise.
25744         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
25745         (vect_supportable_dr_alignment): Likewise.
25746         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
25747         (vectorizable_reduction): Likewise.
25748         * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
25749         (vectorizable_operation, vectorizable_load): Likewise.
25750         * tree.c (build_same_sized_truth_vector_type): Likewise.
25751         * valtrack.c (cleanup_auto_inc_dec): Likewise.
25752         * var-tracking.c (emit_note_insn_var_location): Likewise.
25753         * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
25754         (ADDR_VEC_ALIGN): Likewise.
25756 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25757             Alan Hayward  <alan.hayward@arm.com>
25758             David Sherwood  <david.sherwood@arm.com>
25760         * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
25761         unsigned short.
25762         (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
25763         or if measurement_type is polynomial.
25764         * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
25765         * combine.c (make_extraction): Likewise.
25766         * dse.c (find_shift_sequence): Likewise.
25767         * dwarf2out.c (mem_loc_descriptor): Likewise.
25768         * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
25769         (extract_bit_field, extract_low_bits): Likewise.
25770         * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
25771         (optimize_bitfield_assignment_op, expand_assignment): Likewise.
25772         (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
25773         * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
25774         * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
25775         * reload.c (find_reloads): Likewise.
25776         * reload1.c (alter_reg): Likewise.
25777         * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
25778         * targhooks.c (default_secondary_memory_needed_mode): Likewise.
25779         * tree-if-conv.c (predicate_mem_writes): Likewise.
25780         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
25781         * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
25782         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
25783         * valtrack.c (dead_debug_insert_temp): Likewise.
25784         * varasm.c (mergeable_constant_section): Likewise.
25785         * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
25787 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25788             Alan Hayward  <alan.hayward@arm.com>
25789             David Sherwood  <david.sherwood@arm.com>
25791         * expr.c (expand_assignment): Cope with polynomial mode sizes
25792         when assigning to a CONCAT.
25794 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25795             Alan Hayward  <alan.hayward@arm.com>
25796             David Sherwood  <david.sherwood@arm.com>
25798         * machmode.h (mode_precision): Change from unsigned short to
25799         poly_uint16_pod.
25800         (mode_to_precision): Return a poly_uint16 rather than an unsigned
25801         short.
25802         (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
25803         or if measurement_type is not polynomial.
25804         (HWI_COMPUTABLE_MODE_P): Turn into a function.  Optimize the case
25805         in which the mode is already known to be a scalar_int_mode.
25806         * genmodes.c (emit_mode_precision): Change the type of mode_precision
25807         from unsigned short to poly_uint16_pod.  Use ZERO_COEFFS for the
25808         initializer.
25809         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
25810         for GET_MODE_PRECISION.
25811         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
25812         for GET_MODE_PRECISION.
25813         * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
25814         as polynomial.
25815         (try_combine, find_split_point, combine_simplify_rtx): Likewise.
25816         (expand_field_assignment, make_extraction): Likewise.
25817         (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
25818         (get_last_value): Likewise.
25819         * convert.c (convert_to_integer_1): Likewise.
25820         * cse.c (cse_insn): Likewise.
25821         * expr.c (expand_expr_real_1): Likewise.
25822         * lra-constraints.c (simplify_operand_subreg): Likewise.
25823         * optabs-query.c (can_atomic_load_p): Likewise.
25824         * optabs.c (expand_atomic_load): Likewise.
25825         (expand_atomic_store): Likewise.
25826         * ree.c (combine_reaching_defs): Likewise.
25827         * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
25828         * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
25829         * tree.h (type_has_mode_precision_p): Likewise.
25830         * ubsan.c (instrument_si_overflow): Likewise.
25832 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25833             Alan Hayward  <alan.hayward@arm.com>
25834             David Sherwood  <david.sherwood@arm.com>
25836         * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
25837         polynomial numbers of units.
25838         (SET_TYPE_VECTOR_SUBPARTS): Likewise.
25839         (valid_vector_subparts_p): New function.
25840         (build_vector_type): Remove temporary shim and take the number
25841         of units as a poly_uint64 rather than an int.
25842         (build_opaque_vector_type): Take the number of units as a
25843         poly_uint64 rather than an int.
25844         * tree.c (build_vector_from_ctor): Handle polynomial
25845         TYPE_VECTOR_SUBPARTS.
25846         (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
25847         (uniform_vector_p, vector_type_mode, build_vector): Likewise.
25848         (build_vector_from_val): If the number of units is variable,
25849         use build_vec_duplicate_cst for constant operands and
25850         VEC_DUPLICATE_EXPR otherwise.
25851         (make_vector_type): Remove temporary is_constant ().
25852         (build_vector_type, build_opaque_vector_type): Take the number of
25853         units as a poly_uint64 rather than an int.
25854         (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
25855         VECTOR_CST_NELTS.
25856         * cfgexpand.c (expand_debug_expr): Likewise.
25857         * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
25858         (store_constructor, expand_expr_real_1): Likewise.
25859         (const_scalar_mask_from_tree): Likewise.
25860         * fold-const-call.c (fold_const_reduction): Likewise.
25861         * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
25862         (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
25863         (native_encode_vector, vec_cst_ctor_to_array): Likewise.
25864         (fold_relational_const): Likewise.
25865         (native_interpret_vector): Likewise.  Change the size from an
25866         int to an unsigned int.
25867         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
25868         TYPE_VECTOR_SUBPARTS.
25869         (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
25870         (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
25871         duplicating a non-constant operand into a variable-length vector.
25872         * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
25873         TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
25874         * ipa-icf.c (sem_variable::equals): Likewise.
25875         * match.pd: Likewise.
25876         * omp-simd-clone.c (simd_clone_subparts): Likewise.
25877         * print-tree.c (print_node): Likewise.
25878         * stor-layout.c (layout_type): Likewise.
25879         * targhooks.c (default_builtin_vectorization_cost): Likewise.
25880         * tree-cfg.c (verify_gimple_comparison): Likewise.
25881         (verify_gimple_assign_binary): Likewise.
25882         (verify_gimple_assign_ternary): Likewise.
25883         (verify_gimple_assign_single): Likewise.
25884         * tree-pretty-print.c (dump_generic_node): Likewise.
25885         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
25886         (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
25887         * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
25888         (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
25889         (vect_shift_permute_load_chain): Likewise.
25890         * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
25891         (expand_vector_condition, optimize_vector_constructor): Likewise.
25892         (lower_vec_perm, get_compute_type): Likewise.
25893         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
25894         (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
25895         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
25896         (vect_recog_mask_conversion_pattern): Likewise.
25897         * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
25898         (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
25899         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
25900         (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
25901         (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
25902         (vectorizable_shift, vectorizable_operation, vectorizable_store)
25903         (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
25904         (supportable_widening_operation): Likewise.
25905         (supportable_narrowing_operation): Likewise.
25906         * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
25907         Likewise.
25908         * varasm.c (output_constant): Likewise.
25910 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25911             Alan Hayward  <alan.hayward@arm.com>
25912             David Sherwood  <david.sherwood@arm.com>
25914         * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
25915         so that both the length == 3 and length != 3 cases set up their
25916         own permute vectors.  Add comments explaining why we know the
25917         number of elements is constant.
25918         (vect_permute_load_chain): Likewise.
25920 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25921             Alan Hayward  <alan.hayward@arm.com>
25922             David Sherwood  <david.sherwood@arm.com>
25924         * machmode.h (mode_nunits): Change from unsigned char to
25925         poly_uint16_pod.
25926         (ONLY_FIXED_SIZE_MODES): New macro.
25927         (pod_mode::measurement_type, scalar_int_mode::measurement_type)
25928         (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
25929         (complex_mode::measurement_type, fixed_size_mode::measurement_type):
25930         New typedefs.
25931         (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
25932         (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
25933         or if measurement_type is not polynomial.
25934         * genmodes.c (ZERO_COEFFS): New macro.
25935         (emit_mode_nunits_inline): Make mode_nunits_inline return a
25936         poly_uint16.
25937         (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
25938         Use ZERO_COEFFS when emitting initializers.
25939         * data-streamer.h (bp_pack_poly_value): New function.
25940         (bp_unpack_poly_value): Likewise.
25941         * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
25942         for GET_MODE_NUNITS.
25943         * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
25944         for GET_MODE_NUNITS.
25945         * tree.c (make_vector_type): Remove temporary shim and make
25946         the real function take the number of units as a poly_uint64
25947         rather than an int.
25948         (build_vector_type_for_mode): Handle polynomial nunits.
25949         * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
25950         * emit-rtl.c (const_vec_series_p_1): Likewise.
25951         (gen_rtx_CONST_VECTOR): Likewise.
25952         * fold-const.c (test_vec_duplicate_folding): Likewise.
25953         * genrecog.c (validate_pattern): Likewise.
25954         * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
25955         * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
25956         * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
25957         (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
25958         (expand_vec_cond_expr, expand_mult_highpart): Likewise.
25959         * rtlanal.c (subreg_get_info): Likewise.
25960         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
25961         (vect_grouped_load_supported): Likewise.
25962         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
25963         * tree-vect-loop.c (have_whole_vector_shift): Likewise.
25964         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
25965         (simplify_const_unary_operation, simplify_binary_operation_1)
25966         (simplify_const_binary_operation, simplify_ternary_operation)
25967         (test_vector_ops_duplicate, test_vector_ops): Likewise.
25968         (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
25969         instead of CONST_VECTOR_NUNITS.
25970         * varasm.c (output_constant_pool_2): Likewise.
25971         * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
25972         explicit-encoded elements in the XVEC for variable-length vectors.
25974 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25976         * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
25978 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25979             Alan Hayward  <alan.hayward@arm.com>
25980             David Sherwood  <david.sherwood@arm.com>
25982         * coretypes.h (fixed_size_mode): Declare.
25983         (fixed_size_mode_pod): New typedef.
25984         * builtins.h (target_builtins::x_apply_args_mode)
25985         (target_builtins::x_apply_result_mode): Change type to
25986         fixed_size_mode_pod.
25987         * builtins.c (apply_args_size, apply_result_size, result_vector)
25988         (expand_builtin_apply_args_1, expand_builtin_apply)
25989         (expand_builtin_return): Update accordingly.
25991 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
25993         * cse.c (hash_rtx_cb): Hash only the encoded elements.
25994         * cselib.c (cselib_hash_rtx): Likewise.
25995         * expmed.c (make_tree): Build VECTOR_CSTs directly from the
25996         CONST_VECTOR encoding.
25998 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
25999             Jeff Law  <law@redhat.com>
26001         PR target/83641
26002         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
26003         noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
26004         only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
26005         and add REG_CFA_ADJUST_CFA notes in that case to both insns.
26007         PR target/83641
26008         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
26009         explicitly probe *sp in a noreturn function if there were any callee
26010         register saves or frame pointer is needed.
26012 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
26014         PR debug/83621
26015         * cfgexpand.c (expand_debug_expr): Return NULL if mode is
26016         BLKmode for ternary, binary or unary expressions.
26018         PR debug/83645
26019         * var-tracking.c (delete_vta_debug_insn): New inline function.
26020         (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
26021         insns from get_insns () to NULL instead of each bb separately.
26022         Use delete_vta_debug_insn.  No longer static.
26023         (vt_debug_insns_local, variable_tracking_main_1): Adjust
26024         delete_vta_debug_insns callers.
26025         * rtl.h (delete_vta_debug_insns): Declare.
26026         * final.c (rest_of_handle_final): Call delete_vta_debug_insns
26027         instead of variable_tracking_main.
26029 2018-01-03  Martin Sebor  <msebor@redhat.com>
26031         PR tree-optimization/83603
26032         * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
26033         arguments past the endof the argument list in functions declared
26034         without a prototype.
26035         * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
26036         Avoid checking when arguments are null.
26038 2018-01-03  Martin Sebor  <msebor@redhat.com>
26040         PR c/83559
26041         * doc/extend.texi (attribute const): Fix a typo.
26042         * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
26043         issuing -Wsuggest-attribute for void functions.
26045 2018-01-03  Martin Sebor  <msebor@redhat.com>
26047         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
26048         offset_int::from instead of wide_int::to_shwi.
26049         (maybe_diag_overlap): Remove assertion.
26050         Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
26051         * gimple-ssa-sprintf.c (format_directive): Same.
26052         (parse_directive): Same.
26053         (sprintf_dom_walker::compute_format_length): Same.
26054         (try_substitute_return_value): Same.
26056 2018-01-03  Jeff Law  <law@redhat.com>
26058         PR middle-end/83654
26059         * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
26060         non-constant residual for zero at runtime and avoid probing in
26061         that case.  Reorganize code for trailing problem to mirror handling
26062         of the residual.
26064 2018-01-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
26066         PR tree-optimization/83501
26067         * tree-ssa-strlen.c (get_string_cst): New.
26068         (handle_char_store): Call get_string_cst.
26070 2018-01-03  Martin Liska  <mliska@suse.cz>
26072         PR tree-optimization/83593
26073         * tree-ssa-strlen.c: Include tree-cfg.h.
26074         (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
26075         (strlen_dom_walker): Add new member variable m_cleanup_cfg.
26076         (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
26077         to false.
26078         (strlen_dom_walker::before_dom_children): Call
26079         gimple_purge_dead_eh_edges. Dump tranformation with details
26080         dump flags.
26081         (strlen_dom_walker::before_dom_children): Update call by adding
26082         new argument cleanup_eh.
26083         (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
26085 2018-01-03  Martin Liska  <mliska@suse.cz>
26087         PR ipa/83549
26088         * cif-code.def (VARIADIC_THUNK): New enum value.
26089         * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
26090         thunks.
26092 2018-01-03  Jan Beulich  <jbeulich@suse.com>
26094         * sse.md (mov<mode>_internal): Tighten condition for when to use
26095         vmovdqu<ssescalarsize> for TI and OI modes.
26097 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
26099         Update copyright years.
26101 2018-01-03  Martin Liska  <mliska@suse.cz>
26103         PR ipa/83594
26104         * ipa-visibility.c (function_and_variable_visibility): Skip
26105         functions with noipa attribure.
26107 2018-01-03  Jakub Jelinek  <jakub@redhat.com>
26109         * gcc.c (process_command): Update copyright notice dates.
26110         * gcov-dump.c (print_version): Ditto.
26111         * gcov.c (print_version): Ditto.
26112         * gcov-tool.c (print_version): Ditto.
26113         * gengtype.c (create_file): Ditto.
26114         * doc/cpp.texi: Bump @copying's copyright year.
26115         * doc/cppinternals.texi: Ditto.
26116         * doc/gcc.texi: Ditto.
26117         * doc/gccint.texi: Ditto.
26118         * doc/gcov.texi: Ditto.
26119         * doc/install.texi: Ditto.
26120         * doc/invoke.texi: Ditto.
26122 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26124         * vector-builder.h (vector_builder::m_full_nelts): Change from
26125         unsigned int to poly_uint64.
26126         (vector_builder::full_nelts): Update prototype accordingly.
26127         (vector_builder::new_vector): Likewise.
26128         (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
26129         (vector_builder::operator ==): Likewise.
26130         (vector_builder::finalize): Likewise.
26131         * int-vector-builder.h (int_vector_builder::int_vector_builder):
26132         Take the number of elements as a poly_uint64 rather than an
26133         unsigned int.
26134         * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
26135         from unsigned int to poly_uint64.
26136         (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
26137         (vec_perm_indices::new_vector): Likewise.
26138         (vec_perm_indices::length): Likewise.
26139         (vec_perm_indices::nelts_per_input): Likewise.
26140         (vec_perm_indices::input_nelts): Likewise.
26141         * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
26142         number of elements per input as a poly_uint64 rather than an
26143         unsigned int.  Use the original encoding for variable-length
26144         vectors, rather than clamping each individual element.
26145         For the second and subsequent elements in each pattern,
26146         clamp the step and base before clamping their sum.
26147         (vec_perm_indices::series_p): Handle polynomial element counts.
26148         (vec_perm_indices::all_in_range_p): Likewise.
26149         (vec_perm_indices_to_tree): Likewise.
26150         (vec_perm_indices_to_rtx): Likewise.
26151         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
26152         * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
26153         (tree_vector_builder::new_binary_operation): Handle polynomial
26154         element counts.  Return false if we need to know the number
26155         of elements at compile time.
26156         * fold-const.c (fold_vec_perm): Punt if the number of elements
26157         isn't known at compile time.
26159 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26161         * vec-perm-indices.h (vec_perm_builder): Change element type
26162         from HOST_WIDE_INT to poly_int64.
26163         (vec_perm_indices::element_type): Update accordingly.
26164         (vec_perm_indices::clamp): Handle polynomial element_types.
26165         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
26166         (vec_perm_indices::all_in_range_p): Likewise.
26167         (tree_to_vec_perm_builder): Check for poly_int64 trees rather
26168         than shwi trees.
26169         * vector-builder.h (vector_builder::stepped_sequence_p): Handle
26170         polynomial vec_perm_indices element types.
26171         * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
26172         * fold-const.c (fold_vec_perm): Likewise.
26173         * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
26174         * tree-vect-generic.c (lower_vec_perm): Likewise.
26175         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
26176         * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
26177         element type to HOST_WIDE_INT.
26179 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26180             Alan Hayward  <alan.hayward@arm.com>
26181             David Sherwood  <david.sherwood@arm.com>
26183         * alias.c (addr_side_effect_eval): Take the size as a poly_int64
26184         rather than an int.  Use plus_constant.
26185         (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
26186         Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
26188 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26189             Alan Hayward  <alan.hayward@arm.com>
26190             David Sherwood  <david.sherwood@arm.com>
26192         * calls.c (emit_call_1, expand_call): Change struct_value_size from
26193         a HOST_WIDE_INT to a poly_int64.
26195 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26196             Alan Hayward  <alan.hayward@arm.com>
26197             David Sherwood  <david.sherwood@arm.com>
26199         * calls.c (load_register_parameters): Cope with polynomial
26200         mode sizes.  Require a constant size for BLKmode parameters
26201         that aren't described by a PARALLEL.  If BLOCK_REG_PADDING
26202         forces a parameter to be padded at the lsb end in order to
26203         fill a complete number of words, require the parameter size
26204         to be ordered wrt UNITS_PER_WORD.
26206 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26207             Alan Hayward  <alan.hayward@arm.com>
26208             David Sherwood  <david.sherwood@arm.com>
26210         * reload1.c (spill_stack_slot_width): Change element type
26211         from unsigned int to poly_uint64_pod.
26212         (alter_reg): Treat mode sizes as polynomial.
26214 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26215             Alan Hayward  <alan.hayward@arm.com>
26216             David Sherwood  <david.sherwood@arm.com>
26218         * reload.c (complex_word_subreg_p): New function.
26219         (reload_inner_reg_of_subreg, push_reload): Use it.
26221 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26222             Alan Hayward  <alan.hayward@arm.com>
26223             David Sherwood  <david.sherwood@arm.com>
26225         * lra-constraints.c (process_alt_operands): Reject matched
26226         operands whose sizes aren't ordered.
26227         (match_reload): Refer to this check here.
26229 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26230             Alan Hayward  <alan.hayward@arm.com>
26231             David Sherwood  <david.sherwood@arm.com>
26233         * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
26234         that the mode size is in the set {1, 2, 4, 8, 16}.
26236 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26237             Alan Hayward  <alan.hayward@arm.com>
26238             David Sherwood  <david.sherwood@arm.com>
26240         * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
26241         Use plus_constant instead of gen_rtx_PLUS.
26243 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26244             Alan Hayward  <alan.hayward@arm.com>
26245             David Sherwood  <david.sherwood@arm.com>
26247         * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
26248         * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
26249         * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
26250         * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
26251         * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
26252         * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
26253         * config/i386/i386-protos.h (ix86_push_rounding): Declare.
26254         * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
26255         * config/i386/i386.c (ix86_push_rounding): ...this new function.
26256         * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
26257         a poly_int64.
26258         * config/m32c/m32c.c (m32c_push_rounding): Likewise.
26259         * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
26260         * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
26261         * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
26262         * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
26263         * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
26264         * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
26265         * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
26266         * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
26267         * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
26268         function.
26269         * expr.c (emit_move_resolve_push): Treat the input and result
26270         of PUSH_ROUNDING as a poly_int64.
26271         (emit_move_complex_push, emit_single_push_insn_1): Likewise.
26272         (emit_push_insn): Likewise.
26273         * lra-eliminations.c (mark_not_eliminable): Likewise.
26274         * recog.c (push_operand): Likewise.
26275         * reload1.c (elimination_effects): Likewise.
26276         * rtlanal.c (nonzero_bits1): Likewise.
26277         * calls.c (store_one_arg): Likewise.  Require the padding to be
26278         known at compile time.
26280 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26281             Alan Hayward  <alan.hayward@arm.com>
26282             David Sherwood  <david.sherwood@arm.com>
26284         * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
26285         Use plus_constant instead of gen_rtx_PLUS.
26287 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26288             Alan Hayward  <alan.hayward@arm.com>
26289             David Sherwood  <david.sherwood@arm.com>
26291         * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
26292         rather than an int.
26294 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26295             Alan Hayward  <alan.hayward@arm.com>
26296             David Sherwood  <david.sherwood@arm.com>
26298         * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
26299         instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
26300         via stack temporaries.  Treat the mode size as polynomial too.
26302 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26303             Alan Hayward  <alan.hayward@arm.com>
26304             David Sherwood  <david.sherwood@arm.com>
26306         * expr.c (expand_expr_real_2): When handling conversions involving
26307         unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
26308         multiplying int_size_in_bytes by BITS_PER_UNIT.  Treat GET_MODE_BISIZE
26309         as a poly_uint64 too.
26311 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26312             Alan Hayward  <alan.hayward@arm.com>
26313             David Sherwood  <david.sherwood@arm.com>
26315         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
26317 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26318             Alan Hayward  <alan.hayward@arm.com>
26319             David Sherwood  <david.sherwood@arm.com>
26321         * combine.c (can_change_dest_mode): Handle polynomial
26322         REGMODE_NATURAL_SIZE.
26323         * expmed.c (store_bit_field_1): Likewise.
26324         * expr.c (store_constructor): Likewise.
26325         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
26326         and polynomial REGMODE_NATURAL_SIZE.
26327         (gen_lowpart_common): Likewise.
26328         * reginfo.c (record_subregs_of_mode): Likewise.
26329         * rtlanal.c (read_modify_subreg_p): Likewise.
26331 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26332             Alan Hayward  <alan.hayward@arm.com>
26333             David Sherwood  <david.sherwood@arm.com>
26335         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
26336         numbers of elements.
26338 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26339             Alan Hayward  <alan.hayward@arm.com>
26340             David Sherwood  <david.sherwood@arm.com>
26342         * match.pd: Cope with polynomial numbers of vector elements.
26344 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26345             Alan Hayward  <alan.hayward@arm.com>
26346             David Sherwood  <david.sherwood@arm.com>
26348         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
26349         in a POINTER_PLUS_EXPR.
26351 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26352             Alan Hayward  <alan.hayward@arm.com>
26353             David Sherwood  <david.sherwood@arm.com>
26355         * omp-simd-clone.c (simd_clone_subparts): New function.
26356         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
26357         (ipa_simd_modify_function_body): Likewise.
26359 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26360             Alan Hayward  <alan.hayward@arm.com>
26361             David Sherwood  <david.sherwood@arm.com>
26363         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
26364         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
26365         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
26366         (expand_vector_condition, vector_element): Likewise.
26367         (subparts_gt): New function.
26368         (get_compute_type): Use subparts_gt.
26369         (count_type_subparts): Delete.
26370         (expand_vector_operations_1): Use subparts_gt instead of
26371         count_type_subparts.
26373 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26374             Alan Hayward  <alan.hayward@arm.com>
26375             David Sherwood  <david.sherwood@arm.com>
26377         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
26378         (vect_compile_time_alias): ...this new function.  Do the calculation
26379         on poly_ints rather than trees.
26380         (vect_prune_runtime_alias_test_list): Update call accordingly.
26382 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26383             Alan Hayward  <alan.hayward@arm.com>
26384             David Sherwood  <david.sherwood@arm.com>
26386         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
26387         numbers of units.
26388         (vect_schedule_slp_instance): Likewise.
26390 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26391             Alan Hayward  <alan.hayward@arm.com>
26392             David Sherwood  <david.sherwood@arm.com>
26394         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
26395         constant and extern definitions for variable-length vectors.
26396         (vect_get_constant_vectors): Note that the number of units
26397         is known to be constant.
26399 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26400             Alan Hayward  <alan.hayward@arm.com>
26401             David Sherwood  <david.sherwood@arm.com>
26403         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
26404         of units as polynomial.  Choose between WIDE and NARROW based
26405         on multiple_p.
26407 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26408             Alan Hayward  <alan.hayward@arm.com>
26409             David Sherwood  <david.sherwood@arm.com>
26411         * tree-vect-stmts.c (simd_clone_subparts): New function.
26412         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
26414 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26415             Alan Hayward  <alan.hayward@arm.com>
26416             David Sherwood  <david.sherwood@arm.com>
26418         * tree-vect-stmts.c (vectorizable_call): Treat the number of
26419         vectors as polynomial.  Use build_index_vector for
26420         IFN_GOMP_SIMD_LANE.
26422 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26423             Alan Hayward  <alan.hayward@arm.com>
26424             David Sherwood  <david.sherwood@arm.com>
26426         * tree-vect-stmts.c (get_load_store_type): Treat the number of
26427         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
26428         for variable-length vectors.
26429         (vectorizable_mask_load_store): Treat the number of units as
26430         polynomial, asserting that it is constant if the condition has
26431         already been enforced.
26432         (vectorizable_store, vectorizable_load): Likewise.
26434 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26435             Alan Hayward  <alan.hayward@arm.com>
26436             David Sherwood  <david.sherwood@arm.com>
26438         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
26439         of units as polynomial.  Punt if we can't tell at compile time
26440         which vector contains the final result.
26442 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26443             Alan Hayward  <alan.hayward@arm.com>
26444             David Sherwood  <david.sherwood@arm.com>
26446         * tree-vect-loop.c (vectorizable_induction): Treat the number
26447         of units as polynomial.  Punt on SLP inductions.  Use an integer
26448         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
26449         cast of such a series for variable-length floating-point
26450         reductions.
26452 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26453             Alan Hayward  <alan.hayward@arm.com>
26454             David Sherwood  <david.sherwood@arm.com>
26456         * tree.h (build_index_vector): Declare.
26457         * tree.c (build_index_vector): New function.
26458         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
26459         of units as polynomial, forcibly converting it to a constant if
26460         vectorizable_reduction has already enforced the condition.
26461         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
26462         to create a {1,2,3,...} vector.
26463         (vectorizable_reduction): Treat the number of units as polynomial.
26464         Choose vectype_in based on the largest scalar element size rather
26465         than the smallest number of units.  Enforce the restrictions
26466         relied on above.
26468 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26469             Alan Hayward  <alan.hayward@arm.com>
26470             David Sherwood  <david.sherwood@arm.com>
26472         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
26473         number of units as polynomial.
26475 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26476             Alan Hayward  <alan.hayward@arm.com>
26477             David Sherwood  <david.sherwood@arm.com>
26479         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
26480         * target.def (autovectorize_vector_sizes): Return the vector sizes
26481         by pointer, using vector_sizes rather than a bitmask.
26482         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
26483         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
26484         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
26485         Likewise.
26486         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
26487         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
26488         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
26489         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
26490         * omp-general.c (omp_max_vf): Likewise.
26491         * omp-low.c (omp_clause_aligned_alignment): Likewise.
26492         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
26493         * tree-vect-loop.c (vect_analyze_loop): Likewise.
26494         * tree-vect-slp.c (vect_slp_bb): Likewise.
26495         * doc/tm.texi: Regenerate.
26496         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
26497         to a poly_uint64.
26498         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
26499         the vector size as a poly_uint64 rather than an unsigned int.
26500         (current_vector_size): Change from an unsigned int to a poly_uint64.
26501         (get_vectype_for_scalar_type): Update accordingly.
26502         * tree.h (build_truth_vector_type): Take the size and number of
26503         units as a poly_uint64 rather than an unsigned int.
26504         (build_vector_type): Add a temporary overload that takes
26505         the number of units as a poly_uint64 rather than an unsigned int.
26506         * tree.c (make_vector_type): Likewise.
26507         (build_truth_vector_type): Take the number of units as a poly_uint64
26508         rather than an unsigned int.
26510 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26511             Alan Hayward  <alan.hayward@arm.com>
26512             David Sherwood  <david.sherwood@arm.com>
26514         * target.def (get_mask_mode): Take the number of units and length
26515         as poly_uint64s rather than unsigned ints.
26516         * targhooks.h (default_get_mask_mode): Update accordingly.
26517         * targhooks.c (default_get_mask_mode): Likewise.
26518         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
26519         * doc/tm.texi: Regenerate.
26521 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26522             Alan Hayward  <alan.hayward@arm.com>
26523             David Sherwood  <david.sherwood@arm.com>
26525         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
26526         * omp-general.c (omp_max_vf): Likewise.
26527         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
26528         (expand_omp_simd): Handle polynomial safelen.
26529         * omp-low.c (omplow_simd_context): Add a default constructor.
26530         (omplow_simd_context::max_vf): Change from int to poly_uint64.
26531         (lower_rec_simd_input_clauses): Update accordingly.
26532         (lower_rec_input_clauses): Likewise.
26534 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26535             Alan Hayward  <alan.hayward@arm.com>
26536             David Sherwood  <david.sherwood@arm.com>
26538         * tree-vectorizer.h (vect_nunits_for_cost): New function.
26539         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
26540         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
26541         (vect_analyze_slp_cost): Likewise.
26542         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
26543         (vect_model_load_cost): Likewise.
26545 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26546             Alan Hayward  <alan.hayward@arm.com>
26547             David Sherwood  <david.sherwood@arm.com>
26549         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
26550         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
26551         from an unsigned int * to a poly_uint64_pod *.
26552         (calculate_unrolling_factor): New function.
26553         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
26555 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26556             Alan Hayward  <alan.hayward@arm.com>
26557             David Sherwood  <david.sherwood@arm.com>
26559         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
26560         from an unsigned int to a poly_uint64.
26561         (_loop_vec_info::slp_unrolling_factor): Likewise.
26562         (_loop_vec_info::vectorization_factor): Change from an int
26563         to a poly_uint64.
26564         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
26565         (vect_get_num_vectors): New function.
26566         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
26567         (vect_get_num_copies): Use vect_get_num_vectors.
26568         (vect_analyze_data_ref_dependences): Change max_vf from an int *
26569         to an unsigned int *.
26570         (vect_analyze_data_refs): Change min_vf from an int * to a
26571         poly_uint64 *.
26572         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
26573         than an unsigned HOST_WIDE_INT.
26574         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
26575         (vect_analyze_data_ref_dependence): Change max_vf from an int *
26576         to an unsigned int *.
26577         (vect_analyze_data_ref_dependences): Likewise.
26578         (vect_compute_data_ref_alignment): Handle polynomial vf.
26579         (vect_enhance_data_refs_alignment): Likewise.
26580         (vect_prune_runtime_alias_test_list): Likewise.
26581         (vect_shift_permute_load_chain): Likewise.
26582         (vect_supportable_dr_alignment): Likewise.
26583         (dependence_distance_ge_vf): Take the vectorization factor as a
26584         poly_uint64 rather than an unsigned HOST_WIDE_INT.
26585         (vect_analyze_data_refs): Change min_vf from an int * to a
26586         poly_uint64 *.
26587         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
26588         vfm1 as a poly_uint64 rather than an int.  Make the same change
26589         for the returned bound_scalar.
26590         (vect_gen_vector_loop_niters): Handle polynomial vf.
26591         (vect_do_peeling): Likewise.  Update call to
26592         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
26593         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
26594         be constant.
26595         * tree-vect-loop.c (vect_determine_vectorization_factor)
26596         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
26597         (vect_get_known_peeling_cost): Likewise.
26598         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
26599         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
26600         (vect_transform_loop): Likewise.  Use the lowest possible VF when
26601         updating the upper bounds of the loop.
26602         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
26603         rather than an int.
26604         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
26605         polynomial unroll factors.
26606         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
26607         (vect_make_slp_decision): Likewise.
26608         (vect_supported_load_permutation_p): Likewise, and polynomial
26609         vf too.
26610         (vect_analyze_slp_cost): Handle polynomial vf.
26611         (vect_slp_analyze_node_operations): Likewise.
26612         (vect_slp_analyze_bb_1): Likewise.
26613         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
26614         than an unsigned HOST_WIDE_INT.
26615         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
26616         (vectorizable_load): Handle polynomial vf.
26617         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
26618         a poly_uint64.
26619         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
26621 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26622             Alan Hayward  <alan.hayward@arm.com>
26623             David Sherwood  <david.sherwood@arm.com>
26625         * match.pd: Handle bit operations involving three constants
26626         and try to fold one pair.
26628 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
26630         * tree-vect-loop-manip.c: Include gimple-fold.h.
26631         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
26632         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
26633         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
26634         Add a path that uses a step of VF instead of 1, but disable it
26635         for now.
26636         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
26637         and niters_no_overflow parameters.  Update calls to
26638         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
26639         Create a new SSA name if the latter choses to use a ste other
26640         than zero, and return it via niters_vector_mult_vf_var.
26641         * tree-vect-loop.c (vect_transform_loop): Update calls to
26642         vect_do_peeling, vect_gen_vector_loop_niters and
26643         slpeel_make_loop_iterate_ntimes.
26644         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
26645         (vect_gen_vector_loop_niters): Update declarations after above changes.
26647 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
26649         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
26650         128-bit round to integer instructions.
26651         (ceil<mode>2): Likewise.
26652         (btrunc<mode>2): Likewise.
26653         (round<mode>2): Likewise.
26655 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26657         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
26658         unaligned VSX load/store on P8/P9.
26659         (expand_block_clear): Allow the use of unaligned VSX
26660         load/store on P8/P9.
26662 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
26664         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
26665         New function.
26666         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
26667         swap associated with both a load and a store.
26669 2018-01-02  Andrew Waterman  <andrew@sifive.com>
26671         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
26672         * config/riscv/riscv.md (clear_cache): Use it.
26674 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
26676         * web.c: Remove out-of-date comment.
26678 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26680         * expr.c (fixup_args_size_notes): Check that any existing
26681         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
26682         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
26683         (emit_single_push_insn): ...here.
26685 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26687         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
26688         (const_vector_encoded_nelts): New function.
26689         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
26690         (const_vector_int_elt, const_vector_elt): Declare.
26691         * emit-rtl.c (const_vector_int_elt_1): New function.
26692         (const_vector_elt): Likewise.
26693         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
26694         of CONST_VECTOR_ELT.
26696 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26698         * expr.c: Include rtx-vector-builder.h.
26699         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
26700         directly on the tree encoding.
26701         (const_vector_from_tree): Likewise.
26702         * optabs.c: Include rtx-vector-builder.h.
26703         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
26704         sequence of "u" values.
26705         * vec-perm-indices.c: Include rtx-vector-builder.h.
26706         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
26707         directly on the vec_perm_indices encoding.
26709 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26711         * doc/rtl.texi (const_vector): Describe new encoding scheme.
26712         * Makefile.in (OBJS): Add rtx-vector-builder.o.
26713         * rtx-vector-builder.h: New file.
26714         * rtx-vector-builder.c: Likewise.
26715         * rtl.h (rtx_def::u2): Add a const_vector field.
26716         (CONST_VECTOR_NPATTERNS): New macro.
26717         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
26718         (CONST_VECTOR_DUPLICATE_P): Likewise.
26719         (CONST_VECTOR_STEPPED_P): Likewise.
26720         (CONST_VECTOR_ENCODED_ELT): Likewise.
26721         (const_vec_duplicate_p): Check for a duplicated vector encoding.
26722         (unwrap_const_vec_duplicate): Likewise.
26723         (const_vec_series_p): Check for a non-duplicated vector encoding.
26724         Say that the function only returns true for integer vectors.
26725         * emit-rtl.c: Include rtx-vector-builder.h.
26726         (gen_const_vec_duplicate_1): Delete.
26727         (gen_const_vector): Call gen_const_vec_duplicate instead of
26728         gen_const_vec_duplicate_1.
26729         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
26730         (gen_const_vec_duplicate): Use rtx_vector_builder.
26731         (gen_const_vec_series): Likewise.
26732         (gen_rtx_CONST_VECTOR): Likewise.
26733         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
26734         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
26735         Build a new vector rather than modifying a CONST_VECTOR in-place.
26736         (handle_special_swappables): Update call accordingly.
26737         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
26738         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
26739         Build a new vector rather than modifying a CONST_VECTOR in-place.
26740         (handle_special_swappables): Update call accordingly.
26742 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26744         * simplify-rtx.c (simplify_const_binary_operation): Use
26745         CONST_VECTOR_ELT instead of XVECEXP.
26747 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26749         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
26750         the selector elements to be different from the data elements
26751         if the selector is a VECTOR_CST.
26752         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
26753         ssizetype for the selector.
26755 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26757         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
26758         before testing each element individually.
26759         * tree-vect-generic.c (lower_vec_perm): Likewise.
26761 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26763         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
26764         * selftest-run-tests.c (selftest::run_tests): Call it.
26765         * vector-builder.h (vector_builder::operator ==): New function.
26766         (vector_builder::operator !=): Likewise.
26767         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
26768         (vec_perm_indices::all_from_input_p): New function.
26769         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
26770         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
26771         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
26772         instead of reading the VECTOR_CST directly.  Detect whether both
26773         vector inputs are the same before constructing the vec_perm_indices,
26774         and update the number of inputs argument accordingly.  Use the
26775         utility functions added above.  Only construct sel2 if we need to.
26777 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26779         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
26780         the broadcast of the low byte.
26781         (expand_mult_highpart): Use an explicit encoding for the permutes.
26782         * optabs-query.c (can_mult_highpart_p): Likewise.
26783         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
26784         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26785         (vectorizable_bswap): Likewise.
26786         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
26787         explicit encoding for the power-of-2 permutes.
26788         (vect_permute_store_chain): Likewise.
26789         (vect_grouped_load_supported): Likewise.
26790         (vect_permute_load_chain): Likewise.
26792 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26794         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
26795         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
26796         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
26797         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
26798         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
26799         (vect_gen_perm_mask_any): Likewise.
26801 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26803         * int-vector-builder.h: New file.
26804         * vec-perm-indices.h: Include int-vector-builder.h.
26805         (vec_perm_indices): Redefine as an int_vector_builder.
26806         (auto_vec_perm_indices): Delete.
26807         (vec_perm_builder): Redefine as a stand-alone class.
26808         (vec_perm_indices::vec_perm_indices): New function.
26809         (vec_perm_indices::clamp): Likewise.
26810         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
26811         (vec_perm_indices::new_vector): New function.
26812         (vec_perm_indices::new_expanded_vector): Update for new
26813         vec_perm_indices class.
26814         (vec_perm_indices::rotate_inputs): New function.
26815         (vec_perm_indices::all_in_range_p): Operate directly on the
26816         encoded form, without computing elided elements.
26817         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
26818         encoding.  Update for new vec_perm_indices class.
26819         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
26820         the given vec_perm_builder.
26821         (expand_vec_perm_var): Update vec_perm_builder constructor.
26822         (expand_mult_highpart): Use vec_perm_builder instead of
26823         auto_vec_perm_indices.
26824         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
26825         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
26826         or double series encoding as appropriate.
26827         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
26828         vec_perm_indices instead of auto_vec_perm_indices.
26829         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
26830         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
26831         (vect_permute_store_chain): Likewise.
26832         (vect_grouped_load_supported): Likewise.
26833         (vect_permute_load_chain): Likewise.
26834         (vect_shift_permute_load_chain): Likewise.
26835         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
26836         (vect_transform_slp_perm_load): Likewise.
26837         (vect_schedule_slp_instance): Likewise.
26838         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
26839         (vectorizable_mask_load_store): Likewise.
26840         (vectorizable_bswap): Likewise.
26841         (vectorizable_store): Likewise.
26842         (vectorizable_load): Likewise.
26843         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
26844         vec_perm_indices instead of auto_vec_perm_indices.  Use
26845         tree_to_vec_perm_builder to read the vector from a tree.
26846         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
26847         vec_perm_builder instead of a vec_perm_indices.
26848         (have_whole_vector_shift): Use vec_perm_builder and
26849         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
26850         truncation to calc_vec_perm_mask_for_shift.
26851         (vect_create_epilog_for_reduction): Likewise.
26852         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
26853         from auto_vec_perm_indices to vec_perm_indices.
26854         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
26855         instead of changing individual elements.
26856         (aarch64_vectorize_vec_perm_const): Use new_vector to install
26857         the vector in d.perm.
26858         * config/arm/arm.c (expand_vec_perm_d::perm): Change
26859         from auto_vec_perm_indices to vec_perm_indices.
26860         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
26861         instead of changing individual elements.
26862         (arm_vectorize_vec_perm_const): Use new_vector to install
26863         the vector in d.perm.
26864         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
26865         Update vec_perm_builder constructor.
26866         (rs6000_expand_interleave): Likewise.
26867         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
26868         (rs6000_expand_interleave): Likewise.
26870 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26872         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
26873         to qimode could truncate the indices.
26874         * optabs.c (expand_vec_perm_var): Likewise.
26876 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
26878         * Makefile.in (OBJS): Add vec-perm-indices.o.
26879         * vec-perm-indices.h: New file.
26880         * vec-perm-indices.c: Likewise.
26881         * target.h (vec_perm_indices): Replace with a forward class
26882         declaration.
26883         (auto_vec_perm_indices): Move to vec-perm-indices.h.
26884         * optabs.h: Include vec-perm-indices.h.
26885         (expand_vec_perm): Delete.
26886         (selector_fits_mode_p, expand_vec_perm_var): Declare.
26887         (expand_vec_perm_const): Declare.
26888         * target.def (vec_perm_const_ok): Replace with...
26889         (vec_perm_const): ...this new hook.
26890         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
26891         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
26892         * doc/tm.texi: Regenerate.
26893         * optabs.def (vec_perm_const): Delete.
26894         * doc/md.texi (vec_perm_const): Likewise.
26895         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
26896         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
26897         expand_vec_perm for constant permutation vectors.  Assert that
26898         the mode of variable permutation vectors is the integer equivalent
26899         of the mode that is being permuted.
26900         * optabs-query.h (selector_fits_mode_p): Declare.
26901         * optabs-query.c: Include vec-perm-indices.h.
26902         (selector_fits_mode_p): New function.
26903         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
26904         is defined, instead of checking whether the vec_perm_const_optab
26905         exists.  Use targetm.vectorize.vec_perm_const instead of
26906         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
26907         fit in the vector mode before using a variable permute.
26908         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
26909         vec_perm_indices instead of an rtx.
26910         (expand_vec_perm): Replace with...
26911         (expand_vec_perm_const): ...this new function.  Take the selector
26912         as a vec_perm_indices rather than an rtx.  Also take the mode of
26913         the selector.  Update call to shift_amt_for_vec_perm_mask.
26914         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
26915         Use vec_perm_indices::new_expanded_vector to expand the original
26916         selector into bytes.  Check whether the indices fit in the vector
26917         mode before using a variable permute.
26918         (expand_vec_perm_var): Make global.
26919         (expand_mult_highpart): Use expand_vec_perm_const.
26920         * fold-const.c: Includes vec-perm-indices.h.
26921         * tree-ssa-forwprop.c: Likewise.
26922         * tree-vect-data-refs.c: Likewise.
26923         * tree-vect-generic.c: Likewise.
26924         * tree-vect-loop.c: Likewise.
26925         * tree-vect-slp.c: Likewise.
26926         * tree-vect-stmts.c: Likewise.
26927         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
26928         Delete.
26929         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
26930         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
26931         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
26932         (aarch64_vectorize_vec_perm_const): ...this new function.
26933         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26934         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26935         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
26936         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
26937         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26938         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26939         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
26940         into...
26941         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
26942         check for NEON modes.
26943         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
26944         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
26945         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
26946         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
26947         into...
26948         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
26949         the old VEC_PERM_CONST conditions.
26950         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
26951         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
26952         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
26953         (ia64_vectorize_vec_perm_const_ok): Merge into...
26954         (ia64_vectorize_vec_perm_const): ...this new function.
26955         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
26956         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
26957         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
26958         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
26959         * config/mips/mips.c (mips_expand_vec_perm_const)
26960         (mips_vectorize_vec_perm_const_ok): Merge into...
26961         (mips_vectorize_vec_perm_const): ...this new function.
26962         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
26963         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
26964         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
26965         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
26966         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
26967         (rs6000_expand_vec_perm_const): Delete.
26968         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
26969         Delete.
26970         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26971         (altivec_expand_vec_perm_const_le): Take each operand individually.
26972         Operate on constant selectors rather than rtxes.
26973         (altivec_expand_vec_perm_const): Likewise.  Update call to
26974         altivec_expand_vec_perm_const_le.
26975         (rs6000_expand_vec_perm_const): Delete.
26976         (rs6000_vectorize_vec_perm_const_ok): Delete.
26977         (rs6000_vectorize_vec_perm_const): New function.
26978         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26979         an element count and rtx array.
26980         (rs6000_expand_extract_even): Update call accordingly.
26981         (rs6000_expand_interleave): Likewise.
26982         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
26983         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
26984         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
26985         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
26986         (rs6000_expand_vec_perm_const): Delete.
26987         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
26988         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
26989         (altivec_expand_vec_perm_const_le): Take each operand individually.
26990         Operate on constant selectors rather than rtxes.
26991         (altivec_expand_vec_perm_const): Likewise.  Update call to
26992         altivec_expand_vec_perm_const_le.
26993         (rs6000_expand_vec_perm_const): Delete.
26994         (rs6000_vectorize_vec_perm_const_ok): Delete.
26995         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
26996         reference to the SPE evmerge intructions.
26997         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
26998         an element count and rtx array.
26999         (rs6000_expand_extract_even): Update call accordingly.
27000         (rs6000_expand_interleave): Likewise.
27001         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
27002         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
27003         new function.
27004         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
27006 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27008         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
27009         vector mode and that that mode matches the mode of the data
27010         being permuted.
27011         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
27012         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
27013         directly using expand_vec_perm_1 when forcing selectors into
27014         registers.
27015         (expand_vec_perm_var): New function, split out from expand_vec_perm.
27017 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27019         * optabs-query.h (can_vec_perm_p): Delete.
27020         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
27021         * optabs-query.c (can_vec_perm_p): Split into...
27022         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
27023         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
27024         particular selector is valid.
27025         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
27026         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
27027         (vect_grouped_load_supported): Likewise.
27028         (vect_shift_permute_load_chain): Likewise.
27029         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
27030         (vect_transform_slp_perm_load): Likewise.
27031         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
27032         (vectorizable_bswap): Likewise.
27033         (vect_gen_perm_mask_checked): Likewise.
27034         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
27035         implementations of variable permutation vectors into account
27036         when deciding which selector to use.
27037         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
27038         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
27039         with a false third argument.
27040         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
27041         to test whether the constant selector is valid and can_vec_perm_var_p
27042         to test whether a variable selector is valid.
27044 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27046         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
27047         * optabs-query.c (can_vec_perm_p): Likewise.
27048         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
27049         instead of vec_perm_indices.
27050         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
27051         (vect_gen_perm_mask_checked): Likewise,
27052         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
27053         (vect_gen_perm_mask_checked): Likewise,
27055 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
27057         * optabs-query.h (qimode_for_vec_perm): Declare.
27058         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
27059         (qimode_for_vec_perm): ...this new function.
27060         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
27062 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27064         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
27065         does not have a conditional at the top.
27067 2018-01-02  Richard Biener  <rguenther@suse.de>
27069         * ipa-inline.c (big_speedup_p): Fix expression.
27071 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
27073         PR target/81616
27074         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
27075         for generic 4->6.
27077 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
27079         PR target/81616
27080         Generic tuning.
27081         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
27082         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
27083         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
27084         cond_taken_branch_cost 3->4.
27086 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
27088         PR tree-optimization/83581
27089         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
27090         TODO_cleanup_cfg if any changes have been made.
27092         PR middle-end/83608
27093         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
27094         convert_modes if target mode has the right side, but different mode
27095         class.
27097         PR middle-end/83609
27098         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
27099         last argument when extracting from CONCAT.  If either from_real or
27100         from_imag is NULL, use expansion through memory.  If result is not
27101         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
27102         the parts directly to inner mode, if even that fails, use expansion
27103         through memory.
27105         PR middle-end/83623
27106         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
27107         check for bswap in mode rather than HImode and use that in expand_unop
27108         too.
27110 Copyright (C) 2018 Free Software Foundation, Inc.
27112 Copying and distribution of this file, with or without modification,
27113 are permitted in any medium without royalty provided the copyright
27114 notice and this notice are preserved.